JP7516861B2 - Direct mail creation support device, direct mail creation support method, and program - Google Patents

Description

The present invention relates to a direct mail creation support device, a direct mail creation support method, and a program.

In recent years, direct mail has come into widespread use as a method of sales promotion. A known prior art for creating direct mail is a technique for selecting a pre-registered print pattern according to the attributes of a customer (see, for example, Patent Document 1).

JP 2006-48413 A

However, direct mail is complex and diverse, with changes to the shape, type, content, and position of the mailing. For this reason, it has been difficult with the conventional technology described above to propose optimal conditions for such complex and diverse direct mail.

The present invention has been made to solve the above problems, and its purpose is to provide a direct mail creation support device, a direct mail creation support method, and a program that can propose more optimal conditions for direct mail.

In order to solve the above problem, one aspect of the present invention includes a condition acquisition unit that acquires attribute information indicating attributes related to a target of a direct mail as an input condition, an analytical model that estimates an open rate generated by machine learning from learning data of past results in which the attribute information , element information indicating physical elements related to the medium of the direct mail, which is a mail item, and open information indicating whether the direct mail has been opened are associated with each other, an estimation processing unit that estimates the open rate based on the input condition acquired by the condition acquisition unit, and a pre-opening process that recommends a pre-opening process for the input condition based on the open rate estimated by the estimation processing unit. and an optimization processing unit that determines conditions for the above-mentioned elements , wherein the input conditions are attribute information selected from a predetermined number of the attribute information, the estimation processing unit estimates the open rate based on the input conditions, element information selected from a predetermined number of the element information as conditions for the element, and the analytical model, and the optimization processing unit changes the conditions for the element to have the estimation processing unit estimate the open rate, and determines the condition of the element having the highest open rate among the changed conditions for the element as the recommended condition for the element.

In one aspect of the present invention, in the direct mail creation support device described above, the optimization processing unit executes a first process of selecting conditions for a predetermined number of the elements, and selecting the condition for the element with the highest open rate from among the conditions for the predetermined number of the elements, and a second process of selecting conditions for the predetermined number of the elements by changing some elements of the conditions for the selected element with the highest open rate, and selecting the condition for the element with the highest open rate from among the conditions for the predetermined number of the elements, and repeats the second process a predetermined number of times to determine the condition for the element with the highest open rate.

In one aspect of the present invention, the direct mail creation support device is characterized in that it includes a learning processing unit that generates the analytical model from the learning data by machine learning.

In one aspect of the present invention, in the above direct mail creation support device, the learning data includes action information regarding actions taken in response to the direct mail, the analysis model estimates the open rate and an action rate indicating the probability of an action being taken in response to the direct mail, and the optimization processing unit determines the conditions of the elements recommended for the input conditions based on the open rate and action rate estimated by the estimation processing unit.

Also , one aspect of the present invention includes a condition acquisition step in which a condition acquisition unit acquires attribute information indicating attributes related to a target of a direct mail as an input condition, an estimation processing step in which an estimation processing unit estimates an open rate based on an analytical model generated by machine learning from learning data of past results in which the attribute information, element information indicating physical elements related to a medium of the direct mail, and open information indicating whether the direct mail has been opened are associated with each other, and the input condition acquired by the condition acquisition step, and an optimization processing unit estimates an open rate of the element recommended for the input condition based on the open rate estimated by the estimation processing step. and an optimization processing step for determining conditions, wherein the input conditions are attribute information selected from a predetermined plurality of the attribute information, and in the estimation processing step, the estimation processing unit estimates the open rate based on the input conditions, element information selected from a predetermined plurality of the element information as conditions for the element, and the analysis model, and in the optimization processing step, the optimization processing unit changes the conditions for the element to cause the estimation processing unit to estimate the open rate, and determines the condition of the element having the highest open rate among the changed conditions for the element as the recommended condition for the element.

In one aspect of the present invention, a computer- implemented method includes a condition acquisition step of acquiring attribute information indicating attributes related to a target of a direct mail as an input condition, an analysis model for estimating an open rate generated by machine learning from learning data of past results in which the attribute information, element information indicating physical elements related to the medium of the direct mail, which is a mail item, including the type of delivery medium of the direct mail and written contents, and open information indicating whether the direct mail has been opened are associated with each other, and an estimation processing step of estimating the open rate based on the input condition acquired by the condition acquisition step, and the input condition acquired by the condition acquisition step. and an optimization processing step of determining conditions of the element recommended for the input conditions based on an open rate, the input conditions being attribute information selected from a predetermined plurality of the attribute information; in the estimation processing step, the open rate is estimated based on the input conditions, element information selected from a predetermined plurality of the element information as conditions of the element, and the analytical model; in the optimization processing step, the conditions of the element are changed to estimate the open rate in the estimation processing step; and the condition of the element having the highest open rate among the changed conditions of the element is determined as the recommended condition of the element.

The present invention makes it possible to propose more optimal conditions for direct mail.

1 is a block diagram showing an example of a DM creation support system according to a first embodiment. 4 is a diagram showing an example of items of DM attributes in the first embodiment; FIG. FIG. 4 is a diagram illustrating an example of element items according to the first embodiment. FIG. 4 is a diagram illustrating an example of data in a learning data storage unit according to the first embodiment. 5A and 5B are diagrams illustrating an example of input conditions and output results according to the first embodiment. 5 is a flowchart illustrating an example of a learning process according to the first embodiment. 10 is a flowchart illustrating an example of an optimization process according to the first embodiment. FIG. 11 is a block diagram showing an example of a DM creation support device according to a second embodiment.

The direct mail creation support device and direct mail creation support method according to one embodiment of the present invention will be described below with reference to the drawings.

[First embodiment]
1 is a block diagram showing an example of a direct mail creation support system 100 according to a first embodiment. In the following description, direct mail may be abbreviated as DM.

As shown in FIG. 1, the DM creation support system 100 includes a DM creation support device 1 and a model generating device 2 .
The model generation device 2 also generates an analysis model for estimating an open rate and an action rate from performance data of past direct mail deliveries. The model generation device 2 includes a storage unit 21 and a control unit 22.

The storage unit 21 stores various types of information used by the model generation device 2. The storage unit 21 includes, for example, a training data storage unit 211 and an analytical model storage unit 212.
The learning data storage unit 211 stores learning data in which DM attribute information, element information, and at least opening information are associated with each other in the past direct mail sending results.

Here, the DM attribute information is attribute information that indicates attributes related to the target of the direct mail. The target of the direct mail includes the recipient of the direct mail, the target product of the direct mail, etc. The items of the DM attribute include, for example, the items shown in FIG. 2.

FIG. 2 is a diagram showing an example of DM attribute items in this embodiment.
In FIG. 2, the "DM attribute classification" includes items such as "industry type,""reply,""individual/corporation,""product,""occupation,""gender,""age," and "purpose."

In addition, the DM attribute information for each item is set to "1" if it is selected as an input condition, and is set to "0" if it is not selected as an input condition. The learning data storage unit 211 stores multiple DM attribute information items with "0" or "1" set as input conditions in association with each other.

The element information is information that indicates the physical elements related to the direct mail medium, and includes, for example, the type of delivery medium, such as a postcard, and the contents written on the mail. The element items include, for example, the items shown in FIG. 3.

FIG. 3 is a diagram showing an example of element items in this embodiment.
In FIG. 3, the terms separated by "_" in the heading of each element indicate the classification (major classification, medium classification, minor classification, etc.) of each element.

In addition, the element information for each item is set to "1" if it is selected as an element condition, and is set to "0" if it is not selected as an element condition. The learning data storage unit 211 stores multiple pieces of element information with "0" or "1" set as element conditions in association with each other.

FIG. 4 is a diagram showing an example of data in the learning data storage unit 211 in this embodiment.
4, the learning data storage unit 211 stores, as learning data, a number of samples of past performance data in which a number of pieces of DM attribute information, a number of pieces of element information, and an "opening flag" are associated with each other. Here, the "opening flag" is an example of opening information that indicates whether or not a direct mail has been opened, and when it is "1", it indicates that it has been opened, and when it is "0", it indicates that it has not been opened.

The learning data also includes an "action flag," which indicates whether or not the intended recipient took any action in response to the direct mail. When the "action flag" is "1," it indicates that some action was taken, and when it is "0," it indicates that no action was taken.

Additionally, "A summary_reply_web", "A summary_reply_phone", and "A summary_reply_mail" indicate the type of action, and when each is "1", it indicates that the action has been taken. "Action flag", "A summary_reply_web", "A summary_reply_phone", and "A summary_reply_mail" are examples of action information regarding actions taken in response to direct mail.

In this embodiment, the learning data storage unit 211 is assumed to have previously stored therein learning data based on past performance, such as that shown in FIG. 4.

1, the analytical model storage unit 212 stores an analytical model generated by machine learning based on the above-mentioned learning data. Here, the analytical model is, for example, a learning model that estimates the open rate and the action rate from the above-mentioned input conditions and element conditions.
Here, the open rate indicates the probability that a direct mail piece is opened, and the action rate indicates the probability that an action is taken in response to the direct mail piece.

The control unit 22 is, for example, a processor including a CPU (Central Processing Unit) and generally controls the model generation device 2. The control unit 22 includes a learning data acquisition unit 221 and a learning processing unit 222.

The learning data acquisition unit 221 acquires learning data on past performance that associates multiple DM attribute information (input conditions), multiple element information (element conditions), opening information, and action information as shown in FIG. 4 described above, which is stored in the learning data storage unit 211.

The learning processing unit 222 generates an analysis model that estimates the open rate and the action rate by machine learning from the learning data acquired by the learning data acquisition unit 221. The learning processing unit 222 generates, as an analysis model, a model that calculates the probability that the open flag and the action flag, which are the objective variables, will be "1", using, for example, LightGBM, a machine learning framework. Here, LightGBM is a gradient boosting modeling method based on a decision tree.

In addition, the learning processing unit 222 stores the generated analytical model in the analytical model storage unit 212.
The control unit 22 outputs the analytical model stored in the analytical model storage unit 212 to the outside in response to a request from an external device (for example, a direct mail creation support device, etc.).

The DM creation support device 1 (an example of a direct mail creation support device) uses DM attribute information selected from a predetermined number of pieces of DM attribute information as input conditions, and outputs information on elements recommended for direct mail for the input conditions as output results. The DM creation support device 1 also includes a memory unit 11 and a control unit 12.

The memory unit 11 stores various information used by the DM creation support device 1. The memory unit 11 includes an analysis model memory unit 111, an input condition memory unit 112, an element condition memory unit 113, and an estimation result memory unit 114.

The analytical model storage unit 111 stores the analytical model generated by the learning processing unit 222 of the model generation device 2 described above.
The input condition storage unit 112 stores the above-mentioned input conditions (DM attribute information selected from a plurality of predetermined DM attribute information).

The element condition storage unit 113 stores element conditions (element information selected from a plurality of predetermined element information). The element condition storage unit 113 stores, for example, 100 types (a predetermined number) of element conditions.
The estimation result storage unit 114 stores the open rate and the action rate for the above-mentioned 100 (predetermined number) element conditions based on the above-mentioned analytical model.

The control unit 12 is, for example, a processor including a CPU and generally controls the DM creation support device 1. The control unit 12 stores the analytical model output by the above-mentioned model generation device 2 in the analytical model storage unit 111. The control unit 12 includes a condition acquisition unit 121, an estimation processing unit 122, and an optimization processing unit 123.

The condition acquisition unit 121 acquires DM attribute information (attribute information) indicating attributes related to the target of the direct mail as an input condition. The condition acquisition unit 121 acquires the information from outside as a force condition and stores it in the input condition storage unit 112.

The estimation processing unit 122 estimates the open rate and action rate based on the analytical model stored in the analytical model storage unit 111 and the input conditions acquired by the condition acquisition unit 121. For example, the estimation processing unit 122 estimates 100 open rates and action rates using an analytical model from the input conditions stored in the input condition storage unit 112 and each of 100 (a predetermined number) element conditions stored in the element condition storage unit 113. The estimation processing unit 122 stores the estimation results (open rate and action rate) in the estimation result storage unit 114.

The optimization processing unit 123 determines the conditions of the element recommended for the input conditions based on the open rate (and action rate) estimated by the estimation processing unit 122. The optimization processing unit 123 changes the conditions of the element to have the estimation processing unit 122 estimate the open rate, and determines the condition of the element with the highest open rate (and/or action rate) among the changed conditions of the element as the condition of the recommended element.

The optimization processing unit 123 determines the conditions of the optimal elements that have the highest open rate (and/or action rate) by using, for example, a genetic algorithm. Note that the genetic algorithm can be realized by using, for example, the DEAP library of Python, which is a general-purpose programming language.
Specifically, the optimization processor 123 executes the following processes (1) to (3).

(1) As a first process, the optimization processing unit 123 first selects conditions for a predetermined number of elements, and selects the condition of the element with the highest open rate (and/or action rate) from among the conditions for the predetermined number of elements. Here, the optimization processing unit 123 randomly generates 100 element conditions with different combination patterns, for example, using random numbers, and causes the estimation processing unit 122 to generate the open rate (and action rate) corresponding to each element condition. The optimization processing unit 123 selects the condition of the element with the highest open rate (and/or action rate) from among the 100 open rates (and action rates).

(2) As a second process, the optimization processing unit 123 selects conditions for a predetermined number of elements by changing some of the elements in the conditions for the element with the highest open rate (and/or action rate) that was further selected, and selects the condition for the element with the highest open rate from the conditions for the predetermined number of elements. That is, the optimization processing unit 123 generates 100 new conditions for elements by randomly changing some of the conditions for the element selected in (1), for example, using random numbers, and causes the estimation processing unit 122 to generate an open rate (and action rate) corresponding to the condition for each element. The optimization processing unit 123 selects the condition for the element with the highest open rate (and/or action rate) from the 100 open rates (and action rates).

(3) The optimization processing unit 123 repeats the process (2) (second process) a predetermined number of times (e.g., 100 times) and determines the conditions of the element with the highest open rate (and/or action rate) as the conditions of the recommended element.

The optimization processing unit 123 may determine the conditions for the recommended elements based on either the open rate or the action rate, or may determine the conditions for the recommended elements based on both the open rate and the action rate. The optimization processing unit 123 may also determine the conditions for the recommended elements that correspond to each of the open rate and the action rate.

Moreover, the optimization processor 123 outputs the determined conditions of the recommended elements to the outside as an output result.
FIG. 5 is a diagram showing an example of input conditions and output results in this embodiment.

Fig. 5(a) shows an example of input conditions, which are selected DM attribute information, and Fig. 5(b) shows an example of an output result (recommended element conditions) output by the DM creation support device 1 in response to the input conditions in Fig. 5(a).
That is, in FIG. 5, the DM creation support device 1 outputs the output result shown in FIG. 5(b) in response to the input conditions shown in FIG. 5(a).

Next, the operation of the DM creation support system 100 (the DM creation support device 1 and the model generating device 2) according to this embodiment will be described with reference to the drawings.
First, the learning process for generating an analytical model in this embodiment will be described with reference to FIG.

FIG. 6 is a flowchart showing an example of the learning process in this embodiment.
6, the learning data acquisition unit 221 of the model generating device 2 acquires learning data (step S101). The learning data acquisition unit 221 acquires the learning data from the learning data storage unit 211, for example.

The learning data is learning data from past performance, and includes multiple data sets (direct mail performance data) that associate multiple DM attribute information, multiple element information, and opening flags and action flags. The learning data is, for example, a data string represented by multiple "0"s and "1"s, where selected DM attribute information and element information are "1" and unselected DM attribute information and element information are "0". In addition, the opening flag, which is the objective variable in the learning data, is set so that "1" indicates opening and "0" indicates unopened. In addition, the action flag, which is the objective variable in the learning data, is set so that "1" indicates that some action has been taken and "0" indicates that no action has been taken.

Next, the learning processing unit 222 of the model generation device 2 generates an analysis model that estimates the open rate and the action rate from the learning data by machine learning (step S102). For example, the learning processing unit 222 uses LightGBM to generate, as the analysis model, a model that calculates the probability that the open flag and the action flag, which are the objective variables, will be "1".

Next, the learning processing unit 222 stores the generated analytical model in the analytical model storage unit 212 (step S103). After the processing of step S103, the learning processing unit 222 ends the learning process.

The model generating device 2 transmits the analytical model stored in the learning processing unit 222 to the DM creation support device 1, and the control unit 12 of the DM creation support device 1 stores the analytical model generated by the model generating device 2 in the analytical model storage unit 111.

Next, with reference to FIG. 7, an optimization process in this embodiment for outputting conditions for direct mail elements recommended based on input conditions will be described.
FIG. 7 is a flowchart showing an example of the optimization process in this embodiment.

As shown in FIG. 7, the condition acquisition unit 121 of the DM creation support device 1 acquires input conditions (selected DM attribute information) (step S201). The condition acquisition unit 121 acquires input conditions (e.g., see FIG. 5(a)) from outside the DM creation support device 1, and stores the acquired input conditions in the input condition storage unit 112.

Next, the optimization processing unit 123 of the DM creation support device 1 randomly selects element information and generates conditions for a predetermined number of elements with different combination patterns (step S202). The optimization processing unit 123 randomly generates, for example, 100 element conditions. The optimization processing unit 123 stores the generated 100 element conditions in the element condition storage unit 113.

Next, the estimation processing unit 122 of the DM creation support device 1 generates an open rate and an action rate from the input conditions, each of the conditions for a predetermined number of elements, and the analytical model (step S203). For example, in response to a request from the optimization processing unit 123, the estimation processing unit 122 generates an open rate and an action rate corresponding to each of the 100 element conditions from the input conditions stored in the input condition storage unit 112, each of the 100 element conditions stored in the element condition storage unit 113, and the analytical model stored in the analytical model storage unit 111. The estimation processing unit 122 stores the generated open rate and action rate in the estimation result storage unit 114.

Next, the optimization processing unit 123 selects the element condition with the highest open rate and action rate from among the conditions for a predetermined number of elements (step S204). The optimization processing unit 123 selects the element condition with the highest open rate (and/or action rate) from among the open rates and action rates corresponding to each of the 100 element conditions stored in the estimation result storage unit 114.

Next, the optimization processing unit 123 determines whether or not the process has been repeated a predetermined number of times (step S205). The optimization processing unit 123 determines whether or not the process of step S204 described above has been executed, for example, 100 times (predetermined number of times). If the process has been executed 100 times (predetermined number of times) (step S205: YES), the optimization processing unit 123 advances the process to step S207. If the process has not been executed 100 times (predetermined number of times) (step S205: NO), the optimization processing unit 123 advances the process to step S206.

In step S206, the optimization processing unit 123 randomly changes some of the conditions of the selected elements to generate conditions for a predetermined number of elements with different combination patterns. For example, the optimization processing unit 123 generates 100 types of element conditions by randomly changing some of the conditions of the selected elements. The optimization processing unit 123 stores the generated 100 types of element conditions in the element condition storage unit 113. After the processing of step S206, the optimization processing unit 123 returns the processing to step S203, and in step S205, the processing of step S206 and steps S203 to S205 is repeated until it has been executed 100 times (a predetermined number of times).

Furthermore, in step S207, the optimization processing unit 123 determines the condition of the selected element as the condition of the element recommended for the input condition.
Next, the optimization processing unit 123 outputs the determined conditions of the elements (step S208). That is, the optimization processing unit 123 outputs the determined conditions of the elements as an output result (for example, see FIG. 5(b)) to the outside of the DM creation support device 1. After the process of step S208, the optimization processing unit 123 ends the optimization process.

In the flowchart shown in FIG. 7, the processes from the first step S202 to step S204 correspond to the first process described above, and the processes in steps S206, S203, and S204 correspond to the second process described above.

As described above, the DM creation support device 1 (direct mail creation support device) according to this embodiment includes a condition acquisition unit 121, an estimation processing unit 122, and an optimization processing unit 123. The condition acquisition unit 121 acquires DM attribute information (attribute information) indicating attributes related to the target of the direct mail as an input condition. The estimation processing unit 122 estimates the open rate based on an analysis model that estimates the open rate generated by machine learning from learning data in past performance in which the DM attribute information, element information indicating physical elements related to the medium of the direct mail, and open information (e.g., an open flag) indicating whether the direct mail has been opened are associated with each other, and the input condition acquired by the condition acquisition unit 121. The optimization processing unit 123 determines the condition of the element recommended for the input condition based on the open rate estimated by the estimation processing unit 122.

As a result, the DM creation support device 1 according to this embodiment estimates the open rate using an analytical model generated from DM attribute information of direct mail from past performance records and element information indicating physical elements related to the direct mail medium, and determines the element conditions recommended for the input conditions based on the open rate. Therefore, the DM creation support device 1 according to this embodiment can determine the optimal element conditions for complex and diverse direct mail. Therefore, the DM creation support device 1 according to this embodiment can propose more optimal conditions for direct mail.

In addition, in this embodiment, the input conditions are DM attribute information selected from a predetermined number of pieces of DM attribute information. The estimation processing unit 122 estimates the open rate based on the input conditions, element information selected from a predetermined number of pieces of element information as element conditions, and the analysis model. The optimization processing unit 123 changes the element conditions to have the estimation processing unit 122 estimate the open rate, and determines the element conditions with the highest open rate among the changed element conditions as the recommended element conditions.

As a result, the DM creation support device 1 according to this embodiment can efficiently determine appropriate element conditions by a simple method of changing the element conditions and determining the element conditions with the highest open rate as the recommended element conditions.

In this embodiment, the optimization processing unit 123 executes a first process and a second process, and repeats the second process a predetermined number of times to determine the conditions of the element with the highest open rate. As the first process, the optimization processing unit 123 selects a predetermined number of element conditions (e.g., 100 ways), and selects the condition of the element with the highest open rate from the conditions of the predetermined number of elements (e.g., 100 ways). As the second process, the optimization processing unit 123 further selects a predetermined number of element conditions (e.g., 100 ways) in which some elements of the conditions of the selected element with the highest open rate are changed, and selects the condition of the element with the highest open rate from the conditions of the predetermined number of elements (e.g., 100 ways). That is, the optimization processing unit 123 uses a genetic algorithm to determine the conditions of the element with the highest open rate.

As a result, the DM creation support device 1 according to this embodiment can determine appropriate element conditions as conditions for recommended elements more efficiently and accurately.

In addition, in this embodiment, the learning data includes action information regarding actions taken in response to direct mail, and the analysis model estimates the opening rate and the action rate indicating the probability of whether or not an action will be taken in response to the direct mail. The optimization processing unit 123 determines the conditions of the elements recommended for the input conditions based on the opening rate and action rate estimated by the estimation processing unit 122.

As a result, the DM creation support device 1 according to this embodiment can propose conditions for recommended elements that take into account the action rate as well as the open rate, making it possible to propose even more optimal conditions for direct mail.

The direct mail creation support method according to this embodiment includes a condition acquisition step, an estimation processing step, and an optimization processing step. In the condition acquisition step, the condition acquisition unit 121 acquires DM attribute information indicating attributes related to the target of the direct mail as an input condition. In the estimation processing step, the estimation processing unit 122 estimates the open rate based on the analysis model and the input condition acquired in the condition acquisition step. Here, the analysis model is a model that is generated by machine learning from learning data in past performance that associates DM attribute information, element information indicating physical elements related to the direct mail medium, and open information indicating whether the direct mail has been opened, and estimates the open rate. In the optimization processing step, the optimization processing unit 123 determines the condition of the element recommended for the input condition based on the open rate estimated in the estimation processing step.

As a result, the direct mail creation support method according to this embodiment has the same effect as the DM creation support device 1 described above, and can propose more optimal conditions for direct mail.

Second Embodiment
Next, a DM creation support device 1a according to a second embodiment will be described with reference to the drawings.
In this embodiment, a modified example will be described in which a DM creation support device 1a includes the functions of the model generating device 2 described above.

FIG. 8 is a block diagram showing an example of a DM creation support device 1a according to this embodiment.
As shown in FIG. 8, the DM creation support device 1a includes a storage unit 11a and a control unit 12a.
In FIG. 8, the same components as those in FIG. 1 are given the same reference numerals and the description thereof will be omitted.

The storage unit 11a includes an analysis model storage unit 111, an input condition storage unit 112, an element condition storage unit 113, an estimation result storage unit 114, and a learning data storage unit 211. That is, the storage unit 11a has the same functions as the storage unit 11 in the first embodiment described above, and also includes a learning data storage unit 211 similar to that of the model generating device 2.

The control unit 12a includes a condition acquisition unit 121, an estimation processing unit 122, an optimization processing unit 123, a learning data acquisition unit 221, and a learning processing unit 222. That is, the control unit 12a has the same functions as the control unit 12 in the first embodiment described above, and is additionally provided with a learning data acquisition unit 221 and a learning processing unit 222 similar to those of the model generation device 2.

In this embodiment, the learning processing unit 222 causes the learning data storage unit 211 to generate an analytical model from the learning data through machine learning, and stores the generated analytical model in the analytical model storage unit 111.
The functions of the learning data storage unit 211, the learning data acquisition unit 221, and the learning processing unit 222 are similar to those in the first embodiment, and therefore will not be described here.

Furthermore, the operation of the DM creation support device 1a according to this embodiment is basically the same as the learning process and optimization process according to the first embodiment shown in Figures 6 and 7, so a description thereof will be omitted here.

As described above, the DM creation support device 1a according to this embodiment (or the model generation device 2 according to the first embodiment) includes a learning data acquisition unit 221 and a learning processing unit 222. The learning data acquisition unit 221 acquires learning data from past performance that associates DM attribute information indicating attributes related to the target of the direct mail, element information indicating physical elements related to the medium of the direct mail, and opening information indicating whether the direct mail has been opened. The learning processing unit 222 generates an analysis model that estimates the opening rate by machine learning from the learning data acquired by the learning data acquisition unit 221.

As a result, the DM creation support device 1a (or model generation device 2) according to this embodiment generates an analytical model that estimates the open rate from learning data including attribute information (DM attribute information) of direct mail from past performance and element information indicating physical elements related to the direct mail medium, making it possible to appropriately estimate the open rate using the analytical model. Therefore, the DM creation support device 1a (or model generation device 2) according to this embodiment is able to determine the conditions of elements recommended for the input conditions based on the appropriate open rate, and can determine the optimal element conditions for complex and diverse direct mail. Therefore, the DM creation support device 1a (or model generation device 2) according to this embodiment can propose more optimal conditions for direct mail.

The present invention is not limited to the above-described embodiment, and can be modified without departing from the spirit of the present invention.
For example, in each of the above embodiments, an example has been described in which the optimization processing unit 123 generates conditions for 100 elements and generates conditions for 100 elements by changing some of the conditions of the element with the highest open rate (and/or action rate), but this is not limited to this and may be a specified number other than 100.

In addition, in each of the above embodiments, the optimization processing unit 123 has been described as repeating the optimization 100 times, but this is not limited to this and may be a specified number other than 100.

In addition, in each of the above embodiments, the learning data acquisition unit 221 has been described as acquiring the learning data from the learning data storage unit 211, but this is not limited thereto, and the learning data may be acquired from an external source other than the learning data storage unit 211. In addition, the learning data storage unit 211 may be provided outside the DM creation support device 1a (or the model generation device 2), and for example, a server device (such as a cloud server) connected to a network may be provided with the learning data storage unit 211.

In addition, in each of the above embodiments, an example has been described in which the analysis model is a model that estimates both the open rate and the action rate, but this is not limited to this, and the model may be a model that estimates only one of the open rate and the action rate.

In addition, in each of the above embodiments, examples of DM attribute items and element items have been described with reference to Figures 2 and 3, but the DM attribute items and element items are not limited to Figures 2 and 3 and may include other items.

Each of the components of the DM creation support system 100 and the DM creation support device 1 (1a) described above has a computer system inside. A program for implementing the functions of each of the components of the DM creation support system 100 and the DM creation support device 1 (1a) described above may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read into a computer system and executed to perform processing in each of the components of the DM creation support system 100 and the DM creation support device 1 (1a) described above. Here, "reading a program recorded on a recording medium into a computer system and executing it" includes installing the program into a computer system. The "computer system" referred to here includes hardware such as an OS and peripheral devices.
Furthermore, a "computer system" may include a plurality of computer devices connected via a network including the Internet or communication lines such as a WAN, LAN, or dedicated line. Furthermore, a "computer-readable recording medium" refers to portable media such as a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, or a storage device such as a hard disk built into a computer system. In this manner, the recording medium storing the program may be a non-transitory recording medium such as a CD-ROM.

The recording medium also includes internal or external recording media accessible from a distribution server to distribute the program. The program may be divided into multiple parts, downloaded at different times, and then combined by the components of the DM creation support system 100 and the DM creation support device 1 (1a), or each divided program may be distributed by a different distribution server. Furthermore, the term "computer-readable recording medium" includes a recording medium that holds a program for a certain period of time, such as a volatile memory (RAM) inside a computer system that becomes a server or client when a program is transmitted over a network. The program may also be a recording medium for implementing part of the above-mentioned functions. Furthermore, the program may be a so-called differential file (differential program) that can realize the above-mentioned functions in combination with a program already recorded in the computer system.

In addition, some or all of the above-mentioned functions may be realized as an integrated circuit such as an LSI (Large Scale Integration). Each of the above-mentioned functions may be individually processed, or some or all of the functions may be integrated into a processor. The integrated circuit method is not limited to LSI, and may be realized by a dedicated circuit or a general-purpose processor. Furthermore, if an integrated circuit technology that can replace LSI appears due to advances in semiconductor technology, an integrated circuit based on that technology may be used.

Reference Signs List 1, 1a DM creation support device 2 Model generating device 11, 11a, 21 Storage unit 12, 12a, 22 Control unit 100 DM creation support system 111, 212 Analysis model storage unit 112 Input condition storage unit 113 Element condition storage unit 114 Estimation result storage unit 121 Condition acquisition unit 122 Estimation processing unit 123 Optimization processing unit 211 Learning data storage unit 221 Learning data acquisition unit 222 Learning processing unit

Claims (6)

a condition acquisition unit that acquires attribute information indicating attributes related to a target of the direct mail as an input condition;
an analysis model that estimates an open rate generated by machine learning from learning data of past results in which the attribute information, element information indicating physical elements related to the medium of the direct mail, and open information indicating whether the direct mail has been opened are associated with each other, and an estimation processing unit that estimates the open rate based on the input conditions acquired by the condition acquisition unit;
an optimization processing unit that determines conditions of the element recommended for the input conditions based on the open rate estimated by the estimation processing unit,
the input condition is attribute information selected from a predetermined plurality of pieces of attribute information;
The estimation processing unit estimates the opening rate based on the input condition, element information selected from a predetermined plurality of element information as a condition of the element, and the analysis model;
The direct mail creation support device is characterized in that the optimization processing unit changes the conditions of the elements, causes the estimation processing unit to estimate the open rate, and determines the condition of the element that has the highest open rate among the changed conditions of the elements as the recommended condition of the element. The optimization processing unit:
a first process of selecting a predetermined number of conditions for the elements, and selecting, from among the predetermined number of conditions for the elements, a condition for the element having the highest opening rate;
and selecting conditions for the predetermined number of elements by changing some of the conditions for the element with the highest opening rate, and selecting the condition for the element with the highest opening rate from among the conditions for the predetermined number of elements.
2. The direct mail creation support device according to claim 1 , further comprising: determining a condition for the element with the highest open rate by repeating the second process a predetermined number of times. The direct mail creation support device according to claim 1 or 2 , further comprising a learning processing unit that generates the analysis model from the learning data by machine learning. The learning data includes action information regarding behavior regarding the direct mail;
The analytical model estimates the open rate and an action rate indicating the probability of an action being taken in response to the direct mail;
The direct mail creation support device according to any one of claims 1 to 3, characterized in that the optimization processing unit determines the conditions of the elements recommended for the input conditions based on the open rate and action rate estimated by the estimation processing unit. a condition acquisition step in which a condition acquisition unit acquires attribute information indicating attributes related to a target of the direct mail as an input condition;
an estimation processing step in which an estimation processing unit estimates the open rate based on an analytical model that estimates an open rate by machine learning from learning data of past results in which the attribute information, element information indicating physical elements related to the medium of the direct mail, and open information indicating whether the direct mail has been opened are associated with each other, and the input conditions acquired by the condition acquisition step;
an optimization processing step of determining conditions of the element recommended for the input condition based on the open rate estimated by the estimation processing step,
the input condition is attribute information selected from a predetermined plurality of pieces of attribute information;
In the estimation processing step, the estimation processing unit estimates the opening rate based on the input condition, element information selected from a predetermined plurality of element information as a condition of the element, and the analysis model;
A direct mail creation support method, characterized in that, in the optimization processing step, the optimization processing unit changes the conditions of the elements, causes the estimation processing unit to estimate the open rate, and determines the conditions of the elements having the highest open rate among the changed conditions of the elements as the conditions of the elements to be recommended. On the computer,
a condition acquisition step of acquiring attribute information indicating attributes related to a target of the direct mail as an input condition;
an analysis model for estimating an opening rate generated by machine learning from learning data of past results that associates the attribute information, element information indicating physical elements related to the medium of the direct mail, which is a mailing, including the type of delivery medium of the direct mail and the contents written therein, with opening information indicating whether the direct mail has been opened or not, and an estimation processing step for estimating the opening rate based on the input conditions acquired by the condition acquisition step;
an optimization process step of determining conditions of the element recommended for the input condition based on the open rate estimated by the estimation process step;
the input condition is attribute information selected from a predetermined plurality of pieces of attribute information;
In the estimation process, the opening rate is estimated based on the input condition, element information selected as the element condition from among a plurality of predetermined element information, and the analysis model;
A program that, in the optimization processing step, changes the conditions of the element, estimates the open rate in the estimation processing step, and determines the condition of the element that has the highest open rate among the changed conditions of the element as the recommended condition of the element.

Images