CN114218476B

CN114218476B - Content recommendation method and device and terminal equipment

Info

Publication number: CN114218476B
Application number: CN202111341495.0A
Authority: CN
Inventors: 孔令军; 宋广军; 俞淇纲
Original assignee: Shenzhen Qianhai Pengying Digital Software Operation Co ltd
Current assignee: Shenzhen Qianhai Pengying Digital Software Operation Co ltd
Priority date: 2021-11-12
Filing date: 2021-11-12
Publication date: 2022-10-04
Anticipated expiration: 2041-11-12
Also published as: CN114218476A

Abstract

The application provides a content recommendation method, a content recommendation device and terminal equipment, which are applicable to the technical field of electronic commerce, and the method comprises the following steps: acquiring operation data of a first user in a platform; performing behavior analysis on the first user based on the operation data, and selecting at least one user tag from the user portrait according to the user behavior; obtaining use sample data corresponding to the user tag, and predicting user requirements of the first user based on the user behavior and the use sample data; screening out data to be analyzed related to user requirements from the use sample data; and analyzing the data to be analyzed to determine the content to be recommended corresponding to the first user. According to the method and the device, personalized function recommendation of the user can be achieved, and the function utilization rate of the e-commerce platform is improved.

Description

Content recommendation method and device and terminal equipment

Technical Field

The application belongs to the technical field of electronic commerce, and particularly relates to a content recommendation method, a content recommendation device and terminal equipment.

Background

With the development of internet technology, the functions of O2O (Online To Offline) e-commerce platforms are becoming more powerful, and a multifunctional network platform with functions of shopping, entertainment, traveling and the like has evolved from a single Online shopping platform. The richness of the functions of the O2O e-commerce platform (hereinafter referred to as the platform) aims to provide convenience for the life of users. However, in practical applications, many users do not actively seek to try functions unfamiliar with themselves. Resulting in a platform with low user usage of many functions.

To provide user usage of platform functionality. One way is that in the process of using the platform by the user, the platform actively pushes the introduction of new functions for the user, or prompts some new functions existing in the user platform and provides shortcut operation of starting the functions, so as to attract the user to try the new functions. Although the method can attract users to try new functions of the platform to a certain extent, the method is not strong in pertinence to the users, and the actual effect is poor. Meanwhile, the experience of the user on the platform is also influenced, and the development of the platform is not facilitated. Therefore, a method for recommending platform functions to a user in a personalized manner is needed, so that the utilization rate of the platform functions by the user is improved.

Disclosure of Invention

In view of this, embodiments of the present application provide a content recommendation method, a content recommendation device, and a terminal device, which can solve the problem that the prior art cannot personally recommend a platform function to a user.

A first aspect of an embodiment of the present application provides a content recommendation method, including:

operation data of a first user in a platform is obtained.

And performing behavior analysis on the first user based on the operation data, predicting the user behavior of the first user, and selecting at least one user tag from the user portrait of the first user according to the user behavior.

And acquiring the use sample data corresponding to all the selected user tags, and predicting the user requirement of the first user based on the user behavior and the use sample data. The historical use data of the first user on the functions in the platform is recorded in the use sample data.

And screening out the data to be analyzed related to the user requirements from the use sample data.

And analyzing the data to be analyzed to determine the content to be recommended corresponding to the first user. The content to be recommended is a function or a functional content.

In a possible implementation manner of the first aspect, analyzing data to be analyzed to determine content to be recommended corresponding to the first user includes:

and analyzing the data to be analyzed, and determining functions to which the data to be analyzed belongs and function contents used by the first user under the functions.

Analyzing attribute data of each determined function and function content from data to be analyzed, wherein the attribute data comprises:

the average monthly usage time period T11 of the function, the average monthly usage times N11 of the function, the total time period T12 and the total times N12 of the function usage in the last month, the average monthly usage time period T21 of the function content, the average monthly usage times N21 of the function content, and the total time period T22 and the total times N22 of the function content in the last month.

And calculating the average value T01 of the average monthly use time of all the determined functions, the average value N01 of the average monthly use times, the average time T02 of all the determined functions in the last month and the average number N02 of all the determined functions. Average value of the average usage time of all the functional contents in the month T03, average value of the number of times of usage in the month N03, average time of usage of all the functional contents in the last month T04, and average number of times N04.

The demand score for each function is calculated according to the following formula.

The demand score for each functional content is calculated according to the following formula.

F1 is the demand score of the function, F2 is the demand score of the function content, a, b, c and d are weighting coefficients and are used for adjusting the score proportion of the using time length and the using times, e is an adjusting coefficient and is used for adjusting the demand score of the function content, and c > a, d > b and e >1.

And taking the function or the function content with the highest demand score as the content to be recommended corresponding to the first user.

In a second possible implementation manner of the first aspect, before the selecting, according to the user behavior, at least one user tag from the user representation of the first user, the method further includes:

user information of the first user and historical use data of each function in the platform of the first user are obtained.

And sampling historical use data of each function to obtain use sample data.

And determining user tags based on the use sample data to obtain a plurality of user tags corresponding to the first user, and completing construction of the user portrait of the first user. Wherein each user tag has associated therewith one or more functions.

In a third possible implementation manner of the first aspect, after obtaining the user information of the first user and the historical usage data of each function in the platform by the first user, sampling the historical usage data of each function, and before obtaining usage sample data, the method further includes:

and if the first user has the to-be-processed function without the historical use data, finding out at least one second user with the historical use data for the to-be-processed function.

And acquiring user information of each second user and historical use data of each function in the platform.

And removing the historical use data of the to-be-processed function of each second user from the historical use data of the second users.

And calculating the similarity of the first user and each second user according to the user information of the first user, the historical use data of the first user for each function of the platform, the user information of each second user and the historical use data of each second user for the function in the platform after the elimination operation.

And screening out the second user with the highest similarity, and taking the screened historical use data of the second user on the function to be processed as the historical use data of the first user on the function to be processed.

In a fourth possible implementation manner of the first aspect, calculating a similarity between the first user and each second user according to the user information of the first user, the historical usage data of the first user on each function of the platform, the user information of each second user, and the historical usage data of each second user on the function in the platform after the eliminating operation includes:

and calculating the similarity between the user information of the first user and the user information of the second user.

And respectively counting the total use time length and the total use times of each function by the first user, and respectively counting the total use time length and the total use times of each function by the second user except the function to be processed.

Calculating the similarity between the first user and the second user by using the following formula:

here, the first user is referred to as user b, and the second user is referred to as user c. S _bc1 Representing the similarity of user b and user c, N _(b) Indicating the set of functions used by user b, N _(c) Indicating that user c has used a set of functions other than the pending function, T _bi Indicates the total time length of the use of the ith function by the user b, T _ci Indicates the total time of using the ith function by the user c, N _bi Indicates the total number of times of use of the ith function by the user b, N _ci Indicates the total number of times of use of the ith function by the user c, S _bc2 Representing the user information similarity of user b and user c, | N _(b) I and | N _(c) Respectively represents N _(b) And N _(c) Number of functions involved, | N _(b) |∪|N _(c) I represents N _(b) And N _(c) The sum of the number of functions involved.

Alpha is a weight factor used for adjusting the similarity of the user b and the user c in function use and the similarity weight of the user information, beta is a time factor, and delta is a quantity factor.

In a fifth possible implementation manner of the first aspect, the sampling historical usage data of each function to obtain usage sample data includes:

the functions in the platform are divided into four functional types, namely a food dimension, an entertainment dimension, a trip dimension and a health dimension. Wherein the food dimension comprises diet related functions, the entertainment dimension comprises recreational and recreational related functions, the travel dimension comprises travel and accommodation related functions, and the health dimension comprises diet related functions and user visit and medication related functions.

And acquiring different sampling strategies corresponding to each function type.

And according to the sampling strategy respectively corresponding to each function type, respectively sampling historical use data of the functions in each function type to obtain use sample data.

A second aspect of an embodiment of the present application provides a content recommendation apparatus, including:

and the data acquisition module is used for acquiring the operation data of the first user in the platform.

And the label selecting module is used for performing behavior analysis on the first user based on the operation data, predicting the user behavior of the first user, and selecting at least one user label from the user portrait of the first user according to the user behavior.

And the demand forecasting module is used for acquiring the use sample data corresponding to all the selected user tags and forecasting the user demand of the first user based on the user behavior and the use sample data. The use sample data records historical use data of the first user for the functions in the platform.

And the data screening module is used for screening the data to be analyzed related to the user requirements from the use sample data.

And the content determining module is used for analyzing the data to be analyzed and determining the content to be recommended corresponding to the first user. The content to be recommended is a function or a functional content.

A third aspect of embodiments of the present application provides a terminal device, where the terminal device includes a memory and a processor, where the memory stores a computer program that is executable on the processor, and the processor implements the steps of the content recommendation method according to any one of the first aspect when executing the computer program.

A fourth aspect of embodiments of the present application provides a computer-readable storage medium, comprising: there is stored a computer program, characterized in that the computer program, when being executed by a processor, carries out the steps of the content recommendation method according to any one of the above-mentioned first aspects.

A fifth aspect of embodiments of the present application provides a computer program product, which, when run on a terminal device, causes the terminal device to execute any one of the content recommendation methods of the first aspect.

Compared with the prior art, the embodiment of the application has the advantages that: according to the method and the device, the user label and the use sample data under the label are positioned by predicting the user behavior, the user requirement is determined by using the sample data, and then the use data are secondarily screened according to the user requirement. The data to be analyzed screened out at the moment is sample data screened out through the user behavior and the user demand, and the correlation degree of the data to be analyzed and the actual demand of the user is extremely high. The selection of the use sample data and the secondary screening process are adaptively realized according to the actual condition of the user, so that a large amount of weakly related use data can be reduced, and the influence of human subjective factors can be avoided. The selected use sample data can better meet the individual requirements of the user. And then screening the contents to be recommended which are suitable for the user based on the data to be analyzed. Therefore, the final recommended content of the embodiment of the application is more accurate.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required to be used in the embodiments or the prior art description will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings may be obtained according to these drawings without inventive labor.

FIG. 1 is a schematic diagram illustrating an implementation flow of a user portrait stage in a content recommendation method according to an embodiment of the present application;

FIG. 2 is a schematic diagram illustrating an implementation flow of a user portrait stage in a content recommendation method according to an embodiment of the present application;

fig. 3 is a schematic flow chart illustrating an implementation of calculating user similarity in the content recommendation method according to the embodiment of the present application;

fig. 4 is a schematic view of an implementation flow of content recommendation in a content recommendation method provided in an embodiment of the present application;

fig. 5 is a schematic structural diagram of a content recommendation device provided in an embodiment of the present application;

fig. 6 is a schematic diagram of a terminal device provided in an embodiment of the present application.

Detailed Description

In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

In order to explain the technical solution described in the present application, the following description will be given by way of specific examples.

With the continuous increase of the actual demands of users, users gradually hope to find more life services on line, and therefore the original online shopping function of the platform cannot meet the actual demands of users. Meanwhile, with the continuous development of the internet technology, a technical basis is provided for the richness of the platform functions. On the basis, the platform gradually evolves from a single online shopping platform to a multifunctional network platform. The common functions are as follows: take-out, group-buying gourmet, group-buying vegetable, online dish market, online supermarket, lodging, leisure entertainment, taxi trip, train ticket or air ticket, medicine buying, online hospital and the like.

The increasing abundance of platform functions brings more life convenience to users. However, in real life, many users do not try to explore the habit of platform functions, so that the frequency of use of the actual users of the platform with many functions is not high. At this time, the abundant functions do not bring convenience to the user, but may cause the user to feel the platform bloated, thereby reducing the user experience.

In order to guide the user to use the platform functions, an alternative way is that the platform actively pushes the introduction of new functions for the user during the use of the platform by the user, or prompts some new functions existing in the platform of the user and provides a shortcut operation for starting the functions, so as to attract the user to try the new functions. However, the actual situation is different for each user, and not all functions are required to be used by a single user. The general recommendation of new functions is poor in user pertinence, so that the final actual effect is not good. Therefore, how to recommend some content really interested or having a demand for the user is a problem which needs to be solved urgently and has a great practical significance, so that the user can start to gradually adapt to the platform function on the basis of the actual demand of the user and use the platform function.

In order to mine the actual requirements of the user, the embodiment of the application firstly analyzes historical use data of the user in the platform, and determines a plurality of labels actually possessed by the user and use data corresponding to each label, so that the user portrait is established. These labels reflect the characteristics of the user in different dimensions. On this basis, the embodiment of the application records the operation data of the user in the platform, and performs behavior analysis on the user based on the operation data to determine some behaviors which are actually possible for the user. Tags within the user representation are then matched based on the possible behaviors, and tags associated with the behaviors and corresponding usage data under the tags are determined. At the moment, some related use sample data in the actual state of the user can be obtained, and preliminary screening of the sample data is realized.

After determining the relevant use sample data in the actual state of the user, the embodiment of the application may perform analysis based on the use sample data to predict the actual possible needs of the user, and determine what the user actually needs. And secondly, performing secondary screening on the use sample data based on the predicted actual demand, and selecting the use data (namely the data to be analyzed) strongly related to the user demand. Finally, the embodiment of the application analyzes the data to be analyzed, determines the function suitable for the user or the specific content in the function as the content to be recommended, and recommends the content to be recommended to the user.

The embodiment of the application has the following beneficial effects:

1. since the usage data of a user may involve many user behaviors, there is a large difference in correlation between these behaviors. Therefore, if all the usage data are used as sample data to predict the user's needs and analyze the recommended content, some usage data corresponding to almost irrelevant behaviors may be used together, which may result in poor effectiveness of the sample data. And if the user uses the data of the user to be classified in advance, different types of use data are associated with the user behaviors, and then the use data corresponding to the user behaviors are selected as sample data according to the association condition so as to predict the user requirements and analyze the recommended content. This is subject to human subjectivity. The division standards of each person may be different, that is, the accuracy of division cannot be guaranteed, so that the validity of sample data is still poor.

According to the method and the device, the user label and the use sample data under the label are positioned by predicting the user behavior, the user requirement is determined by using the sample data, and then the use data are secondarily screened according to the user requirement. The data to be analyzed screened out at the moment is sample data screened out through user behaviors and requirements, and the correlation degree of the data to be analyzed and the actual requirements of the user is extremely high. And the selection and secondary screening processes of the use data are adaptively realized according to the actual situation of the user, and compared with the full selection of the use data, a large amount of weakly related use data can be reduced, so that the accuracy of final recommendation can be improved. Compared with the method that the usage data are selected by artificially dividing some usage data types and associating behaviors, the method and the device for selecting the usage data can avoid the influence of artificial subjective factors, and enable the selected usage data to be more suitable for the personalized requirements of the user. Therefore, the final recommended content of the embodiment of the application is more accurate.

2. The user portrait is a multi-dimensional label portrait established based on the actual use condition of the user, and meanwhile, the user portrait is more flexible in use. It is not a single-dimensional recommendation, nor a combination of fixed dimensions. The labels in the user portrait are flexibly selected according to the actual user condition, different users and even different operations of the same user can cause different labels of the users selected by the scheme. Therefore, interference of some useless factors can be eliminated, and accuracy is improved. While the user is not a single thin user with only one type of feature. Therefore, a real user can be obtained, and the analysis of the behavior, the prediction of the demand and the recommendation of the content are more accurate and reliable.

3. We are not simple functional recommendations. The content to be recommended can be the function itself or the specific content under the function. The granularity of the recommended content can be finer, and the recommended content is more user-friendly, so that the recommendation effect is better.

The embodiment of the application can be divided into two stages of user portrait and content recommendation according to the practical technical implementation situation, and the following description is respectively given.

Stage one: the user portrays the image.

Fig. 1 shows a flowchart of implementing a user portrait phase in a content recommendation method according to an embodiment of the present application, which is detailed as follows:

s101, obtaining user information of a first user and historical use data of the first user on each function of the platform.

In this embodiment of the application, the first user refers to a user who needs to perform content recommendation. The user information refers to some attribute information of the user. The specific information content contained in the user information may be different according to different platform situations. For example, in some alternative embodiments, the user information may include the gender and age of the user. In other alternative embodiments, the user information may also include the user's gender, age, and shipping address.

The embodiment of the application does not excessively limit the functions of the platform, and can be determined according to the actual application condition. For example, in some alternative embodiments, the functionality of the platform may include any one or more of the following: take-out, group-buying gourmet, group-buying vegetable, online dish market, online supermarket, lodging, leisure entertainment, taxi trip, train ticket or air ticket, buying medicine and online hospital.

The historical use data refers to data of historical use of each function in the platform by a user. Such as which content within the function was clicked, and the specific click time, etc. In some alternative embodiments, the historical usage data may further include a portal type corresponding to the historical usage function of the user. For example by means of a web site or by means of a terminal application, such as an application installed in a mobile phone.

And S102, sampling historical use data of the function to obtain use sample data.

The time span and amount of data may be relatively large due to historical usage. For example, if the user is a multi-year user of the platform, the historical usage data may theoretically have a time span of several years, and the corresponding data amount may be relatively large. However, in practical applications, it is found that the behavior of the user has certain periodicity and regularity. The periodicity and regularity of the behaviors generated under different requirements also have certain difference. Therefore, the data validity is poor for the historical use data which occurs in a long time. If the sample data is directly used as sample data for subsequent user requirement analysis, on one hand, the data processing amount is large, and on the other hand, the analysis accuracy is also reduced.

Therefore, after the historical usage data is acquired in S101, the embodiment of the present application performs data sampling on the historical usage data, and selects data with higher effectiveness as usage sample data. The embodiment of the present application does not excessively limit the specific sampling method, and can be determined by those skilled in the art according to actual needs. For example, in some alternative embodiments, a fixed sampling window length may be set to sample all historical usage data. If the length of the sampling window is set to be one month, historical use data in the last month is intercepted and used as use sample data. Different sampling window lengths can be set for sampling according to the characteristics of each type of historical use data.

In order to analyze the platform function more accurately, and improve the accuracy of the subsequent analysis prediction of the user requirement. In an alternative embodiment of the present application, the functions are classified into four function types, a food dimension, an entertainment dimension, a travel dimension, and a health dimension, according to their characteristics. Wherein the food dimension includes all diet related functions, the entertainment dimension includes functions related to leisure and entertainment, the travel dimension includes functions related to travel and accommodation, and the health dimension includes diet related functions as well as functions related to the user seeing a doctor and taking medicine. The following are examples of functions that four function types may contain respectively:

the food dimensions may comprise: take-out, group-buying gourmet, group-buying vegetable, online dish market, online supermarket.

The entertainment dimension may include: and (5) leisure and entertainment.

The travel dimensions may include: taxi taking, train ticket or air ticket, accommodation, recreation and entertainment.

The health dimension may include: the method comprises the steps of online hospitals, medicine buying, taking out, group-buying of gourmet, online supermarkets, online dish markets and group-buying of dishes.

On the basis, the function types actually included by the platform can be determined according to the function-included conditions of the four function types. For example, assume that in an alternative embodiment the platform includes: taking out, group buying gourmet, leisure entertainment, train ticket or air ticket, because taking out belongs to food dimension and healthy dimension simultaneously, leisure entertainment belongs to the amusement dimension, train ticket or air ticket belongs to the trip dimension, therefore four kinds of functional types of food dimension, amusement dimension, trip dimension and healthy dimension are contained simultaneously to the platform at this moment.

And sampling historical use data contained in the four function types in a sampling window mode. However, because the characteristics of each function type have certain differences, the sampling window length is set according to the natural activity rule of human beings when sampling. So that the lengths of the sampling windows corresponding to the four function types are not identical. The specific sampling method or strategy is as follows:

1. historical usage data for food dimensions. A default value for the length of the sampling window is first set. The default value may be set to one week or one month. Considering the real life, there may be taste variation for users, and there may be difference in the period of taste variation for different users. Therefore, a fixed sampling window length is difficult to satisfy the actual requirements of different users. Based on this, the embodiment of the application performs regular analysis on historical use data of food dimensions of the user in the past half year every month to determine the diet cycle of the user. The meal cycle is then updated to the length of the sampling window for historical usage data for the food dimension.

By way of example, historical usage data for a user's past half-year food dimension may be regularly analyzed on the first day of each month. Determining how long the user typically tastes changes, i.e. the eating period of the user's taste. The meal cycle is then set to the length of the sampling window for the historical usage data for the food dimension. If the analysis result is assumed, the taste of the user has a remarkable change once in two months, namely the drinking period is two months. At this time, the embodiment of the present application may set the sampling window length corresponding to the food dimension to two months. That is, in S102, for the function in the food dimension, the historical usage data of the last two months is intercepted as the corresponding usage sample data.

As an alternative embodiment of the present application. In consideration of the actual life, the taste changes caused by the factors of the user. Some external factors may also have some influence on the taste of the user, thereby causing the taste of the user to change. Such as seasonal changes, sudden changes in weather, and large changes in user geographic location transmissions. To address this situation, the user's dietary cycle is determined by regular analysis of the historical usage data of the user's past half-year food dimension at regular intervals per month as described above. Some trigger conditions are also set in the embodiment of the application. When the trigger conditions are detected, regular analysis is performed on historical use data of food dimensions of the user in the past half year to determine the diet cycle of the user, and the diet cycle is updated to the length of a sampling window of the historical use data of the food dimensions. The specific triggering condition content is not limited in this embodiment, and can be selected or set by a technician according to actual needs. Examples may include, but are not limited to: a seasonal switch (i.e., the last day of each season) is detected, a change in weather for the current day is detected, and the spatial distance of the user's location from the previous location is greater than a distance threshold. The actual value of the distance threshold may be set by the skilled person according to actual requirements, and is not limited herein. Such as may be set at 1000 kilometers.

The food dimension historical use data rule analysis method is not limited too much, and specifically can be selected or set by technical staff according to actual requirements. For example, in some alternative embodiments, some data law analysis algorithms, such as SMCA algorithm and cluster analysis, may be used. In other alternative embodiments, the food may also be classified in advance. And clustering food types contained in the historical use data of the half-year in the user operation according to the occurrence time of the historical use data to determine how often the user has obvious changes to the food type. I.e. to determine the value of the period during which the type of food has changed.

2. Historical usage data for entertainment dimensions. A default value for the sampling window length is first set to one month. At the moment, historical use data in the last month is intercepted and used as use sample data. Meanwhile, the long-term regularity of the interests and hobbies of the user is considered. That is, a person may have different hobbies over a period of time, but a long life/work habit may make some hobbies persistent. Therefore, the embodiment of the application can also analyze historical use data of the entertainment dimension of the past half year and screen out some existing long-term hobbies. And add historical usage data corresponding to these preferences to the usage sample data. The analysis method of the long preference is not limited too much in the embodiment of the application, and the analysis method can be selected or set by technicians according to actual requirements. For example, as an alternative analysis method, the number of times of entertainment items involved in the historical usage data of the entertainment dimension can be counted, and the number of times is larger than a threshold number of times, or the top n entertainment items with the largest number of times are used as the long-time preference of the user. Wherein the number threshold and n may be set at the discretion of the skilled person. Such as the number threshold, may be set to any of 6-10. n may take any value from 1 to 3.

Meanwhile, the embodiment of the application can remove the duplicate of the historical use data of the entertainment dimension, so that the data added correspondingly to the long-term preference can not be repeatedly calculated. If singing is liked and KTV is frequently sent, the KTV is added to the use sample data. However, if the data for removing KTV already exists in the sample data, the inserted data for removing KTV is deleted.

3. Historical usage data for travel dimensions. The method and the device for identifying the travel situation of the user can firstly identify the travel situation of the user and judge whether the user frequently leaves a resident city. The specific determination method may be to determine the resident city of the user according to the setting or location data of the user. And determining the times of leaving the resident city according to the positioning data of the user. If the number of times is larger than the travel number threshold value, the user is determined to frequently leave the resident city, namely the user is an active user. Otherwise, if the number of times is less than or equal to the threshold value of the travel times, the user is determined to leave the resident city infrequently, that is, the user is an inactive user. The effective time range of the positioning data and the specific size of the time threshold are not limited too much here, and can be set by a technician according to actual needs. For example, in some embodiments, the positioning data valid time range is 2 years. That is, at this time, it is determined whether the user frequently leaves the resident city or is an active user according to the positioning data of the user within 2 years. And the number threshold may be set to any one of values 7 to 10.

For active users, travel is often required. Therefore, when historical use data sampling of the travel dimension is performed on the data, the length of a sampling window needs to be long so as to guarantee the effectiveness of the obtained use sample data. On the contrary, for the inactive users, the frequency of travel is very small, and more effective use sample data can be obtained without a very long sampling window length. And excessive redundant data can not be brought, and the calculation amount is saved. Thus, in the present embodiment, two different sampling window lengths are set, and the sampling window length for active users is longer than the sampling window length for inactive users. So as to realize the self-adaptive processing of the actual travel situation of the user. After whether the user is an active user is identified, the corresponding sampling window length is selected, and historical use data of the travel dimension is sampled to obtain corresponding use sample data. The actual values of the lengths of the sampling windows corresponding to the active users and the inactive users are not limited too much, and can be set by technicians according to actual requirements. For example, in some alternative embodiments, the active user sampling window length may be set to any value from 6 months to 12 months, while the inactive user sampling window length may be set to any value from 1 month to 3 months.

To illustrate by way of example, assume that the positioning data is valid for a time period of 2 years, a time threshold of 7, a sampling window length of 6 months for active users, and a sampling window length of 1 month for inactive users. At this time, according to the embodiment of the application, the city A with the largest positioning times of the user can be determined as a resident city according to the positioning data of the user within two years. And counting the times that the positioning data is not the city A within two years. If the number of times is greater than 7, the user is determined to be an active user, and at this time, historical usage data of the travel dimension of the user in the past 6 months is captured and used as usage sample data of the travel dimension. And if the number of times is less than or equal to 7, determining that the user is an inactive user, and capturing historical use data of the travel dimension of the user in the past 1 month at the moment as use sample data of the travel dimension. Therefore, the effectiveness of using the sample data can be guaranteed, more redundant data can be prevented from being introduced, and the calculated amount is increased.

4. Historical usage data for the health dimension. From the above classification of functions, the health dimension encompasses all functions within the food dimension. But differs from the food dimension in that the food dimension is more concerned with the user's dietary preferences. For diet-related functions, the health dimension is concerned about the diet health of the user. Meanwhile, the health dimension also concerns the health condition of the body of the user, so that the health dimension also comprises the function of the user in seeing a doctor and taking a medicine. I.e. the health dimension is different from the focus of food dimension attention. The historical usage data sampling method of the diet-related function in the health dimension may refer to the historical usage data sampling method of the food dimension, and is not described herein again. The correlation between the medical treatment and the audited health condition of the user is extremely high, so that the historical use data sampling of the function in the aspect is to keep all the historical use data. In this case, the application window is not required to be arranged.

In addition, as an alternative embodiment of the present application. It is contemplated that more data is contained within the historical usage data. For example, historical usage data for diet-related functions may include specific foods purchased by the user, time of purchase, price, merchant of sale, and number of purchases. The historical usage data for the medical care medication related functions may include user specific medications, time of purchase, price, sales pharmacy or hospital, and number of purchases. The data has large correlation difference and high discreteness with the physical health of the user. Therefore, if all data are retained and subsequent user requirement analysis is performed, on one hand, the data size is large, and the calculation amount is large, and on the other hand, the difficulty in user requirement analysis is increased due to the discrete data with different relevancy. Therefore, in the embodiment of the present application, data extraction is performed on historical usage data during sampling, which is specifically as follows:

it is important to consider the health of the diet and the type, number, and source of food consumed by the user. Therefore, the embodiment of the application can classify the food into the healthy food and the garbage food, classify and count the food according to the historical use data of the intercepted food related function, and determine the type of the food (namely, the healthy food or the garbage food), the times of each type of the food and the source of the food. Wherein the source is divided into take-out cooked food and self-purchased food material. All cooked foods purchased through the platform are taken-out cooked food sources, and all non-cooked food materials purchased through the platform are self-purchased food materials. The takeout cooked food indicates that the user is mainly for takeout and has lower health degree for the user. And the self-purchase of food materials shows that the user can cook by himself, so that the relative health degree is high.

For the case of medical care medication, the type and number of medications, and the number of uses of the user's medical care and medication related functions in the near term are important. Therefore, the embodiment of the application analyzes the historical use data of the functions related to the medical diagnosis and the medicine purchase to determine the types of the medicines, the times of each type of the medicines and the times of using the functions related to the medical diagnosis and the medicine use of the user within the preset time period. The specific time period of the preset time period can be set by a technician, and is not limited too much here. For example, the current time may be set to within a half year of the end time.

So far, the embodiment of the present application completes sampling of all historical use data of the functions under the four function types, and all use sample data obtained by the function downsampling are the use sample data required in S102.

As an alternative embodiment of the present application. In practical applications, the platform is rich in functions for a single user, but not all functions are used. Therefore, when the historical usage data is sampled, part of the historical usage data may be missing, and normal sampling may not be performed. Resulting in the loss of usage sample data. If the user sample data with the missing data is used for subsequent data processing and the analysis and prediction of the user requirement, the reliability of the analysis and prediction of the requirement is reduced. Based on this, the embodiment of the present application provides a method for filling up historical usage data of a user, which is detailed as follows:

referring to fig. 2, after S101 and before S102, the method further includes:

s201, if the first user has the function to be processed without the historical use data, at least one second user with the historical use data for the function to be processed is found out.

In S101, if the historical usage data of the first user is obtained, it is searched whether a function (hereinafter referred to as a to-be-processed function) exists in the platform function, where the first user does not have corresponding historical usage data. If yes, the first user does not use the function to be processed. At this time, the embodiment of the present application finds out the user who uses the pending function (i.e. the second user) from the users existing in the platform. Specifically, only the user having the historical usage data of the function to be processed needs to be found.

It should be understood that a single user may have one or more functions that have not been used, and thus the pending function in the embodiments of the present application may be one or more functions. Also, each function may be used by one or more users. Therefore, in the embodiment of the present application, the number of the second users found out is uncertain, and may be one or more. The concrete needs to be determined according to actual conditions.

S202, obtaining user information of each second user and historical use data of each second user on each function of the platform.

After the second users are determined, the embodiment of the application may acquire user information of the second users and historical usage data of each function of the platform. The specific operation details, principles, and the like are the same as those of the step S101 of acquiring the user information and the historical usage data of the first user, and therefore, details are not repeated here, and reference may be made to the description of the step S101.

S203, historical use data of the to-be-processed functions of the second users are removed from the acquired historical use data. And calculating the similarity between the first user and each second user according to the user information of the first user, the historical use data of the first user for each function of the platform, the user information of each second user and the historical use data of each second user for each function of the platform after the operation is eliminated.

In the embodiment of the present application, it is considered that the first user does not have historical usage data of the function to be processed. When the user similarity calculation is carried out, the historical use data of the functions to be processed does not have great reference significance. Therefore, before calculating the similarity between the first user and the second user, the historical use data of the second user on the function to be processed is eliminated. To reduce unnecessary computational effort. It should be understood that culling herein does not represent deleting the second user's historical usage data of the pending function, but merely illustrates not using these data to perform the calculation of user similarity.

According to the embodiment of the application, the similarity calculation method of the first user and the second user is not limited too much, and specifically, the similarity calculation method can be selected or set by a technician according to actual requirements. Referring to fig. 3, as an alternative embodiment of the present application, there is provided a similarity calculation method for a first user and a single second user, which is detailed as follows:

s2031, calculating a similarity between the user information of the first user and the user information of the second user.

The method for calculating the similarity of the user information is not limited too much, and can be selected or set by technical personnel according to actual requirements. For example, the method may be implemented by using an algorithm such as cosine similarity, pearson coefficient, or modified cosine similarity, which is not limited herein.

S2032, respectively counting the total using time length and the total using times of each function by the first user, and respectively counting the total using time length and the total using times of each function by the second user except the function to be processed.

S2033, processing the similarity of the user information, the total using time length and the total using times of each function by the first user and the total using time length and the total using times of each function by the second user by using the formula (1), and obtaining the similarity of the first user and the second user:

in the embodiment of the present application, the first user is referred to as user b, and the second user is referred to as user c. In the formula (1), S _bc1 Representing the similarity of user b and user c. N is a radical of _(b) Indicates the set of functions used by user b, N _(c) Indicating a set of functions used by user c (not including the functions to be processed). T is a unit of _bi Indicates the total time length of the use of the ith function by the user b, T _ci Indicating the total duration of use of the ith function by user c. N is a radical of hydrogen _bi Indicates the total number of times of use of the ith function by the user b, N _ci Indicating the total number of uses of the ith function by user c. S _bc2 Indicating the similarity of the user information of user b and user c. | N _(b) I and I N _(c) Respectively represents N _(b) And N _(c) The number of functions involved. | N _(b) |∪|N _(c) I represents N _(b) And N _(c) The sum of the number of functions involved.

Wherein, α is a weighting factor used for adjusting the similarity of the user b and the user c in the function use and the similarity weighting of the user information. Beta is a time factor, and the larger the value of beta is, the greater the influence of the time length difference of the functions used by the two users on the similarity is. The larger the value of the delta quantity factor is, the greater the influence of the degree of difference in the number of times two users use the function on the similarity is. The values α, β and δ can be set by the skilled person as a function of the actual situation and are not overly limited herein.

As an optional embodiment of the present application, in consideration of the search for similar users in the embodiment of the present application, it is more necessary to fill up the missing historical usage data of the first user. Therefore, the importance of the use similarity condition of the functions is higher than that of the similarity condition of the basic condition of the user. Based on this, in the present embodiment, α >0.5 may be set. In addition, as another embodiment of the present application, the relevance between the use duration of the function by the user and the preference of the user for the function is considered to be weaker than the relevance between the use frequency of the function by the user and the preference of the user for the function. Therefore, in the embodiment of the present application, δ > β can be set to increase the influence of the total number of uses of the function on the similarity of the uses.

In the embodiment of the application, the use condition of the function by a user is quantified through two dimensions of the total use duration of the function and the total use times of the function. So that even if the user has not used a function, it can have initial values in two dimensions, i.e., the total usage time and the total number of uses of the function are both 0. Therefore, the functions used by the first user and the functions used by the second user can be effectively quantitatively compared. The function using condition comparison is more comprehensive and reliable. And further, the calculation of the user similarity is more accurate and reliable.

And S204, screening out the second users with the highest similarity. And taking the screened historical use data of the second user on the function to be processed as the historical use data of the first user on the function to be processed.

After determining the similarity between each second user and the first user, the embodiment of the present application selects one second user with the highest similarity. And taking the historical use data of the second user to the function to be processed as the historical use data of the first user to the function to be processed. At this time, the deficiency of the first user to the historical use data of the function to be processed can be filled. Meanwhile, the reliability of the filled historical use data is ensured, and a sample data basis is further provided for the accuracy of analysis and prediction of subsequent user requirements.

S103, determining user tags based on the use sample data to obtain a plurality of user tags corresponding to the first user, and completing user portrait construction. Wherein each user tag has associated therewith one or more functions.

In the embodiment of the application, the method for determining the user label is not limited too much, and the method can be selected or set by technical personnel according to actual requirements. For example, in some alternative embodiments, some user tags and associations between user tags and functions may be preset by a technician. At this time, only the use sample data corresponding to the function needs to be associated with the user tag. In other alternative embodiments, the sample data may be clustered, and the user tag may be labeled based on the clustering result (the user tag at this time belongs to an abstract user tag, which represents a certain abstract characteristic of the user). And then according to the user label corresponding to the use sample data, the function to which the use sample data belongs and the incidence relation of the user label can be determined.

As an alternative embodiment of the present application, on the basis of the above-described embodiments of four functional types, namely, the food dimension, the entertainment dimension, the travel dimension, and the health dimension. When the user portrays the image, the method and the device operate as follows:

and clustering the use sample data under each function type respectively, and marking clustering results to obtain one or more user labels corresponding to each function type respectively.

In the embodiment of the present application, the function classification is made in consideration of the characteristic that the function type itself is a function. Therefore, the corresponding use sample data has strong relevance. And clustering all the use sample data under a single function type by taking the function type as a unit, so that mutual interference when the sample data is clustered under different function types can be avoided. Therefore, the clustering accuracy can be greatly improved, the user label can be more fit with the personal condition of the user, and the user portrait is more accurate and reliable. The embodiment of the present application does not excessively limit the specific clustering method used, and the method can be selected or set by a technician according to the actual situation. For example, algorithms such as two-step clustering, KMeans clustering, or systematic clustering may be used.

Upon completion of operations S101 through S103, the construction of the stage-one user representation is completed.

The user portrait is a multi-dimensional label portrait established based on the actual use condition of the user, so that a real user can be obtained. And more reliable and effective basic data are provided for the subsequent analysis of user behaviors, the prediction of requirements and the recommendation of contents.

And a second stage: and (3) content recommendation:

on the basis of completing user portrait construction, the embodiment of the application can perform a second stage: and (4) recommending the content. Referring to fig. 4, a flowchart of an implementation of a content recommendation phase in the content recommendation method according to an embodiment of the present application is shown, which is detailed as follows:

s401, obtaining operation data of the first user in the platform.

In the embodiment of the present application, the operation data may include two cases:

case 1, the operation data is the usage data of all functions within the platform for a single time by the first user.

Case 2, the operation data is the usage data of all functions in the platform in the process that the first user uses the platform for multiple times recently. The specific number of times can be set by a technician, and can be set to any value from 2 to 5.

For the case 1, the data analysis of the platform used by the user for a single time can be realized, and the method is suitable for some scenes with real-time analysis requirements. If the user needs to use the platform, the user operation is analyzed in real time while responding to the user operation, and the corresponding appropriate content to be recommended is determined.

For the case 2, the data analysis of the platform used by the user within a period of time can be realized, and the actual behavior tendency of the user can be further excavated. For example, according to the operation data of the user on the platform in a period of time, the potential behavior tendency of the user in the period of time is analyzed, and the corresponding appropriate content to be recommended is determined.

According to the embodiment of the application, the actual situation of the operation data is not limited too much, and the operation data can be determined according to the actual application scene.

S402, performing behavior analysis on the first user based on the operation data, predicting the user behavior of the first user, and selecting at least one user tag corresponding to the first user according to the user behavior.

After the operation data is obtained, the embodiment of the application may further analyze the operation data to predict the next possible behavior of the user, that is, which function or functions the user may use next. The embodiment of the application does not excessively limit the specifically used user behavior prediction method, and the user behavior prediction method can be selected or set by technical personnel according to actual requirements. For example, as an alternative embodiment, a user behavior prediction model may be trained in advance. The method comprises the steps of utilizing actual behavior paths of a large number of users as training sample data, utilizing a neural network model to conduct deep learning training, and obtaining a user behavior prediction model capable of predicting subsequent behaviors of the users based on the existing user behavior paths. And processing the function use behaviors of the first user in the operation data based on the model, and predicting the next possible behaviors of the first user. In another alternative embodiment, a GNN model for user behavior prediction may also be pre-trained. And then the trained GNN model is used for realizing the prediction of the next possible behaviors of the first user.

After determining the next possible user behavior of the user, the embodiment of the present application selects all the user tags related to the behavior. As an alternative to the user tag selection, all functions related to the user behavior may be found, and all user tags associated with the functions may be extracted.

As another optional implementation of user tag selection, on the basis of the above-described embodiment in which the food dimension, the entertainment dimension, the travel dimension, and the health dimension are extracted for each function type, respectively, the user tag is extracted. The method and the device for determining the user behavior comprise the steps of firstly determining the function type of the function in the user behavior. And screening all functions contained in the function types, and completely extracting all user tags related to the screened functions. For example, assume that the user behavior is to use a takeaway function. Since the takeout functions belong to the food dimension and the health dimension, all functions contained in the food dimension and the health dimension are determined, and then all user tags associated with the functions are extracted.

It should be understood that the number of predicted user behaviors is not excessively limited in the embodiments of the present application. That is, the user behavior predicted in S402 may be a single behavior or a plurality of behaviors. For example, only the takeout function may be used, or both the takeout function and the group purchase function may be used.

And S403, obtaining the use sample data corresponding to all the selected user tags, and predicting the user requirement of the first user based on the user behavior and the selected use sample data.

User behavior is the manifestation of user requirements, which inherently hides the actual requirements inherent to the user. For example, although the user is using the take-out function and the group buying and buying function, the user may not know what the user is well eating, and the hidden actual demand is to find a delicious food suitable for the taste of the user. Therefore, after all user tags corresponding to the user behaviors are determined, the embodiment of the application screens all the use sample data of all functions under the user tags, and analyzes the user requirements. The actual requirements of the user are determined. The method and the device for partitioning the network data do not limit the partitioning rule required by the user too much, and can be determined according to the actual application condition. For example, in some optional embodiments, the user requirement may be that a technician manually sets a plurality of requirement tags in advance, and the user requirement at this time has a relatively clear meaning of daily life reality. Such as searching for food, cooking, shopping, relaxing body, traveling, going on business, seeing a doctor, etc. In other alternative embodiments, some clusters may be used to cluster the user requirements, where each user requirement obtained by the clustering has a certain common feature data set label. It does not necessarily correspond exactly to the human daily living needs.

The embodiment of the application does not excessively limit the specific analysis and prediction method of the user requirement, and technicians can select or set the method according to actual requirements.

As an optional embodiment of the present application, in order to implement prediction of user requirements, a requirement prediction model capable of predicting user requirements is trained in advance. In corresponding S403, the operation is as follows:

and generating the feature vectors of the functions of the use sample data based on the screened use sample data.

And generating a feature vector of the function corresponding to the user behavior.

And inputting all the obtained feature vectors into a demand forecasting model to obtain the user demand of the first user.

The training operation of the demand forecasting model is as follows:

an initial demand prediction model is preset, the demand prediction model is a neural network model, and a machine learning method is adopted. Alternative machine learning methods include, but are not limited to, such as: the learning method comprises the steps of logistic regression, self-adaptive enhancement, support vector machine, random forest and long and short term memory network.

Collecting a plurality of sets of training data for a plurality of sample users, wherein each set of training data comprises two parts:

part a: the usage data for one or more platform functions by the sample user, and the corresponding usage data for the last time the platform function was used by the sample user.

And b part: and (4) corresponding user requirements after the sample user uses the platform function for the last time. Wherein, a plurality of user requirements are preset by technical personnel, such as searching for food, cooking, shopping, relaxing body, traveling, going on business and seeing a doctor, etc. The sample user only needs to select one or more user requirements according to the actual situation after using the platform function for the last time.

And vectorizing the part a of the training data to obtain a feature vector corresponding to each function in the training data. For training data, the feature vector corresponding to the part a is used as input, the user requirement corresponding to the part b is used as a corresponding label, and an initial requirement prediction model is trained. And obtaining a demand prediction model which can be used for user demand prediction until the demand prediction model meets the preset convergence condition, and completing model pre-training.

As a single user may not have used all of the functionality of the platform. Therefore, if the behavior prediction is performed only based on the fact that the user actually uses the function, the behavior predicted at a large probability is also the function that the user has used is operated again. It is difficult for the behavior prediction itself to jump out of the range of "functions that the user has used". There is still no involvement with functions that are not used by the user. In order to solve the problem, in the embodiment of the present application, for each user behavior, all the corresponding user tags are selected. And will screen out all user's label that elect the use sample data. Therefore, even if the user only wants to perform a certain action, all functions related to the action can be found and processed. So that users unfamiliar with even not used functions at all can be taken into the category of possible recommendations. And then the content to be recommended is recommended, and the effect of improving the utilization rate of the platform function by the user can be really achieved.

S404, screening out the data to be analyzed related to the user requirement from the use sample data.

Since all the relevant usage sample data of the user behavior is acquired in S403, the data content is more, and not all the data has strong correlation with the user requirement.

Assume, for example, that in some alternative embodiments, the user action is to use a take-away function. While assuming that the user demand is to cook. The takeaway function is attributed to the food dimension and the health dimension. At the moment, the embodiment of the application extracts the use sample data of all functions contained in the food dimension and the health dimension for analyzing the user requirements. However, in actual conditions, the functions of hospital online, medicine buying, takeaway and group food buying are relatively poor in correlation with cooking. The functions of the online supermarket, the online dish market and the group buying and buying of dishes relate to food materials, seasonings, kitchenware and the like required for cooking, so that the relevance to cooking is high.

As can be seen from this, if the recommended content suitable for the user is analyzed by directly using the usage sample data acquired in S403, a large amount of usage sample data that is weakly correlated with the user' S demand is involved. This may reduce the accuracy of the final recommended content. Therefore, after the user requirement is determined, the embodiment of the application performs secondary screening on the use sample data, and only the use sample data (namely, the data to be analyzed) strongly related to the user requirement is selected.

The embodiment of the application does not limit the screening method of the data to be analyzed too much. Can be selected or set by a technician according to actual requirements. For example, in some alternative embodiments, the association of user requirements with functions may be preset by a technician. In S404, all functions associated with the user requirement are determined, and then the use sample data of the associated functions is screened from the use sample data and used as the data to be analyzed.

S405, analyzing the data to be analyzed, and determining the content to be recommended corresponding to the first user. The content to be recommended is a function or a functional content.

In the embodiment of the present application, the content that needs to be recommended to the user is referred to as content to be recommended. The content to be recommended may be a function, or may be specific content within the function (i.e., functional content). Therefore, in the embodiment of the application, the data to be analyzed is firstly analyzed, and the functions to which the data to be analyzed belongs and the function contents used by the first user under the functions are determined. And the functions and the function contents are taken as objects to be analyzed, and the contents to be recommended are determined.

For example, it is assumed that, in one embodiment, the data to be analyzed is the usage sample data of functions such as supermarket online, dish market online and group buying. At the moment, the embodiment of the application can determine the functions of the online supermarket, the online dish market and the group buying and buying of dishes. And simultaneously, determining the functional contents used by the user according to the actual use conditions of the user in the data to be analyzed on the functions of the online supermarket, the online dish market and the group buying and buying. If the data to be analyzed are recorded with a supermarket store in an online supermarket, a seafood aquatic product in an online dish store and a hot pot food material in a group buying dish, the functional contents are used by the user. Then the supermarket stores, seafood, aquatic products and hot pot food materials can be selected. Therefore, in this embodiment, the 3 functions and 3 functional contents of the online supermarket, the online dish market, the group buying dish, the supermarket store, the seafood and the hot pot food material are taken as objects, and a total of 6 analyzed objects are obtained.

After the functions and the function contents are determined, the embodiment of the application can analyze the attribute data of each function and function content from the data to be analyzed. Wherein the attribute data includes:

And simultaneously calculating the average value T01 of the average monthly use time of all the functions, the average value N01 of the monthly use times, the average time T02 of all the functions used in the last month and the average times N02. Average value T03 of the average usage time of all the functional contents in the month, average value N03 of the average usage times of the month, average time T04 of the usage of all the functional contents in the last month and average number N04.

And finally, calculating the demand score of the first user for each function and function content according to the attribute data of each function and function content. Wherein, the function demand score F1 is calculated by the following formula (2):

the requirement score F2 of the functional content is calculated by equation (3) as follows:

wherein, a, b, c and d are all weighting coefficients used for adjusting the fraction proportion of the use duration and the use times. And e is an adjusting coefficient used for adjusting the demand score of the functional content, so that the demand score of the functional content is comparable to the demand score of the function. The method has certain timeliness in consideration of user requirements. Therefore, in practical application, c > a and d > b are set, and on the basis, the specific values of a, b, c and d can be set by a skilled person according to the self-setting. Meanwhile, the use duration and the use times of the function content are considered to be less than the functions to which the function content belongs. However, in view of the recommendation effect, if the user uses the function content after recommending the function content, the function to which the function content belongs must be entered first. And thus is advantageous to the user's habit in using the function. Based on this, in the present embodiment, e >1 is set. The specific value of e on the basis of the above can be set by a technician according to the self-setting.

And finally, the function or function content with the highest demand score in the function and function contents is used as the content to be recommended, so that the content to be recommended which is most suitable for the actual demand of the first user can be obtained.

S406, pushing the content to be recommended to the first user.

The method and the time for pushing the content to be recommended are not limited too much. Can be freely set by technicians according to actual requirements. For example, the push mode may be that the user preferentially displays the content to be recommended when using the platform. Such as preferentially displaying in a highlighted area of a first page of the platform, displaying in a search interface of the platform, or increasing the display sequence of the contents to be recommended in the platform, such as advancing from the second page to the first page. Or the user pushes the content to be recommended to the user in a form of pop-up window or the like in the process of using the platform, so as to attract the user to use the content to be recommended. The push time may be real-time push in the process that the user uses the platform, or push when the user enters the platform next time.

According to the embodiment of the application, the user label and the use sample data under the label are positioned by predicting the user behavior, and the user requirement is determined by using the sample data. And then the use data is screened secondarily according to the user requirements. The data to be analyzed screened out at the moment is sample data screened out through the user behavior and the user demand, and the correlation degree of the data to be analyzed and the actual demand of the user is extremely high. And the selection and secondary screening process of the used sample data are adaptively realized according to the actual condition of the user, and compared with the full selection of the used sample data, a large amount of weakly related used sample data can be reduced, so that the accuracy of final recommendation can be improved. Compared with the method for selecting the use data by artificially dividing some use data types and associating behaviors, the method and the device for selecting the use data can avoid the influence of artificial subjective factors, so that the selected use data can better meet the individual requirements of users. Therefore, the final recommended content of the embodiment of the application is more accurate.

The method and the device for recommending the functions or the function contents which may be required by the user can recommend the functions or the function contents to the user. This function may be used by the user or may be a function or content of a function that the user has not used but may need. Such recommendation is more suitable for the actual situation of the user, and the useless functions of the user are not recommended forcibly in order to attract the user to use the new functions, so that personalized recommendation of the actual situation of the user can be realized. When the contents to be recommended are functions or functional contents that the user has tried, since these recommended contents are contents that the user needs. Therefore, the user can not feel the discomfort of the user and can be provided with convenience. The user becomes familiar with these functions after attempting to use them or their content. And then the utilization rate of the platform function by the user is improved, and the user experience of the platform is improved.

In addition, we are not simple functional recommendations. The content to be recommended may be the function itself or specific function content under the function (the user is using the function itself while operating the function content). The granularity of the recommended content can be finer, and the recommended content is more user-friendly, so that the recommendation effect is better.

Fig. 6 shows a block diagram of a content recommendation device provided in an embodiment of the present application, which corresponds to the method in the foregoing embodiment, and only shows a part related to the embodiment of the present application for convenience of description. The content recommendation apparatus illustrated in fig. 6 may be an execution subject of the content recommendation method provided in the first embodiment.

Referring to fig. 5, the content recommendation apparatus includes:

and a data acquiring module 51, configured to acquire operation data of the first user in the platform.

And the tag selecting module 52 is configured to perform behavior analysis on the first user based on the operation data, predict user behavior of the first user, and select at least one user tag from the user image of the first user according to the user behavior.

The demand forecasting module 53 is configured to obtain usage sample data corresponding to all the selected user tags, and forecast the user demand of the first user based on the user behavior and the usage sample data. The historical use data of the first user on the functions in the platform is recorded in the use sample data.

And the data screening module 54 is configured to screen data to be analyzed related to the user requirement from the use sample data.

And the content determining module 55 is configured to analyze the data to be analyzed, and determine the content to be recommended corresponding to the first user. The content to be recommended is a function or a functional content.

Further, a content determination module, comprising:

and the object determining module is used for analyzing the data to be analyzed and determining the functions to which the data to be analyzed belongs and the function contents used by the first user under the functions.

The data analysis module is used for analyzing the attribute data of each determined function and function content from the data to be analyzed, and the attribute data comprises:

the method comprises the following steps of enabling the average monthly use duration T11 of the function, enabling the average monthly use frequency N11 of the function, enabling the total monthly use duration T12 and the total number N12 of the function to be in the last month, enabling the average monthly use duration T21 of the function content, the average monthly use frequency N21 of the function content, and enabling the total monthly use duration T22 and the total number N22 of the function content to be in the last month.

And calculating the average value T01 of the monthly average use time of all the determined functions, the average value N01 of the monthly average use times, the average time T02 of all the determined functions in the last month and the average times N02 of all the determined functions in the last month. Average value T03 of the average usage time of all the functional contents in the month, average value N03 of the average usage times of the month, average time T04 of the usage of all the functional contents in the last month and average number N04.

And the demand calculating module is used for calculating the demand scores of all the functions according to the following formula.

And the content selection module is used for taking the function or the function content with the highest demand score as the content to be recommended corresponding to the first user.

Further, the content recommendation apparatus further includes:

the first information acquisition module is used for acquiring the user information of the first user and the historical use data of the first user on each function in the platform.

And the sampling module is used for sampling the historical use data of each function to obtain use sample data.

And the user portrait module is used for determining the user tags based on the use sample data, obtaining a plurality of user tags corresponding to the first user and completing user portrait construction of the first user. Wherein each user tag has associated therewith one or more functions.

Further, the content recommendation apparatus further includes:

the user searching module is used for searching at least one second user with historical use data for the function to be processed when the first user has the function to be processed without the historical use data.

And the second information acquisition module is used for acquiring the user information of each second user and historical use data of each function in the platform.

And the data removing module is used for removing the historical use data of the to-be-processed function of each second user from the historical use data of the second users.

And the similarity calculation module is used for calculating the similarity between the first user and each second user according to the user information of the first user, the historical use data of the first user for each function of the platform, the user information of each second user and the historical use data of each second user for the function in the platform after the operation is eliminated.

And the data filling module is used for screening out the second user with the highest similarity and taking the screened historical use data of the second user on the function to be processed as the historical use data of the first user on the function to be processed.

Further, a similarity calculation module comprising:

and the first similarity module is used for calculating the similarity between the user information of the first user and the user information of the second user.

And the counting module is used for respectively counting the total using time length and the total using times of each function by the first user and respectively counting the total using time length and the total using times of each function by the second user except the function to be processed.

A second similarity module for calculating the similarity between the first user and the second user by using the following formula:

here, the first user is referred to as user b, and the second user is referred to as user c. S. the _bc1 Representing the degree of similarity of user b and user c, N _(b) Indicates the set of functions used by user b, N _(c) Indicating that user c has used a set of functions other than the pending function, T _bi Indicates the total time length of the use of the ith function by the user b, T _ci Indicates the total time of using the ith function by the user c, N _bi Indicates the total number of times of use of the ith function by the user b, N _ci Indicates the total number of times of use of the ith function by the user c, S _bc2 Representing the user information similarity of user b and user c, | N _(b) I and | N _(c) Respectively represents N _(b) And N _(c) Number of functions involved, | N _(b) |∪|N _(c) I represents N _(b) And N _(c) The sum of the number of functions involved.

Further, a module is employed comprising:

and the function classification module is used for classifying the functions in the platform into four functional types, namely a food dimension, an entertainment dimension, a trip dimension and a health dimension. Wherein the food dimension includes diet-related functions, the entertainment dimension includes recreational-related functions, the travel dimension includes travel and accommodation-related functions, and the health dimension includes diet-related functions as well as user-seeing and taking medicine-related functions.

And the strategy acquisition module is used for acquiring different sampling strategies corresponding to each function type.

And the sampling submodule is used for sampling the historical use data of the functions in each function type according to the sampling strategy respectively corresponding to each function type to obtain the use sample data.

The process of implementing each function by each module in the content recommendation device provided in the embodiment of the present application may specifically refer to the description of the embodiments of fig. 1 to 4 and other method embodiments, and is not described herein again.

It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

It should also be understood that the term "and/or" as used in this specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items and includes such combinations.

As used in this specification and the appended claims, the term "if" may be interpreted contextually as "when", "upon" or "in response to a determination" or "in response to a detection". Similarly, the phrase "if it is determined" or "if a [ described condition or event ] is detected" may be interpreted contextually to mean "upon determining" or "in response to determining" or "upon detecting [ described condition or event ]" or "in response to detecting [ described condition or event ]".

Furthermore, in the description of the present application and the appended claims, the terms "first," "second," "third," and the like are used for distinguishing between descriptions and not necessarily for describing a relative importance or importance. It will also be understood that, although the terms first, second, etc. may be used herein to describe various elements in some embodiments of the application, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first table may be named a second table, and similarly, a second table may be named a first table, without departing from the scope of various described embodiments. The first table and the second table are both tables, but they are not the same table.

Reference throughout this specification to "one embodiment" or "some embodiments," or the like, means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," or the like, in various places throughout this specification are not necessarily all referring to the same embodiment, but rather "one or more but not all embodiments" unless specifically stated otherwise. The terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless expressly specified otherwise.

The content recommendation method provided by the embodiment of the application can be applied to terminal devices such as a mobile phone, a tablet computer, a wearable device, a vehicle-mounted device, an Augmented Reality (AR)/Virtual Reality (VR) device, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a Personal Digital Assistant (PDA), and the like, and the specific type of the terminal device is not limited at all in the embodiment of the application.

For example, the terminal device may be a Station (ST) in a WLAN, and may be a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA) device, a handheld device with Wireless communication capability, a computing device or other processing device connected to a Wireless modem, a vehicle mounted device, a vehicle networking terminal, a computer, a laptop, a handheld communication device, a handheld computing device, a satellite Wireless device, a Wireless modem card, a Set Top Box (STB), a Customer Premises Equipment (CPE), and/or other devices for communicating over a Wireless system and a next generation communication system, such as a Mobile terminal in a 5G Network or a Mobile terminal in a future evolved Public Land Mobile Network (PLMN) Network, and the like.

By way of example and not limitation, when the terminal device is a wearable device, the wearable device may also be a generic term for intelligently designing daily wearing by applying a wearable technology, developing wearable devices, such as glasses, gloves, watches, clothes, shoes, and the like. A wearable device is a portable device that is worn directly on the body or integrated into the clothing or accessories of the user. The wearable device is not only a hardware device, but also realizes powerful functions through software support, data interaction and cloud interaction. The general formula smart machine of wearing includes that the function is complete, the size is big, can not rely on the smart mobile phone to realize complete or partial function, like intelligent wrist-watch or intelligent glasses etc to and only be absorbed in a certain class of application function, need use like the smart mobile phone cooperation with other equipment, like all kinds of intelligent bracelet, intelligent ornament etc. that carry out the sign monitoring.

Fig. 6 is a schematic structural diagram of a terminal device according to an embodiment of the present application. As shown in fig. 6, the terminal device 6 of this embodiment includes: at least one processor 60 (only one shown in fig. 6), a memory 61, said memory 61 having stored therein a computer program 62 executable on said processor 60. The processor 60, when executing the computer program 62, implements the steps in the various content recommendation method embodiments described above, such as steps 101 to 103 shown in fig. 1. Alternatively, the processor 60, when executing the computer program 62, implements the functions of the modules/units in the above-mentioned device embodiments, such as the functions of the modules 51 to 55 shown in fig. 5.

The terminal device 6 may be a desktop computer, a notebook, a palm computer, a cloud server, or other computing devices. The terminal device may include, but is not limited to, a processor 60, a memory 61. It will be appreciated by those skilled in the art that fig. 6 is merely an example of a terminal device 6 and does not constitute a limitation of the terminal device 6 and may include more or less components than those shown, or some components may be combined, or different components, for example the terminal device may also include an input transmitting device, a network access device, a bus, etc.

The Processor 60 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory 61 may in some embodiments be an internal storage unit of the terminal device 6, such as a hard disk or a memory of the terminal device 6. The memory 61 may also be an external storage device of the terminal device 6, such as a plug-in hard disk provided on the terminal device 6, smart Media Card (SMC), secure Digital (SD) Card, flash memory Card (Flash Card), and the like. Further, the memory 61 may also include both an internal storage unit and an external storage device of the terminal device 6. The memory 61 is used for storing an operating system, an application program, a BootLoader (BootLoader), data, and other programs, such as program codes of the computer programs. The memory 61 may also be used to temporarily store data that has been transmitted or is to be transmitted.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit may be implemented in the form of hardware, or may also be implemented in the form of a software functional unit.

The embodiment of the present application further provides a terminal device, where the terminal device includes at least one memory, at least one processor, and a computer program that is stored in the at least one memory and is executable on the at least one processor, and when the processor executes the computer program, the terminal device is enabled to implement the steps in any of the method embodiments.

The embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the steps in the above-mentioned method embodiments.

The embodiments of the present application provide a computer program product, which when running on a terminal device, enables the terminal device to implement the steps in the above method embodiments when executed.

The integrated modules/units, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow in the method of the embodiments described above can be realized by a computer program, which can be stored in a computer-readable storage medium and can realize the steps of the embodiments of the methods described above when the computer program is executed by a processor. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, usb disk, removable hard disk, magnetic disk, optical disk, computer Memory, read-Only Memory (ROM), random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution medium, and the like.

In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the embodiments of the present application, and they should be construed as being included in the present application.

Claims

1. A content recommendation method, comprising:

acquiring operation data of a first user in a platform; the operation data is the use data of the first user on all functions in the platform once, or the operation data is the use data of the first user on all functions in the platform in the process of using the platform for multiple times recently;

performing behavior analysis on the first user based on the operation data, predicting user behavior of the first user, and selecting at least one user tag from the user portrait of the first user according to the user behavior;

obtaining the selected use sample data corresponding to all the user tags, and predicting the user requirement of the first user based on the user behavior and the use sample data; historical use data of the first user on functions in the platform are recorded in the use sample data; historical use data, data of historical use of each function in the platform for the first user;

screening out data to be analyzed related to the user requirements from the use sample data;

analyzing the data to be analyzed to determine the content to be recommended corresponding to the first user; the content to be recommended is a function or a functional content;

prior to said selecting at least one user tag from said user representation of said first user in accordance with said user behavior, further comprising:

acquiring user information of the first user and historical use data of the first user on each function in the platform;

sampling historical use data of each function to obtain the use sample data; the functions are divided into food dimension, entertainment dimension, travel dimension and health dimension; setting different sampling window lengths corresponding to historical use data of different functions for sampling;

determining the user tags based on the use sample data to obtain a plurality of user tags corresponding to the first user, and completing user portrait construction of the first user; wherein each of said user tags has associated therewith one or more functions;

after the obtaining of the user information of the first user and the historical use data of the first user on each function in the platform, the sampling of the historical use data of each function is performed, and before the use sample data is obtained, the method further includes:

if the first user has a function to be processed without historical use data, at least one second user with the historical use data for the function to be processed is found out;

acquiring user information of each second user and historical use data of each function in the platform;

historical use data of each second user on the function to be processed is removed from the historical use data of the second users;

calculating the similarity between the first user and each second user according to the user information of the first user, the historical use data of the first user for each function of the platform, the user information of each second user and the historical use data of each second user for the function in the platform after the elimination operation;

screening out the second user with the highest similarity, and taking the screened historical use data of the second user on the function to be processed as the historical use data of the first user on the function to be processed;

the calculating the similarity between the first user and each second user according to the user information of the first user, the historical use data of the first user for each function of the platform, the user information of each second user, and the historical use data of each second user for the function in the platform after the removing operation includes:

calculating the similarity between the user information of the first user and the user information of the second user;

respectively counting the total use time length and the total use times of each function by the first user, and respectively counting the total use time length and the total use times of each function by the second user except the function to be processed;

the first user is called a user b, and the second user is called a user c; s _bc1 Represents user b andsimilarity of users c, N _(b) Indicating the set of functions used by user b, N _(c) Indicating that user c has used a set of functions, T, other than the pending function _bi Indicates the total time length of the use of the ith function by the user b, T _ci Indicates the total time of using the ith function by the user c, N _bi Indicates the total number of times of use of the ith function by the user b, N _ci Indicates the total number of times of use of the ith function by the user c, S _bc2 Representing the user information similarity of user b and user c, | N _(b) I and I N _(c) Respectively represents N _(b) And N _(c) Number of functions involved, | N _(b) |∪|N _(c) I represents N _(b) And N _(c) The sum of the number of functions involved;

2. The content recommendation method according to claim 1, wherein the analyzing the data to be analyzed to determine the content to be recommended corresponding to the first user includes:

analyzing the data to be analyzed, and determining functions to which the data to be analyzed belongs and functional contents used by the first user under the functions;

the method comprises the following steps that the monthly average use duration T11 of the function, the monthly average use frequency N11 of the function, the total use duration T12 and the total use frequency N12 of the function in the last month, the monthly average use duration T21 of the function content, the monthly average use frequency N21 of the function content, and the total use duration T22 and the total use frequency N22 of the function content in the last month;

calculating the determined average value T01 of the month average use time length of all the functions, the determined average value N01 of the number of the month average use times, the determined average time length T02 of all the functions in the last month and the determined average number N02 of all the functions in the last month, calculating the determined average value T03 of the month average use time length of all the function contents, the determined average value N03 of the month average use times, calculating the determined average time length T04 of all the function contents in the last month and calculating the determined average number N04 of the month average use time length of all the function contents in the last month;

calculating the demand score of each function according to the following formula;

calculating the demand score of each functional content according to the following formula;

f1 is a demand score of a function, F2 is a demand score of a function content, a, b, c and d are weighting coefficients used for adjusting the score proportion of the using duration and the using times, e is an adjusting coefficient used for adjusting the demand score of the function content, and c > a, d > b and e >1;

3. The content recommendation method according to claim 1, wherein said sampling historical usage data of each function to obtain said usage sample data comprises:

the platform has four functional types; wherein the food dimension comprises diet-related functions, the entertainment dimension comprises recreational-related functions, the travel dimension comprises travel-and accommodation-related functions, and the health dimension comprises diet-related functions and user-visit and medication-related functions;

acquiring different sampling strategies corresponding to each function type;

and according to the sampling strategy respectively corresponding to each function type, respectively sampling historical use data of the functions in each function type to obtain the use sample data.

4. A content recommendation apparatus characterized by comprising:

the data acquisition module is used for acquiring operation data of a first user in the platform; the operation data is the use data of all functions of the first user in the platform for a single time, or the operation data is the use data of all functions of the first user in the platform in the process that the first user uses the platform for multiple times recently;

the tag selection module is used for performing behavior analysis on the first user based on the operation data, predicting the user behavior of the first user, and selecting at least one user tag from the user portrait of the first user according to the user behavior;

the demand forecasting module is used for acquiring the use sample data corresponding to all the selected user tags and forecasting the user demand of the first user based on the user behavior and the use sample data; historical use data of the first user on functions in the platform are recorded in the use sample data; historical use data, data of each function used by the first user in the platform;

the data screening module is used for screening the data to be analyzed related to the user requirements from the use sample data;

the content determining module is used for analyzing the data to be analyzed and determining the content to be recommended corresponding to the first user; the content to be recommended is a function or a functional content;

the first information acquisition module is used for acquiring the user information of the first user and historical use data of the first user on each function in the platform;

the sampling module is used for sampling historical use data of each function to obtain the use sample data; the functions are divided into a food dimension, an entertainment dimension, a trip dimension and a health dimension; setting different sampling window lengths corresponding to historical use data of different functions for sampling;

the user portrait module is used for determining the user tags based on the use sample data to obtain a plurality of user tags corresponding to the first user and completing user portrait construction of the first user; wherein each of said user tags has associated therewith one or more functions;

the user searching module is used for searching at least one second user with historical use data for the function to be processed when the first user has the function to be processed without the historical use data;

the second information acquisition module is used for acquiring the user information of each second user and historical use data of each function in the platform;

the data removing module is used for removing the historical use data of the to-be-processed function of each second user from the historical use data of the second users;

the similarity calculation module is used for calculating the similarity of the first user and each second user according to the user information of the first user, the historical use data of the first user for each function of the platform, the user information of each second user and the historical use data of each second user for the function in the platform after the operation is eliminated;

the similarity calculation module includes:

the first similarity module is used for calculating the similarity between the user information of the first user and the user information of the second user;

the statistical module is used for respectively counting the total using time length and the total using times of each function by the first user, and respectively counting the total using time length and the total using times of each function by the second user except the function to be processed;

wherein, the first user is called user b, and the second user is called user c; s _bc1 Representing the degree of similarity of user b and user c, N _(b) Indicating the set of functions used by user b, N _(c) Indicating that user c has used a set of functions other than the pending function, T _bi Represents the total usage time of the ith function by the user b, T _ci Indicates the total time of using the ith function by the user c, N _bi Indicates the total number of times of use of the ith function by the user b, N _ci Indicates the total number of times of use of the ith function by the user c, S _bc2 Representing the user information similarity of user b and user c, | N _(b) I and | N _(c) Respectively represents N _(b) And N _(c) Number of functions involved, | N _(b) |∪|N _(c) | denotes N _(b) And N _(c) The sum of the number of functions involved;

5. The content recommendation device of claim 4, wherein the content determination module comprises:

the object determining module is used for analyzing the data to be analyzed and determining functions to which the data to be analyzed belongs and function contents used by the first user under the functions;

the data analysis module is used for analyzing the determined attribute data of each function and the function content from the data to be analyzed, and the attribute data comprises:

the method comprises the following steps that the average monthly use duration T11 of the function, the average monthly use times N11 of the function, the total duration T12 and the total times N12 of the function use in the last month, the average monthly use duration T21 of the function content, the average monthly use times N21 of the function content, and the total duration T22 and the total times N22 of the function use in the last month;

the demand calculating module is used for calculating the demand scores of all the functions according to the following formula;

wherein F1 is the demand score of the function, F2 is the demand score of the function content, a, b, c and d are weighting coefficients used for adjusting the score proportion of the use duration and the use times, e is an adjusting coefficient used for adjusting the demand score of the function content, and c > a, d > b and e >1;

6. A terminal device, characterized in that the terminal device comprises a memory, a processor, a computer program being stored on the memory and being executable on the processor, the processor implementing the steps of the method according to any of claims 1 to 3 when executing the computer program.

7. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 3.