WO2024179127A1

WO2024179127A1 - Navigation route recommendation method and apparatus, familiar-route prediction model training method and apparatus, and medium

Info

Publication number: WO2024179127A1
Application number: PCT/CN2023/138523
Authority: WO
Inventors: 王岳; 徐龙飞; 陈超
Original assignee: 北京高德云图科技有限公司
Priority date: 2023-02-28
Filing date: 2023-12-13
Publication date: 2024-09-06
Also published as: CN116358582A

Abstract

A navigation route recommendation method (100) and apparatus, a familiar-route prediction model training method (200) and apparatus, and a medium. The navigation route recommendation method (100) comprises: acquiring historical travel data of a link comprised in a candidate navigation route (S101); constructing prediction data of the link on the basis of the historical travel data of the link and route planning data of the candidate navigation route, the prediction data comprising planning feature data of the link (S102); and determining a recommendation prediction value of the candidate navigation route on the basis of the prediction data of the link comprised in the candidate navigation route (S103). In this way, a travel preference change of a subject for whom navigation is to be performed can be captured in a timely manner, and a navigation route meeting the preference of said subject can be recommended to said subject more accurately, thereby better realizing personalized route recommendation.

Description

Navigation route recommendation, familiar route prediction model training method, device, and medium

CROSS-REFERENCE TO RELATED APPLICATIONS

This disclosure claims priority to Chinese patent application filed on February 28, 2023 and application number 202310214491.9, the entire contents of which are incorporated herein by reference.

Technical Field

The present disclosure relates to the field of navigation technology, and in particular to a method, device, and medium for recommending a navigation route and training a familiar road prediction model.

Background Art

Existing applications (software) and their service systems that support map navigation functions can calculate and recommend navigation routes for users based on the user's input of starting point, end point and other information.

In order to make the navigation route recommended to the user meet the user's travel preferences (such as liking to take familiar roads, liking roads with fewer traffic lights, etc.), the prior art generally determines the navigation route recommended to the user based on the relevant data representing the user's travel preferences. However, the inventors found that the user's travel preferences will change with time or changes in the environment, etc. When the relevant data representing the user's travel preferences cannot reflect the changes in the user's travel preferences, the navigation route recommended to the user will not meet the user's expectations, and the user experience will deteriorate.

Summary of the invention

The embodiments of the present disclosure provide a method, device, and medium for recommending a navigation route and training a familiar route prediction model.

In a first aspect, an embodiment of the present disclosure provides a method for recommending a navigation route, comprising:

Obtain candidate navigation routes including historical travel data of road segments;

Constructing prediction data of the road section based on historical travel data of the road section and route planning data of the candidate navigation route, the prediction data including planning feature data of the road section;

Based on the prediction data of the candidate navigation route including the road segments, a recommendation prediction value of the candidate navigation route is determined.

In a second aspect, the present disclosure provides a method for training a familiar road prediction model, including:

Obtaining sample historical travel data of a sample navigation route including a sample road section;

Constructing a training sample of the sample road section according to the sample historical travel data of the sample road section and the sample route planning data of the sample navigation route, wherein the training sample includes the sample planning feature data of the sample road section;

The familiar road prediction model is trained using training samples of the sample road segments, and the familiar road prediction model is used to determine the recommended prediction value of the candidate navigation route based on the prediction data of the candidate navigation route including the road segments.

In a third aspect, an embodiment of the present disclosure provides a navigation route recommendation device, including:

A first acquisition module is used to acquire historical travel data of candidate navigation routes including road sections;

A first construction module is used to construct prediction data of the road section according to historical travel data of the road section and route planning data of the candidate navigation route, wherein the prediction data includes planning feature data of the road section;

The determination module is used to determine the recommended prediction value of the candidate navigation route based on the prediction data of the candidate navigation route including the road segment.

In a fourth aspect, an embodiment of the present disclosure provides a training device for a familiar road prediction model, comprising:

A second acquisition module is used to acquire sample historical travel data of a sample navigation route including a sample road section;

A second construction module is used to construct a training sample of the sample road section according to the sample historical travel data of the sample road section and the sample route planning data of the sample navigation route, wherein the training sample includes the sample planning feature data of the sample road section;

The training module is used to train the familiar road prediction model using training samples of the sample road segments, and the familiar road prediction model is used to determine the recommended prediction value of the candidate navigation route based on the prediction data of the candidate navigation route including the road segments.

In a fifth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the method provided by any embodiment of the present disclosure is implemented.

The technical solution disclosed in the present invention obtains the historical travel data of the candidate navigation route including the road segment; constructs the prediction data of the road segment according to the historical travel data of the road segment and the route planning data of the candidate navigation route, and the prediction data includes the planning feature data of the road segment; and determines the recommended prediction value of the candidate navigation route based on the prediction data of the candidate navigation route including the road segment. Since the historical travel data in the present invention is the data generated by the actual travel behavior of the navigated object, the historical travel data represents the actual travel preference of the navigated object for the corresponding road segment. When the travel preference of the navigated object changes with time or environment, the historical travel data can easily capture the change in the preference of the navigated object for the corresponding road segment. Furthermore, since the prediction data of the road segment constructed by the present invention includes the planning feature data of the road segment based on the historical travel data and the route planning data of the candidate navigation routes, that is, the prediction data of the road segment constructed by the present invention associates the features of the road segment included in the candidate navigation route when it is planned with the features of the road segment recorded in the historical travel data, so that the changes in the travel preferences of the navigated object are reflected in the prediction data of the candidate navigation route, thereby improving the accuracy of the recommended prediction value of the candidate route, so that the navigation route recommended to the navigated object is more in line with the current travel preferences of the navigated object, and the navigated object is more likely to choose the recommended navigation route.

The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present disclosure will be readily apparent by reference to the accompanying drawings and the following detailed description.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or related technologies, the following briefly introduces the drawings required for use in the exemplary embodiments or related technical descriptions. Obviously, the drawings described below are some exemplary embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG. 1 is a schematic flow chart of a method for recommending a navigation route in an embodiment of the present disclosure.

FIG2 is a flowchart of a method for training a familiar road prediction model in an embodiment of the present disclosure.

FIG. 3 is a structural block diagram of a navigation route recommendation device in an embodiment of the present disclosure.

FIG4 is a structural block diagram of a training device for a familiar road prediction model in an embodiment of the present disclosure.

FIG. 5 is a block diagram of an electronic device for implementing an embodiment of the present disclosure.

DETAILED DESCRIPTION

In the following, only some exemplary embodiments are briefly described. As can be appreciated by those skilled in the art, The described embodiments may be modified in various different ways without departing from the spirit or scope of the present disclosure.Accordingly, the drawings and description are to be regarded as illustrative in nature and not restrictive.

To facilitate understanding of the technical solutions of the embodiments of the present disclosure, the related technologies of the embodiments of the present disclosure are described below. The following related technologies can be arbitrarily combined with the technical solutions of the embodiments of the present disclosure as optional solutions, and they all belong to the protection scope of the embodiments of the present disclosure.

The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

When using an application with a map navigation function for navigation, the user (also referred to as the navigated object) inputs the starting point and the end point, and the application and its related service system will plan the optimal navigation route and alternative navigation routes under different strategies for the user. In addition, in order to make the navigation route recommended to the user meet the user's expectations, the existing technology also combines the user's travel preferences to recommend a navigation route that meets the user's expectations, for example, recommending a navigation route consisting of a road section that the user is familiar with.

Some concepts involved in this disclosure include road segments, personal road networks, etc.

Link: It is the smallest unit of a road. It can be understood that a road in the real world will be divided into multiple links in an electronic map. Usually, a section of road between two intersections will be divided into one link. Since roads are divided into links in electronic maps, the navigation route planned based on the electronic map usually includes more than one link.

Personal road network: Based on the road information that the user (also called the navigation object) actually traveled (drove or walked, etc.) within a preset historical time (for example, within half a year), the personal road network is recorded according to road sections.

In the related art, a personal road network represents a preferred or familiar road section for a user when traveling. When determining a personal road network, the prior art usually mines it using predefined rules. For example, a road section that a user has walked twice or more is a preferred/familiar road section for the user. However, there are some problems with the personal road network mined using predefined rules. For example, a road section that a user has walked twice or more is defined as a familiar road section for the user. As time goes by or the environment changes, the road section may no longer be a road section that the user often walks on. In other words, as time goes by, the road section is no longer a road section that the user is familiar with or prefers to travel on. This will result in the navigation route containing the road section not being the navigation route expected by the user. Especially in the familiar road mode, the navigation route expected by the user, but because the mining cycle of the personal road network is usually long, when the user's previous frequently traveled section is no longer the current frequently traveled section due to changes in work location or relocation or some special needs, the personal road network will still mark this section as a familiar section for the user due to untimely updates, which will cause the predicted value of the navigation route containing this section to usually exceed other navigation routes, and ultimately the navigation route recommended to the user does not actually meet the user's current expectations.

Therefore, it is necessary to provide a new navigation route recommendation technology that can timely capture changes in user travel preferences and recommend navigation routes that meet user expectations.

Fig. 1 is a flowchart of a method for recommending a navigation route in an embodiment of the present disclosure. As shown in Fig. 1, a method 100 for recommending a navigation route provided in an embodiment of the present disclosure includes steps S101 to S103.

In step S101, historical travel data of the candidate navigation route including the road segments is obtained.

The number of candidate navigation routes may be one or more, and the candidate navigation routes may include several topologically connected road segments. The historical travel data of the road segment may include historical feature data characterizing the historical travel preferences of the navigated object. The historical travel data acquired in step S101 may be the historical travel data of the road segment within a preset historical time range, for example, within six months or three months from the current time. The preset historical time range may be set as needed, and the present disclosure does not impose any restrictions.

The historical travel data corresponds to the navigated object. For example, the historical travel data of the corresponding road section of the navigated object can be obtained based on the identifier of the navigated object. The historical travel data records the historical feature data generated when the navigated object actually walked through the corresponding road section before the route planning time of the candidate navigation route. Since the navigated object actually walked through the road section before and the historical travel data is relatively close to the current time, the historical feature data used in the present disclosure can characterize the recent changes in the historical travel preferences of the navigated object for the road section.

The road sections in the candidate navigation routes may have historical travel data or may not have historical travel data. For example, if the navigated object passes through the road section within a preset historical time range, then the road section has corresponding historical travel data. If the navigated object does not pass through the road section within the preset historical time range, then the road section has no historical travel data. The present disclosure is aimed at the road sections included in the candidate navigation routes for which historical travel data can be obtained. For the calculation of the weights of the road sections for which historical travel data cannot be obtained, relevant existing technologies can be used, and the present disclosure does not impose any restrictions.

In step S102, prediction data of the road segment is constructed based on the historical travel data of the road segment and the route planning data of the candidate navigation route, and the prediction data includes planning feature data of the road segment.

The route planning data in step S102 can be understood as the data that an application with a map navigation function and its service system need to use when planning a candidate navigation route, including but not limited to the starting point, end point, road conditions, etc. The prediction data of the road section is constructed based on the historical travel data of the road section and the route planning data of the candidate navigation route. The prediction data includes planning feature data of the road section, which associates the features of the road section included in the candidate navigation route when it is planned with the features recorded in the historical travel data of the road section, so that the travel preference changes of the navigated object are reflected in the prediction data of the candidate navigation route.

In step S103, based on the prediction data of the candidate navigation route including the road segments, a recommended prediction value of the candidate navigation route is determined.

Among them, the recommendation prediction value refers to the score value of the candidate navigation route recommended to the navigated object. The larger the recommendation prediction value, the higher the probability that the candidate navigation route is recommended to the navigated object. When the recommendation scenario is "familiar road mode", the recommendation prediction value can be understood as the familiarity of the navigated object with the candidate navigation route. The higher the recommendation prediction value, the more familiar the navigated object is with the candidate navigation route. For the navigated object, especially the navigated object traveling by car, driving on a familiar road will make them more aware of the road conditions and feel more at ease. Therefore, recommending a familiar navigation route to the navigated object will reduce the unfamiliarity of the navigated object with the navigation route and be more in line with the expectations of the navigated object. The navigated object can also improve the safety of driving by choosing a familiar navigation route.

The above is the method provided by an embodiment of the present disclosure, which determines the recommended prediction value of the candidate navigation routes. After determining the recommended prediction value of each candidate navigation route, the method of the present disclosure can compare the recommended prediction value of the candidate navigation route with a preset threshold, and select the candidate navigation route recommended to the navigated object from the candidate navigation routes whose recommended prediction value is greater than the preset threshold. For example, when the number of candidate navigation routes whose recommended prediction value is greater than or equal to the preset threshold is at least two, these candidate navigation routes can be sorted according to the size of the recommended prediction value, and one or more of the candidate navigation routes can be recommended to the navigated object as the target navigation route based on the sorting. The number of candidate navigation routes recommended to the navigated object can be set as needed, This disclosure does not impose any limitations.

In the related art, when determining a personal road network, mining is usually performed using predefined rules, for example, a road section that the navigation object has walked twice or more is a road section that the navigation object prefers/is familiar with. However, there are some problems with the personal road network mined using predefined rules. For example, a road section that the navigation object has walked twice or more is defined as a road section that the navigation object is familiar with. However, as time goes by or the environment changes, the road section may no longer be a road section that the navigation object often walks on. In other words, as time goes by, the road section is no longer a road section that the navigation object is familiar with or prefers to travel on, which may result in the navigation route containing the road section not being the navigation route expected by the navigation object. In particular, in the familiar road mode, due to the long mining cycle of the personal road network, the personal road network will still mark this road section as a road section that the navigation object is familiar with due to untimely updates, which will result in the predicted value of the navigation route containing the road section usually exceeding other navigation routes, resulting in the predicted value of the navigation route not being able to accurately reflect the travel preferences of the navigation object, and ultimately making the navigation route recommended to the navigation object actually not meet the current expectations of the navigation object.

The technical solution disclosed in the present invention obtains the historical travel data of the candidate navigation route including the road segment; constructs the prediction data of the road segment according to the historical travel data of the road segment and the route planning data of the candidate navigation route, and the prediction data includes the planning feature data of the road segment; and determines the recommended prediction value of the candidate navigation route based on the prediction data of the candidate navigation route including the road segment. Since the historical travel data in the present invention is the data generated by the actual travel behavior of the navigated object, the historical travel data represents the actual travel preference of the navigated object for the corresponding road segment. When the travel preference of the navigated object changes with time or environment, the historical travel data can easily capture the change in the preference of the navigated object for the corresponding road segment. Furthermore, since the prediction data of the road segment constructed by the present disclosure includes the planning feature data of the road segment based on the historical travel data and the route planning data of the candidate navigation route, that is, the prediction data of the road segment constructed by the present disclosure associates the features of the road segment included in the candidate navigation route when it was planned with the features of the road segment recorded in the historical travel data, so that the travel preference changes of the navigated object are reflected in the prediction data of the candidate navigation route, thereby improving the accuracy of the recommended prediction value of the candidate route, making the navigation route recommended to the navigated object more in line with the current travel preference of the navigated object, and the navigated object is more likely to choose the recommended navigation route. Therefore, by adopting the technical solution of the present disclosure, the travel preference changes of the navigated object can be captured in time, and the navigation route that meets its preferences can be recommended to the navigated object more accurately, so as to better realize personalized route recommendation.

In one embodiment, historical travel data may include historical travel time, which can be understood as the historical time when the navigated object passes through the road section. Route planning data may include route planning time. Route planning time should be understood as the time when an application with navigation function and its service system plan the navigation route according to the starting point and end point input by the navigated object. The planning feature data of the road section includes time feature and distance feature.

Based on the historical travel data of the road segment and the route planning data of the candidate navigation route, the prediction data of the road segment is constructed, which may include:

Determine the time feature corresponding to the historical travel time according to the historical travel time included in the historical travel data of the road section and the route planning time included in the route planning data of the candidate navigation route;

Determine the distance feature corresponding to the historical travel time based on the historical starting point and historical end point corresponding to the historical travel time in the historical travel data of the road section, and the planned starting point and planned end point included in the route planning data of the candidate navigation route; and construct the prediction data of the road section based on at least the time feature and the distance feature.

It is understandable that if the historical travel time of a road section is far away from the route planning time, it means that the navigation object The time from the time when the candidate navigation route was planned to the time when the person walked the road section before is too far away, and the familiarity of the person being navigated with the road section decays with time, that is, the farther the difference between the historical travel time and the route planning time is, the less familiar the person being navigated with the road section is. On the contrary, if the difference between the historical travel time and the route planning time is very close, the more familiar the person being navigated with the road section is. It can be seen that the present disclosure determines the time feature corresponding to the historical travel time based on the historical travel time of the road section and the route planning time of the candidate navigation route, and takes into account the problem of the decay of familiarity with time, so that the constructed prediction data can better reflect the preference or familiarity of the person being navigated with the corresponding road section.

The historical starting point and the historical end point may be the starting point and the end point of the historical travel route corresponding to the historical travel time. The historical travel route is the navigation route corresponding to the navigation object's historical passage through the road section. For example, the navigation route planned by the navigation object from starting point E to end point F 5 days before the route planning time passes through road section 1, and the navigation object actually also passes through road section 1. Then, the historical starting point and the historical end point corresponding to the historical travel time are the starting point and the end point corresponding to the navigation route from starting point E to end point F. The historical travel time may be the specific date of passing the historical travel route, such as November 20, 2022, or the historical travel time may be the number of days of passing the historical travel route relative to the route planning time, such as 5 days ago. The specific form of the historical travel time can be set as needed. Continuing with the above example, the navigation object passed through road section 1 5 days before the candidate navigation route was planned, and the historical travel route through road section 1 was from starting point E to end point F. Then, for road section 1, the historical feature data may include 5 days ago, starting point E and end point F, and the historical travel data may be expressed as {5 days ago, starting point E, end point F}. Each time the navigated object passes through section 1, a historical feature data corresponding to section 1 is generated. The more times the navigated object passes through section 1, the more historical feature data corresponding to section 1. The planned starting point and the planned end point are the starting point and the end point of the planned candidate navigation route.

In the related art, for the familiar road section of the navigated object, the starting point and the end point of the navigation route are not considered. As long as the road section exists in the personal road network, the probability of recommending the navigation route containing the road section to the navigated object will become higher. In this embodiment, the historical travel data includes the historical starting point and the historical end point corresponding to the historical travel time, and the route planning data includes the planned starting point and the planned end point. It can be understood that if the distance between the historical starting point and the planned starting point of the navigation route when the navigated object has walked through a certain road section in the past is very different, and the distance between the historical end point and the planned end point is very different, it means that even if the navigated object has walked through the road section, the navigated object is not familiar with the road section when walking from the planned starting point to the planned end point, or does not prefer the road section. On the contrary, if the distance between the historical starting point and the planned starting point of the navigation route when the navigated object has walked through a certain road section in the past is not much different, and the distance between the historical end point and the planned end point is not much different, it means that the navigated object will prefer to walk on the road section in the process of walking from the planned starting point to the planned end point.

In the above-mentioned embodiment of the present disclosure, the time feature corresponding to the historical travel time is determined according to the historical travel time and the route planning time; the distance feature corresponding to the historical travel time is determined according to the historical starting point, the historical end point, the planned starting point, and the planned end point, and the prediction data of the road section is constructed based on at least the time feature and the distance feature. It can be seen that the prediction data of the road section constructed by the present disclosure not only considers the influence of time on travel preference, but also considers the influence of the starting point and the end point on travel preference, which can further improve the accuracy of the recommended prediction value of the candidate navigation route, so that the recommended prediction value can more accurately reflect the current preference of the navigated object for the candidate navigation route.

It should be noted that the starting point in the present disclosure may be the position coordinates of the starting point, and the end point may be the position coordinates of the end point. When the distance between two starting points is less than or equal to a preset distance, the two starting points may be considered the same; when the distance between two end points is less than or equal to a preset distance, the two end points may be considered the same. The preset distance may be, for example, 100 m.

In one embodiment, the time feature may include the time difference between the historical travel time and the route planning time.

In one embodiment, the distance feature may include a first distance feature and a second distance feature. According to the historical starting point and the historical end point corresponding to the historical travel time in the historical travel data of the road segment, and the planned starting point and the planned end point included in the route planning data of the candidate navigation route, the distance feature corresponding to the historical travel time is determined, including: determining the distance between the historical starting point and the planned starting point as the first distance feature; and determining the distance between the historical end point and the planned end point as the second distance feature.

In the prediction data of a road section, the time difference between the historical travel time of the road section and the route planning time can be the difference in days between the historical travel time and the route planning time. The time difference can represent the difference between the historical travel time of the navigated object passing through the road section and the route planning time. The smaller the time difference, the closer the time when the navigated object passed through the road section to the route planning time, the more familiar the navigated object is with the road section, or the more the navigated object prefers the road section. Therefore, the prediction data of the road section that includes the time difference between the historical travel time and the route planning time of the road section takes into account the impact of the time difference on travel preferences, so that the recommended prediction value of the determined candidate navigation route more accurately reflects the current preference of the navigated object for the candidate navigation route.

The recommended prediction value of the candidate navigation route is related to the first distance feature and the second distance feature. Generally, the smaller the values corresponding to the first distance feature and the second distance feature, the greater the possibility that the navigated object will select the road section. Therefore, the predicted data of the road section includes the first distance feature and the second distance feature. The recommended prediction value of the candidate navigation route determined based on the predicted data of the road section can better reflect the familiarity of the navigated object with the road section, better reflect the current preference of the navigated object for the candidate navigation route, and meet the expectations of the navigated object.

Since the road condition of a section determines the travel time cost of a navigation route, usually, the navigated object will not actively choose a frequently congested section. In order to reflect the navigated object's preference for the section due to the road condition in the process of recommending the predicted value, and further improve the accuracy of the recommended predicted value of the determined candidate navigation route, in one embodiment of the present disclosure, the planning feature data of the section may also include road condition features. The method of the embodiment of the present disclosure may also include:

The road condition characteristics corresponding to the historical travel time are determined based on the historical road condition information corresponding to the historical travel time in the historical travel data of the road section and the planned road condition information of the road section included in the route planning data.

For example, the traffic condition information may include smooth, slow, congested, extremely congested, no traffic condition or under construction. For example, if the historical traffic condition information is smooth and the planned traffic condition information is congested, then the navigation object's current preference for the candidate navigation route is low, and the navigation object is less likely to currently select the candidate navigation route. It should be noted that no traffic condition means that the traffic condition of the road section cannot be known.

In one embodiment, the traffic condition feature may include the difference between the historical traffic condition information corresponding to the historical travel time and the planned traffic condition information of the road section. For example, a value is assigned to each type of traffic condition information, such as 1 for smooth traffic, 0.8 for slow traffic, 0.6 for congestion, 0.4 for extreme congestion, 0.2 for no traffic condition, and 0 for under construction. The difference between the historical traffic condition information and the planned traffic condition information may be determined as the traffic condition feature.

In one embodiment, the traffic condition characteristics may include historical traffic condition information corresponding to historical travel times and planned traffic condition information of road sections included in the route planning data.

It should be noted that the historical traffic information corresponds to the planned traffic information. When the historical traffic information corresponding to the historical travel time is the historical traffic information of the road section, the planned traffic information is the traffic information of the road section in the candidate navigation route at the planned time; when the historical traffic information corresponding to the historical travel time is the overall historical traffic information of the historical travel route, the planned route The traffic information is the overall traffic information of the candidate navigation route at the time of planning.

Furthermore, since the longer the road segment that the navigation object is familiar with accounts for in the candidate navigation route, the more the navigation object prefers the navigation route, in one embodiment of the present disclosure, the planning feature data of the road segment also includes a road segment proportion feature. The method of the embodiment of the present disclosure may also include: determining the road segment proportion feature according to the length proportion of the road segment in the candidate navigation route.

The road segment proportion feature can reflect the preference of the navigated object for the navigation route. When the prediction data of the road segment includes the road segment proportion feature, the recommended prediction value of the candidate navigation route determined based on the candidate navigation route including the prediction data of the road segment can further improve the accuracy of the recommended prediction value, and can better reflect the current preference of the navigated object for the candidate navigation route.

In the disclosed embodiment, the prediction data of the road segment is constructed based on at least the time feature and the distance feature, including:

The time characteristics, distance characteristics, road condition characteristics and road section proportion characteristics corresponding to the same historical travel time are constructed into a prediction data of the road section.

It is understandable that each time the navigated object passes through the road section, a corresponding historical travel data will be generated, each historical travel data corresponds to a historical travel time, and each historical travel time corresponds to a prediction data. A prediction data of a road section can be expressed as {time feature, distance feature, road condition feature, road section proportion feature}, and a prediction data can also be expressed as {time feature, first distance feature, second distance feature, historical road condition information, planned road condition information, road section proportion feature}.

In one embodiment, based on the prediction data of the candidate navigation route including the road segments, the recommended prediction value of the candidate navigation route is determined, including: inputting the prediction data of the candidate navigation route including the road segments into a trained familiar road prediction model to obtain the model prediction value corresponding to the prediction data of the road segments; determining the road segment prediction value of the road segment according to the model prediction value corresponding to the prediction data of the road segment; determining the recommended prediction value of the candidate navigation route according to the road segment prediction value of the candidate navigation route including the road segments.

Each historical travel time of a road section corresponds to a piece of prediction data of the road section. By inputting the prediction data of the road section into the trained familiar road prediction model, the model prediction value corresponding to the prediction data of the road section can be obtained.

The familiar route prediction model can be a neural network model, which can be a single-layer neural network or a multi-layer neural network. Training the familiar route prediction model allows the model to learn the travel preferences of the navigated object at different times and different starting and ending points. Since the prediction data of the road section is based on the historical travel data and the route planning data of the candidate navigation route, the constructed prediction data of the road section includes the planning feature data of the road section. The historical travel data does not need to be specially mined, and the travel of the navigated object will form travel data. Therefore, when the model is trained, the prediction data of the road section is input into the trained familiar route prediction model. The model prediction value output by the familiar route prediction model can better reflect the current preference of the navigated object for the road section, avoiding the problem that the personal road network mined by the pre-defined rules cannot timely capture the changes in the travel preferences of the navigated object, resulting in inaccurate recommended prediction values.

In one embodiment, the candidate navigation route may include multiple sections, each section may correspond to at least one historical travel data, and thus each section may correspond to at least one prediction data. Therefore, the recommended prediction value of the candidate navigation route may be determined based on the prediction data corresponding to each section in the candidate navigation route.

Determining the section prediction value of the road section according to the model prediction value corresponding to the prediction data of the road section includes: performing weighted calculation on the model prediction value corresponding to the prediction data of the road section to obtain the section prediction value of the road section.

Each historical travel time of a road section corresponds to a prediction data of the road section. When the navigation object passes through the road section multiple times, the road section has multiple historical travel times, and then multiple prediction data of the road section can be constructed. Correspondingly, the road section corresponds to multiple model prediction values. The model prediction values corresponding to the road section can be weighted to obtain the road section prediction value of the road section.

For example, based at least on the time feature in the prediction data, the weight value of the model prediction value corresponding to the prediction data is determined, and the product of the model prediction value and the weight value is calculated. The products corresponding to the model prediction values are added together to obtain the section prediction value of the road section.

For example, the model prediction value of the prediction data of a road section is y. Assume that the candidate navigation route A includes road sections 1, 2 and 3, and the prediction data of the three roads corresponding to road section 1, and the model prediction values of the prediction data of the three roads are y1, y2 and y3 respectively. Assume that the weight values of y1, y2 and y3 are determined to be α1, α2 and α3 respectively according to the time characteristics in the prediction data, then the road section prediction value M1 of road section 1 is M1 = y1*α1+y2*α2+y3*α3. It should be noted that when determining the weight value of the model prediction value corresponding to the prediction data, it is not limited to the time characteristics in the prediction data, and the weight value of the model prediction value can also be determined based on distance characteristics, etc.

In one embodiment, the weight values of the model prediction values may all be set to 1, and the segment prediction value of segment 1 M1 = y1 + y2 + y3.

The predicted value of the road section determined by the embodiment of the present disclosure integrates the historical travel data of each time the navigated object passes through the road section, and can better reflect the navigated object's preference for the road section. It can be understood that the more times the navigated object passes through the road section, the more the navigated object prefers the road section, and thus, the predicted values of each model are weighted and calculated. The larger the calculated value, the larger the predicted value of the road section, the stronger the navigated object's preference for the road section, and the more familiar the navigated object is with the road section.

In one embodiment, determining the recommended prediction value of the candidate navigation route based on the segment prediction values of the segments included in the candidate navigation route includes: performing weighted calculation on the segment prediction values of the segments included in the candidate navigation route to obtain the recommended prediction value of the candidate navigation route.

For example, the weight value of the predicted value of the road segment may be determined based at least on the length ratio of the road segment in the candidate navigation route, the product of the predicted value of the road segment and the weight value may be calculated, and the products corresponding to the predicted values of the road segments may be added to obtain the recommended predicted value of the candidate navigation route.

For example, the predicted value of segment 2 is calculated to be M2. There is no historical travel data for segment 3, so there is no predicted value for segment 3 or the predicted value for segment 3 is 0. The weight values of M1 and M2 are determined to be β1 and β2 respectively, then the recommended predicted value of candidate navigation route 1 is Q1=M1*β1+M2*β2. It should be noted that when determining the weight value of the segment prediction value, it is not limited to being based on the length ratio of the segment in the candidate navigation route, and the weight value of the segment prediction value of the segment can also be determined based on other relevant features.

In one embodiment, the weight values of the road segment prediction values may all be set to 1, and the recommended prediction value Q1 of the candidate navigation route A = M1 + M2.

The disclosed embodiment determines the recommended prediction value of the candidate navigation route by integrating the various sections that the navigated object has historically passed through in the candidate navigation route. The more sections that the navigated object has historically passed through in the candidate navigation route, the more familiar the navigated object is with the candidate navigation route. Therefore, the higher the recommended prediction value of the candidate navigation route obtained, the more familiar the navigated object is with the candidate navigation route, and the more the navigated object prefers to choose the candidate navigation route.

In the disclosed embodiment, the prediction data of the road section includes the planned characteristic data of the road section, and the planned characteristic data of the road section includes time characteristics, first distance characteristics, second distance characteristics, historical traffic information, planned traffic information, and road section proportion characteristics. Thus, the prediction data of the road section can be expressed as {time characteristics, first distance characteristics, second distance characteristics, historical traffic information, planned traffic information, and road section proportion characteristics}. The dimension of the prediction data of the road section can be 6 dimensions. The prediction data of the road section in the disclosed embodiment is highly extensible. If a new road section feature is added, the dimension of the prediction data of the road section only needs to be expanded from 6 dimensions. It can be up to 7 dimensions, so the scalability of the technical solution disclosed in the present invention is stronger.

In one embodiment, a method for recommending a navigation route can be applied to a server. Before obtaining the historical travel data of the candidate navigation routes including the road segment, the method for recommending the navigation route can also include: obtaining a planned starting point and a planned end point, and planning at least one candidate navigation route based on the planned starting point and the planned end point. For example, after the navigated object enters the planned starting point and the planned end point in an application with a map navigation function, the client can send the planned starting point, the planned end point and the client identifier to the server, and the client identifier corresponds to the navigated object. The server plans at least one candidate navigation route based on the acquired planned starting point and the planned end point, and the candidate navigation route is a navigation route from the planned starting point to the planned end point. The server obtains the historical travel data of the candidate navigation routes including the road segment based on the client identifier.

The server determines the navigation route recommended to the navigated object based on the recommended prediction values of the candidate navigation routes. When the recommended prediction values of the candidate navigation routes are all less than the preset threshold, the server may not recommend the candidate navigation route to the navigated object. The server may determine the navigation route recommended to the navigated object in other ways and recommend it to the navigated object.

In one embodiment, the method for recommending a navigation route can be applied to a server. Before obtaining the historical travel data of the candidate navigation route including the road segment, the navigation route recommendation method can also include: obtaining the candidate navigation route. For example, after the navigated object of the client enters the planned starting point and the planned end point in an application that supports the map navigation function, the client application plans the candidate navigation route according to the planned starting point and the planned end point, and sends the candidate navigation route and the client identifier to the server. The server obtains the historical travel data of the candidate navigation route including the road segment based on the obtained candidate navigation route and the client identifier. The historical travel data corresponds to the navigated object, and the navigated object corresponds to the client identifier.

It should be noted that when the navigated object uses an application that supports map navigation function for navigation, the various sections of the route the navigated object passes through and the time it passes through, as well as the starting point and end point of the corresponding historical travel route, can be recorded and stored. These data can be used as the historical travel data of the navigated object. The historical travel data corresponds to the navigated object, and different navigated objects may have different historical travel data.

In other embodiments, when the starting point and end point corresponding to other navigated objects are the same as the planned starting point and the planned end point, in order to improve the efficiency of navigation route recommendation, the navigation route recommended to the navigated object can be directly recommended to the other navigated object, or the route with the highest recommendation prediction value among the navigation routes recommended to the navigated object can be recommended to the other navigated object.

In one embodiment, at least one candidate navigation route may be recommended based on the recommendation prediction value of the candidate navigation route and the attribute information of the candidate navigation route. The attribute information of the candidate navigation route may include the total length of the candidate navigation route, the number of traffic lights on the candidate navigation route, the current time to pass the candidate navigation route, etc.

It should be noted that the factors that affect the selection of a candidate navigation route by the navigated object include not only the familiarity of the navigated object with the route, but also the attribute information of the candidate navigation route. It is understandable that if the recommended prediction value of a candidate navigation route is very high, but the current time to pass the candidate navigation route is far beyond the expectations of the navigated object, then the navigated object may not choose the candidate navigation route. Therefore, based on the recommended prediction value of the candidate navigation route and the attribute information of the candidate navigation route, at least one candidate navigation route is recommended, which is not only a route that the navigated object is familiar with, but also takes into account the actual conditions at the time of planning the route, which can further improve the satisfaction of the navigated object.

The following uses a specific embodiment to illustrate the application process of the navigation route recommendation method disclosed in the present invention.

For example, the candidate navigation routes include route A, route B, and route C. The segment sequences in each route are extracted to obtain the segment sequences of three routes: route A {segment 1, segment 2, segment 3}; route B {segment 4, segment 5, segment 6, segment 7}; Segment 7}; Route C {Segment 1, Segment 2, Segment 8, Segment 7}.

For route A, traverse each section of route A. If there is no historical travel data for the section, ignore the section. For example, if there is no historical travel data for section 1, but there is historical travel data for sections 2 and 3, ignore section 1 and obtain the historical travel data for sections 2 and 3. Construct the prediction data for section 2 in route A based on the historical travel data of section 2 and the route planning data of route A. Input the prediction data corresponding to section 2 into the trained familiar road prediction model to obtain the model prediction value corresponding to the prediction data of section 2. Perform weighted calculation on the model prediction values corresponding to the prediction data of section 2 to obtain the section prediction value of section 2. Similarly, obtain the section prediction value of section 3. Perform weighted calculation on the section prediction values of sections 2 and 3 to obtain the recommended prediction value of route A.

In the same way, the recommended prediction values for routes B and C are obtained.

Assume that the recommended prediction value of route A is 45, the recommended prediction value of route B is 50, the recommended prediction value of route C is 65, and the preset threshold is 50. The recommended prediction values of route A, route B, and route C are compared with the preset threshold respectively, and the route with the recommended prediction value greater than or equal to the preset threshold is determined as the target navigation route. The obtained target navigation route includes route B and route C. Route B and route C are sorted according to the recommended prediction value, and the route with the largest recommended prediction value among the target navigation routes can be recommended to the navigated object, for example, route C is recommended to the navigated object.

Fig. 2 is a flowchart of a method for training a familiar road prediction model in an embodiment of the present disclosure. The present disclosure also provides a method 200 for training a familiar road prediction model, as shown in Fig. 2, the training method 200 includes steps S201 to S203.

In step S201, a sample navigation route including sample historical travel data of a sample road segment is obtained.

The number of sample navigation routes may be one or more. The sample navigation routes may include a number of topologically connected sample sections. The sample historical travel data of the sample sections may include sample historical feature data characterizing the historical travel preferences of the navigated object. The sample historical travel data may be sample historical travel data of the sample sections within a preset historical time range, for example, within half a year or three months from the sample planning time. The preset historical time range may be set as needed.

The sample historical travel data corresponds to the navigated object. For example, the sample historical travel data of the corresponding sample road section of the navigated object can be obtained according to the identifier of the navigated object. The sample historical travel data records the sample historical feature data generated when the navigated object actually walked through the corresponding sample road section before the sample route planning time of the sample navigation route. Since the navigated object actually walked through the sample road section before, the sample historical feature data of the sample road section can characterize the changes in the navigated object's recent historical travel preferences for the sample road section.

The sample road sections in the sample navigation route may have sample historical travel data or may not have sample historical travel data. For example, if the navigated object passes through the sample road section within a preset historical time range, then the sample road section has corresponding sample historical travel data. If the navigated object does not pass through the sample road section within the preset historical time range, then the sample road section does not have sample historical travel data. The present disclosure is directed to the sample road sections included in the sample navigation route for which sample historical travel data can be obtained. For the calculation of the weights of the sample road sections for which sample historical travel data cannot be obtained, relevant existing technologies can be used, and the present disclosure does not impose any restrictions.

In step S202, a training sample of the sample road section is constructed according to the sample historical travel data of the sample road section and the sample route planning data of the sample navigation route. The training sample includes the sample planning feature data of the sample road section.

The sample route planning data in step S202 can be understood as the data that the application program with map navigation function and its service system need to use when planning the sample navigation route, including but not limited to the sample starting point, sample end point, sample route, and sample route. Conditions, etc. A training sample of the sample road segment is constructed based on the sample historical travel data of the sample road segment and the sample route planning data of the sample navigation route. The training sample includes sample planning feature data of the sample road segment, and the sample planning feature data of the sample road segment associates the features of the sample road segment included in the sample navigation route when it is planned with the features recorded in the sample historical travel data of the sample road segment, so that the travel preference changes of the navigated object are reflected in the prediction data of the sample navigation route.

In step 203, a familiar road prediction model is trained using training samples of the sample road segments, and the familiar road prediction model is used to determine a recommended prediction value of a candidate navigation route based on prediction data of the candidate navigation route including the road segments.

The training method provided by the disclosed embodiment obtains sample historical travel data of sample sections including sample navigation routes; constructs training samples of sample sections based on the sample historical travel data of the sample sections and sample route planning data of the sample navigation routes, wherein the training samples include sample planning feature data of the sample sections; and uses the training samples of the sample sections to train the familiar road prediction model. Since the sample historical travel data is data generated by the actual travel behavior of the navigated object, the sample historical travel data represents the actual travel preference of the navigated object for the corresponding sample sections. When the travel preference of the navigated object changes with time or environment, the sample historical travel data can easily capture the change in preference of the navigated object for the corresponding sample sections. Furthermore, since the training samples of the sample road segments constructed by the present disclosure are based on the sample historical travel data and the sample route planning data of the sample navigation route, that is, the training samples of the sample road segments constructed by the present disclosure associate the features of the sample road segments included in the sample navigation route when they are planned with the features of the sample road segments recorded in the sample historical travel data, so that the changes in the travel preferences of the navigated object are reflected in the training samples of the sample road segments. Therefore, after the familiar road prediction model is trained using the training samples determined by the embodiments of the present disclosure, the model prediction value output by the familiar road prediction model can more accurately reflect the preference of the navigated object for the sample road segments at the time of sample navigation route planning. The recommended prediction value of the candidate navigation route determined using the model prediction value output by the familiar road prediction model can more accurately reflect the current travel preference of the navigated object for the candidate navigation route. When a navigation route is recommended to the navigated object based on the recommended prediction value, the navigation route recommended to the navigated object can better meet the current travel preference of the navigated object, and the navigated object is more likely to choose the recommended navigation route. Therefore, the familiar route prediction model trained by the technical solution of the present invention can timely capture the changes in the travel preferences of the navigated object, and can more accurately recommend navigation routes that meet the preferences of the navigated object, thereby better realizing personalized route recommendations.

It should be noted that when the familiar road prediction model is used in the navigation route recommendation method in the embodiment of the present disclosure, the input of the familiar road prediction model is the prediction data of the road segment. Therefore, the training samples of the sample road segment can be understood according to the relevant content of the prediction data of the road segment.

In one embodiment, the sample historical travel data may include the sample historical travel time, which may be understood as the sample historical time when the sample navigation object passes through the sample road section. The sample route planning data may include the sample route planning time. The sample route planning time should be understood as the time when the application program with navigation function and its service system plan the sample navigation route according to the sample starting point and sample end point input by the sample navigation object.

The sample planning feature data of the sample road section includes: sample time features and sample distance features. According to the sample historical travel data of the sample road section and the sample route planning data of the sample navigation route, constructing the training sample of the sample road section may include: determining the sample time features corresponding to the sample historical travel time according to the sample historical travel time included in the sample historical travel data of the sample road section and the sample route planning time included in the sample route planning data of the sample navigation route; determining the sample historical travel time corresponding to the sample historical travel time according to the sample historical starting point and sample historical end point corresponding to the sample historical travel time in the sample historical travel data of the sample road section and the sample planning starting point and sample planning end point included in the sample route planning data of the sample navigation route. Sample distance feature: constructing a training sample of the sample road segment based on at least the sample time feature and the sample distance feature.

It is understandable that if the sample historical travel time of a sample road section differs from the sample route planning time, it means that the sample navigation object has walked the sample road section for a long time since the sample route planning time, and the sample navigation object's familiarity with the sample road section decays over time, that is, the farther the sample historical travel time differs from the sample route planning time, the less familiar the sample navigation object is with the sample road section. Conversely, the closer the sample historical travel time is to the sample route planning time, the more familiar the sample navigation object is with the sample road section. It can be seen from this that the sample time feature corresponding to the sample historical travel time determined based on the sample historical travel time of the sample road section and the sample route planning time of the sample navigation route takes into account the problem of familiarity decaying over time, and can better reflect the sample navigation object's preference or familiarity with the corresponding sample road section.

The sample history starting point and the sample history end point may be the starting point and the end point of the sample history travel route corresponding to the sample history travel time. The sample history travel route is the navigation route corresponding to the sample navigation object's history passing through the sample road section. The sample planning starting point and the sample planning end point are the starting point and the end point of the sample navigation route.

In this embodiment, the sample historical travel data includes the sample historical starting point and the sample historical end point corresponding to the sample historical travel time, and the sample route planning data includes the sample planning starting point and the sample planning end point. It can be understood that if the distance between the sample historical starting point and the sample planning starting point of the navigation route when the sample navigation object historically walked through a certain sample road section is very different, and the distance between the sample historical end point and the sample planning end point is very different, it means that even if the sample navigation object has walked through the sample road section, the sample navigation object is not familiar with the sample road section when walking from the sample planning starting point to the sample planning end point, or does not prefer the sample road section. On the contrary, if the distance between the sample historical starting point and the sample planning starting point of the navigation route when the sample navigation object historically walked through a certain sample road section is not much different, and the distance between the sample historical end point and the sample planning end point is not much different, it means that the sample navigation object prefers the sample road section in the process of walking from the sample planning starting point to the sample planning end point.

In the above embodiment of the present disclosure, the sample time feature corresponding to the sample historical travel time is determined according to the sample historical travel time and the sample route planning time; the sample distance feature corresponding to the sample historical travel time is determined according to the sample historical starting point, the sample historical end point, the sample planning starting point, and the sample planning end point, and the training sample of the sample road section is constructed based on at least the sample time feature and the sample distance feature. It can be seen that the training sample of the sample road section constructed by the present disclosure not only considers the influence of time on the travel preference of the sample navigation object, but also considers the influence of the sample planning starting point and the sample planning end point on the travel preference. Therefore, the familiar road prediction model trained by using the training sample considers the influence of time on the travel preference, and considers the influence of the route starting point and the end point on the travel preference. Therefore, the recommended prediction value of the candidate navigation route determined by the familiar road prediction model of the embodiment of the present disclosure can more accurately reflect the current preference of the navigation object for the candidate navigation route, can better reflect the familiarity of the navigation object with the navigation route, can more accurately recommend the route familiar to the navigation object, and better realize personalized route recommendation.

In one embodiment, the sample time feature may include a time difference between the sample historical travel time and the sample route planning time.

In one embodiment, the sample distance feature includes a sample first distance feature and a sample second distance feature. The sample distance feature corresponding to the sample historical travel time is determined based on the sample historical starting point and the sample historical end point corresponding to the sample historical travel time in the sample historical travel data of the sample road section, and the sample planning starting point and the sample planning end point included in the sample route planning data of the sample navigation route, including: determining the distance between the sample historical starting point and the sample planning starting point as the sample first distance feature; determining the distance between the sample historical end point and the sample planning end point as the sample second distance feature.

It should be noted that the greater the difference between the sample historical travel data of the sample road section and the sample route planning data of the sample navigation route, the greater the difference between the sample navigation route and the sample historical travel route, and correspondingly, the less familiar the sample navigation object is with the sample navigation route; conversely, the more familiar the sample navigation object is with the sample navigation route. Therefore, the distance between the sample historical starting point and the sample planning starting point is determined as the sample first distance feature, and the distance between the sample historical end point and the sample planning end point is determined as the sample second distance feature, so that the training sample can better reflect the difference between the sample historical travel data of the sample road section and the sample route planning data of the sample navigation route. After using the training sample to train the familiar road prediction model, the accuracy of the familiar road prediction model output can be improved, so that the obtained recommended prediction value can more accurately reflect the current preference of the navigated object for the planned navigation route.

In one embodiment, the sample planning feature data of the sample road section also includes sample road condition features, and the method also includes: determining the sample road condition features corresponding to the sample historical travel time based on the sample historical road condition information corresponding to the sample historical travel time in the sample historical travel data of the sample road section, and the sample planning road condition information of the sample road section included in the sample route planning data.

The traffic information of a route or road section can reflect or even determine whether the navigated object chooses the route or road section. Therefore, the traffic information can reflect the navigated object's preference for the route or road section. Therefore, the sample traffic characteristics corresponding to the sample historical travel time are determined based on the sample historical traffic information and the sample planned traffic information. The training sample includes the sample traffic characteristics, which can further improve the accuracy of the familiar road prediction model output trained using the training sample.

For example, the traffic condition information may include smooth, slow, congested, extremely congested, no traffic condition or under construction. For example, if the sample historical traffic condition information is smooth and the sample planned traffic condition information is congested, then the sample navigation object has a low preference for the sample navigation route, and the sample navigation object is less likely to choose the sample navigation route. It should be noted that no traffic condition means that the traffic condition of the road section cannot be known.

In one embodiment, the sample traffic condition feature may include the difference between the sample historical traffic condition information corresponding to the sample historical travel time and the sample planned traffic condition information of the sample road section. For example, a value is assigned to each type of traffic condition information, such as 1 for smooth traffic, 0.8 for slow traffic, 0.6 for congestion, 0.4 for extreme congestion, 0.2 for no traffic condition, and 0 for under construction. The difference between the sample historical traffic condition information and the sample planned traffic condition information may be determined as the sample traffic condition feature.

In one embodiment, the traffic condition feature may include sample historical traffic condition information corresponding to the sample historical travel time and sample planned traffic condition information of the sample road section included in the sample route planning data.

It should be noted that the sample historical road condition information and the sample planned road condition information correspond to each other. When the sample historical road condition information corresponding to the sample historical travel time is the sample historical road condition information of the sample section, the sample planned road condition information is the road condition information of the sample section in the sample navigation route at the sample planning time; when the sample historical road condition information corresponding to the sample historical travel time is the overall historical road condition information of the sample historical travel route, the sample planned road condition information is the overall road condition information of the sample navigation route at the planning time.

In one embodiment, the sample planning feature data of the sample road section also includes a sample road section proportion feature, and the method further includes: determining the sample road section proportion feature according to the length proportion of the sample road section in the sample navigation route.

It can be understood that the greater the proportion of the sample road segment in the sample navigation route, the more the navigated object prefers the sample navigation route. Therefore, when the training samples include the sample road segment proportion feature, the accuracy of the output of the familiar road prediction model can be further improved after the training samples including the sample road segment proportion feature are used to train the familiar road prediction model, and the accuracy of the recommended prediction value of the candidate navigation route determined by the familiar road prediction model can be further improved, so as to better reflect the navigated object's preference for the candidate navigation route. The current preference for the route.

In one embodiment, a training sample of a sample road section is constructed based at least on sample time features and sample distance features, including: constructing a training sample of a sample road section with sample time features, sample distance features, sample road condition features and sample road section ratio features corresponding to the same sample historical travel time.

It is understandable that each time the sample navigation object passes through the sample road section, a corresponding sample historical travel data will be generated, each sample historical travel data corresponds to a sample historical travel time, and each sample historical travel time corresponds to a training sample. A training sample of a sample road section can be expressed as {sample time feature, sample distance feature, sample road condition feature, sample road section proportion feature}, and a training sample can also be expressed as {sample time feature, sample first distance feature, sample second distance feature, sample historical road condition information, sample planned road condition information, sample road section proportion feature}.

In one embodiment, a familiar road prediction model is trained using training samples of a sample road section, including: inputting the training samples of the sample road section into the familiar road prediction model to obtain model prediction values corresponding to the training samples of the sample road section; and adjusting parameters of the familiar road prediction model based on the model prediction values corresponding to the training samples of the sample road section, the sample navigation route, and the sample travel route.

In one embodiment, based on the model prediction values corresponding to the training samples of the sample road segment, the sample navigation route and the sample travel route, the parameters of the familiar road prediction model are adjusted, including: determining the travel coverage rate of the sample navigation route, the travel coverage rate is the proportion of the length of the repeated sections of the sample navigation route and the sample travel route in the sample navigation route; determining the sample recommendation prediction value of the sample navigation route according to the model prediction values corresponding to the training samples of the sample road segment; determining the loss value of the familiar road prediction model according to the sample recommendation prediction value of the sample navigation route and the travel coverage rate of the sample navigation route; and adjusting the parameters of the familiar road prediction model based on the loss value.

In one embodiment, a sample recommendation prediction value of a sample navigation route is determined based on the model prediction value corresponding to the training sample of the sample road section, including: inputting the sample navigation route including the training sample of the sample road section into a familiar road prediction model to obtain the model prediction value corresponding to the training sample of the sample road section; determining the sample road section prediction value of the sample road section based on the model prediction value corresponding to the training sample of the sample road section; determining the sample recommendation prediction value of the sample navigation route based on the sample road section prediction value including the sample road section of the sample navigation route.

In one embodiment, determining the sample section prediction value of the sample section according to the model prediction value corresponding to the training sample of the sample section includes: performing weighted calculation on the model prediction value corresponding to the training sample of the sample section to obtain the sample section prediction value of the sample section.

For example, based at least on the time features in the training samples, the weight values of the model prediction values corresponding to the training samples are determined, and the product of the model prediction values and the weight values is calculated. The products corresponding to the model prediction values are added together to obtain the sample road section prediction value of the sample road section.

In one embodiment, the sum of the model prediction values corresponding to the training samples of the sample road section may be used as the sample road section prediction value of the sample road section.

In one embodiment, the sample recommended prediction value of the sample navigation route is determined based on the sample segment prediction value of the sample segment included in the sample navigation route, including: weighted calculation of the sample segment prediction value of the sample navigation route including the sample segment to obtain the sample recommended prediction value of the sample navigation route.

For example, the weight value of the sample road segment prediction value can be determined based on at least the length ratio of the sample road segment in the sample navigation route. The product of the sample road segment prediction value and the weight value is calculated. The products corresponding to the sample road segment prediction values are added together to obtain the sample road segment prediction value. Sample recommended predicted values for this navigation route.

In one embodiment, the sum of the sample segment prediction values of the sample segments included in the sample navigation route may be used as the sample recommendation prediction value of the sample navigation route.

The sample recommendation prediction value of the sample navigation route is determined, the travel coverage rate of the sample navigation route is used as a label, and the parameters of the familiar route prediction model are adjusted.

For example, the training samples include sample time features, sample first distance features, sample second distance features, sample historical road condition information, sample planned road condition information, and sample road section proportion features. The parameters of the familiar road prediction model include the weight value of the sample time features, the weight value of the sample first distance features, the weight value of the sample second distance features, the weight value of the sample historical road condition information, the weight value of the sample planned road condition information, and the weight value of the sample road section proportion feature. Different training samples are used to train the familiar road prediction model to obtain the weight value of the sample time features, the weight value of the sample first distance features, the weight value of the sample second distance features, the weight value of the sample historical road condition information, the weight value of the sample planned road condition information, and the weight value of the sample road section proportion feature that converge the loss value.

In the embodiment of the present disclosure, the training sample includes a sample time feature, a sample first distance feature, a sample second distance feature, a sample historical road condition information, a sample planned road condition information, and a sample road section ratio feature. The training sample can be expressed as {sample time feature, sample first distance feature, sample second distance feature, sample historical road condition information, sample planned road condition information, sample road section ratio feature}. The dimension of the training sample is 6 dimensions. The training sample in the embodiment of the present disclosure has strong scalability. If a new sample planning feature is added, it is only necessary to expand the dimension of the training sample from 6 dimensions to 7 dimensions. Thus, the technical solution of the present disclosure has stronger scalability.

Fig. 3 is a structural block diagram of a navigation route recommendation device in an embodiment of the present disclosure. As shown in Fig. 3 , the navigation route recommendation device 300 includes a first acquisition module 301 , a first construction module 302 and a determination module 303 .

The first acquisition module 301 is used to acquire the historical travel data of the candidate navigation route including the road segments.

The first constructing module 302 is used to construct prediction data of the road segment according to the historical travel data of the road segment and the route planning data of the candidate navigation route, wherein the prediction data includes planning feature data of the road segment.

The determination module 303 is used to determine the recommended prediction value of the candidate navigation route based on the prediction data of the candidate navigation route including the road segments.

In one embodiment, the planning feature data of the road section includes: time feature, distance feature, and the first construction module 302 includes: a first determination submodule, used to determine the time feature corresponding to the historical travel time according to the historical travel time included in the historical travel data of the road section and the route planning time included in the route planning data of the candidate navigation route; a second determination submodule, used to determine the distance feature corresponding to the historical travel time according to the historical starting point and historical end point corresponding to the historical travel time in the historical travel data of the road section, and the planned starting point and planned end point included in the route planning data of the candidate navigation route; a construction submodule, used to construct the prediction data of the road section based on at least the time feature and the distance feature.

In one embodiment, the distance feature includes a first distance feature and a second distance feature, and the second determination submodule is used to: determine the distance between the historical starting point and the planned starting point as the first distance feature; and determine the distance between the historical end point and the planned end point as the second distance feature.

In one embodiment, the planning characteristic data of the road section also includes road condition characteristics, and the first construction module 302 also includes: a third determination submodule for determining the road condition information corresponding to the historical travel time in the historical travel data of the road section and the route planning The planned traffic information of the road sections included in the planning data is used to determine the traffic characteristics corresponding to the historical travel time.

In one embodiment, the planning characteristic data of the road section also includes a road section proportion feature, and the first construction module 302 also includes: a fourth determination submodule, which is used to determine the road section proportion feature according to the length proportion of the road section in the candidate navigation route.

In one embodiment, the construction submodule is used to construct the time characteristics, distance characteristics, road condition characteristics and road section proportion characteristics corresponding to the same historical travel time into a piece of prediction data for the road section.

In one embodiment, the determination module is also used to: input the prediction data of the candidate navigation route including the road segment into the trained familiar road prediction model to obtain the model prediction value corresponding to the prediction data of the road segment; determine the road segment prediction value of the road segment according to the model prediction value corresponding to the prediction data of the road segment; determine the recommended prediction value of the candidate navigation route according to the road segment prediction value of the candidate navigation route including the road segment.

In one embodiment, the determination module is also used to: perform weighted calculation on the model prediction values corresponding to the prediction data of the road section to obtain the road section prediction value of the road section; perform weighted calculation on the road section prediction values including the road section of the candidate navigation route to obtain the recommended prediction value of the candidate navigation route.

Fig. 4 is a block diagram of a training device for a familiar road prediction model in an embodiment of the present disclosure. As shown in Fig. 4 , the training device 400 for a familiar road prediction model includes a second acquisition module 401 , a second construction module 402 and a training module 403 .

The second acquisition module 401 is used to acquire sample historical travel data of a sample navigation route including a sample road section.

The second construction module 402 is used to construct a training sample of the sample road section according to the sample historical travel data of the sample road section and the sample route planning data of the sample navigation route, wherein the training sample includes the sample planning feature data of the sample road section.

The training module 403 is used to train the familiar road prediction model using training samples of the sample road segments, and the familiar road prediction model is used to determine the recommended prediction value of the candidate navigation route based on the prediction data of the candidate navigation route including the road segments.

In one embodiment, the sample planning feature data of the sample road section includes: sample time feature and sample distance feature. The second construction module 402 is further used to: determine the sample time feature corresponding to the sample historical travel time according to the sample historical travel time included in the sample historical travel data of the sample road section and the sample route planning time included in the sample route planning data of the sample navigation route; determine the sample distance feature corresponding to the sample historical travel time according to the sample historical starting point and the sample historical end point corresponding to the sample historical travel time in the sample historical travel data of the sample road section, and the sample planning starting point and the sample planning end point included in the sample route planning data of the sample navigation route; and construct a training sample of the sample road section based on at least the sample time feature and the sample distance feature.

In one embodiment, the sample distance feature includes a sample first distance feature and a sample second distance feature, and the second construction module 402 is further used to: determine the distance between the sample history starting point and the sample planning starting point as the sample first distance feature; and determine the distance between the sample history end point and the sample planning end point as the sample second distance feature.

In one embodiment, the sample planning feature data of the sample road section also includes sample road condition features, and the second construction module 402 is also used to determine the sample road condition features corresponding to the sample historical travel time based on the sample historical road condition information corresponding to the sample historical travel time in the sample historical travel data of the sample road section, and the sample planning road condition information of the sample road section included in the sample route planning data.

In one embodiment, the sample planning feature data of the sample road section also includes a sample road section proportion feature, and the second construction module 402 is further used to determine the sample road section proportion feature according to the length proportion of the sample road section in the sample navigation route.

In one embodiment, the second construction module 402 is further used to: The features of sample distance, sample road condition and sample road section proportion are constructed as a training sample of the sample road section.

In one embodiment, the training module includes: an acquisition submodule, which is used to input the training samples of the sample road section into the familiar road prediction model to obtain the model prediction values corresponding to the training samples of the sample road section; an adjustment submodule, which is used to adjust the parameters of the familiar road prediction model based on the model prediction values corresponding to the training samples of the sample road section, the sample navigation route and the sample travel route.

In one embodiment, the adjustment submodule is used to: determine the travel coverage rate of the sample navigation route, where the travel coverage rate is the proportion of the length of the repeated sections of the sample navigation route and the sample travel route in the sample navigation route; determine the sample recommendation prediction value of the sample navigation route based on the model prediction value corresponding to the training sample of the sample section; determine the loss value of the familiar road prediction model based on the sample recommendation prediction value of the sample navigation route and the travel coverage rate of the sample navigation route; and adjust the parameters of the familiar road prediction model based on the loss value.

FIG5 is a block diagram of an electronic device for implementing an embodiment of the present disclosure. As shown in FIG5 , the electronic device includes: a memory 510 and a processor 520, wherein the memory 510 stores a computer program that can be run on the processor 520. When the processor 520 executes the computer program, the method in the above embodiment is implemented. The number of the memory 510 and the processor 520 can be one or more.

The electronic device also includes:

The communication interface 530 is used to communicate with external devices and perform data exchange transmission.

If the memory 510, the processor 520 and the communication interface 530 are implemented independently, the memory 510, the processor 520 and the communication interface 530 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in FIG5, but it does not mean that there is only one bus or one type of bus.

Optionally, in a specific implementation, if the memory 510, the processor 520 and the communication interface 530 are integrated on a chip, the memory 510, the processor 520 and the communication interface 530 can communicate with each other through an internal interface.

An embodiment of the present disclosure provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present disclosure when the program is executed by a processor.

The embodiment of the present disclosure also provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided by the embodiment of the present disclosure.

The embodiment of the present disclosure also provides a chip, including: an input interface, an output interface, a processor and a memory. The input interface, the output interface, the processor and the memory are connected through an internal connection path. The processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided in the embodiment of the application.

It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the Advanced RISC Machines (ARM) architecture.

Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory, and may also include a non-volatile random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of exemplary but not limiting description, many forms of RAM are available. For example, static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DR RAM).

In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present disclosure is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

Any process or method description in the flowchart or otherwise described herein can be understood to represent a module, segment or portion of code including one or more executable instructions for implementing the steps of a specific logical function or process. And the scope of the preferred embodiments of the present disclosure includes other implementations, in which the functions may not be performed in the order shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved.

The logic and/or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor or other system that can fetch instructions from an instruction execution system, apparatus or device and execute instructions), or used in combination with these instruction execution systems, apparatuses or devices.

It should be understood that the various parts of the present disclosure may be implemented with hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods may be implemented with software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above embodiment method may be completed by instructing the relevant hardware through a program, which may be stored in a computer-readable storage medium, and when the program is executed, includes one or a combination of the steps of the method embodiment.

In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The storage medium may be a read-only memory, a disk or an optical disk, etc.

The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope disclosed in the present disclosure, which should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

A method for recommending a navigation route, characterized by comprising:

Obtain candidate navigation routes including historical travel data of road segments;

Constructing prediction data of the road segment according to the historical travel data of the road segment and the route planning data of the candidate navigation route, wherein the prediction data includes planning feature data of the road segment;

Based on the prediction data of the candidate navigation route including the road segments, a recommended prediction value of the candidate navigation route is determined.
The method according to claim 1 is characterized in that the planning feature data of the road section includes: time features and distance features, and the step of constructing the prediction data of the road section based on the historical travel data of the road section and the route planning data of the candidate navigation route includes:

Determining the time feature corresponding to the historical travel time according to the historical travel time included in the historical travel data of the road section and the route planning time included in the route planning data of the candidate navigation route;

Determine the distance feature corresponding to the historical travel time according to the historical starting point and the historical end point corresponding to the historical travel time in the historical travel data of the road section, and the planned starting point and the planned end point included in the route planning data of the candidate navigation route;

Prediction data of the road segment is constructed based on at least the time feature and the distance feature.
The method according to claim 2 is characterized in that the distance feature includes a first distance feature and a second distance feature, and determining the distance feature corresponding to the historical travel time according to the historical starting point and the historical end point corresponding to the historical travel time in the historical travel data of the road segment, and the planned starting point and the planned end point included in the route planning data of the candidate navigation route, comprises:

determining the distance between the historical starting point and the planned starting point as the first distance feature;

The distance between the historical endpoint and the planned endpoint is determined as the second distance feature.
The method according to claim 2 is characterized in that the planning characteristic data of the road section also includes road condition characteristics, and the method further comprises:

The road condition characteristics corresponding to the historical travel time are determined based on the historical road condition information corresponding to the historical travel time in the historical travel data of the road section and the planned road condition information of the road section included in the route planning data.
The method according to claim 4 is characterized in that the planning characteristic data of the road section also includes road section proportion characteristics, and the method further includes:

The road section proportion feature is determined according to the proportion of the length of the road section in the candidate navigation route.
The method according to claim 5, characterized in that constructing the prediction data of the road section based at least on the time feature and the distance feature comprises:

The time characteristics, distance characteristics, road condition characteristics and the road section proportion characteristics corresponding to the same historical travel time are constructed into a piece of prediction data for the road section.
The method according to any one of claims 1 to 6, characterized in that determining the recommended prediction value of the candidate navigation route based on the prediction data of the candidate navigation route including the road segment comprises:

The candidate navigation route includes prediction data of road segments, which are input into a trained familiar road prediction model to obtain the road segments. The model prediction value corresponding to the predicted data;

Determining a section prediction value of the section according to a model prediction value corresponding to the prediction data of the section;

According to the predicted values of the sections included in the candidate navigation route, a recommended predicted value of the candidate navigation route is determined.
The method according to claim 7, characterized in that

Determining a section prediction value of the section according to a model prediction value corresponding to the prediction data of the section includes:

Performing weighted calculation on the model prediction values corresponding to the prediction data of the road section to obtain the road section prediction value of the road section;

Determining a recommended prediction value of the candidate navigation route according to the predicted value of the segment of the candidate navigation route includes:

The candidate navigation route includes the road segment prediction values of the road segments and performs weighted calculation to obtain the recommended prediction value of the candidate navigation route.
A method for training a familiar road prediction model, characterized by comprising:

Obtaining sample historical travel data of a sample navigation route including a sample road section;

Constructing a training sample of the sample road section according to the sample historical travel data of the sample road section and the sample route planning data of the sample navigation route, wherein the training sample includes the sample planning feature data of the sample road section;

The familiar road prediction model is trained using the training samples of the sample road segments, and the familiar road prediction model is used to determine the recommended prediction value of the candidate navigation route based on the prediction data of the candidate navigation route including the road segments.
The method according to claim 9, characterized in that the familiar road prediction model is trained using the training samples of the sample road section, comprising:

Inputting the training samples of the sample road section into the familiar road prediction model to obtain the model prediction value corresponding to the training samples of the sample road section;

The parameters of the familiar road prediction model are adjusted based on the model prediction values corresponding to the training samples of the sample road section, the sample navigation route and the sample travel route.
The method according to claim 10, characterized in that adjusting the parameters of the familiar road prediction model based on the model prediction value corresponding to the training sample of the sample road section, the sample navigation route and the sample travel route comprises:

Determine the travel coverage rate of the sample navigation route, where the travel coverage rate is the ratio of the length of the sample navigation route and the repeated sections of the sample travel route in the sample navigation route;

Determining a sample recommendation prediction value of the sample navigation route according to the model prediction value corresponding to the training sample of the sample road section;

Determining a loss value of the familiar route prediction model according to the sample recommendation prediction value of the sample navigation route and the travel coverage rate of the sample navigation route;

Based on the loss value, the parameters of the familiar road prediction model are adjusted.
A navigation route recommendation device, characterized by comprising:

A first acquisition module is used to acquire historical travel data of candidate navigation routes including road sections;

A first construction module is used to construct prediction data of the road section according to historical travel data of the road section and route planning data of the candidate navigation route, wherein the prediction data includes planning feature data of the road section;

A determination module is used to determine the predicted route of the candidate navigation route based on the predicted data of the candidate navigation route including the road segment. Recommended predicted value.
A training device for a familiar road prediction model, characterized by comprising:

A second acquisition module is used to acquire sample historical travel data of a sample navigation route including a sample road section;

A second construction module is used to construct a training sample of the sample road section according to the sample historical travel data of the sample road section and the sample route planning data of the sample navigation route, wherein the training sample includes the sample planning feature data of the sample road section;

A training module is used to train a familiar road prediction model using training samples of the sample road segments, wherein the familiar road prediction model is used to determine a recommended prediction value of the candidate navigation route based on prediction data of the candidate navigation route including the road segments.
A computer-readable storage medium having a computer program stored therein, wherein the computer program, when executed by a processor, implements the method according to any one of claims 1 to 11.