Prediction device, prediction method, and program

Abstract

Implement measures in areas where they would be most effective if implemented. [Solution] A prediction device accepts the designation of a specific region where a measure related to the provision of a service will be implemented, identifies a precedent region where the measure has been implemented in the past, and predicts whether potential member users in the specific region who have not yet used the service will use the service after the measure has been implemented. Regional similarity, which is the degree of similarity between the precedent region and the specific region, and user similarity between member users in the precedent region who had not used the service before the measure was implemented and member users in the specific region, are input into a machine learning model to predict whether or not the service will be used. The machine learning model is trained by using as training data data correlating the number and usage rate of the service in the precedent region with the regional similarity between the precedent region and past regions where measures were implemented before the precedent region, and the user similarity between member users in the past region and the precedent region.

Description

[Technical Field]

[0001] The present invention relates to a prediction device, a prediction method, and a program. [Background technology]

[0002] In recent years, delivery services using the Internet, such as online supermarkets, have become popular, and various efforts have been made to promote the use of such services. For example, Patent Document 1 discloses an information processing device that promotes the use of services by providing benefits such as electronic coupons to introducers and introduced persons when an introducer's order is confirmed. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2015-52944 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the above-mentioned information processing device did not take into account the effectiveness of measures such as granting bonuses, and there was room for improvement in implementing measures in areas where they would be most effective.

[0005] The present invention is intended to solve the above-mentioned problems, and aims to provide a prediction device, a prediction method, and a program that enable measures to be implemented in areas where implementation would be most effective. [Means for solving the problem]

[0006] A prediction device according to a first aspect of the present invention comprises: a specific area designation reception unit that receives designation of a specific area where measures related to the provision of services are to be implemented; a precedent region identification unit that identifies precedent regions where the measures have been taken in the past; a prediction unit that predicts the number of potential users of the service who are member users in the specific area and have not yet used the service after the measures are taken, the prediction unit predicts the number of users of the service by inputting into a machine learning model a regional similarity, which is a similarity between the precedent region and the specific region, and a user similarity between a member user in the precedent region who has not used the service before the measure is taken and a member user in the specific region; The machine learning model is a machine learning model that is trained by machine learning using, as training data, data that corresponds the number of users of the service in the precedent region, the region similarity between the precedent region and a past region in which the measure was taken before the precedent region, and the user similarity between member users in the past region and member users in the precedent region. It is characterized by:

[0007] In addition, in the prediction device according to the above aspect, the prediction unit calculates the user similarity based on a user database that records at least the postal code of the area where the member user resides, the user attributes of the member user, and whether or not the member user has used each of a plurality of services including the service; It is characterized by:

[0008] In addition, in the prediction device according to the above aspect, The specific area designation reception unit accepting a designation of the specific area by selecting a polygonal area corresponding to the specific area; Based on a region database in which postal codes are associated with polygonal regions of regions corresponding to the postal codes in a map database, the postal code corresponding to the selected polygonal region is determined as the postal code of the specific region; the prediction unit identifies, based on the user database, a member user corresponding to a postal code of the specific region as a member user of the specific region, and calculates the user similarity; The regional database includes: Obtaining a postal town name associated with the postal code from a postal code database; selecting, from polygonal areas corresponding to map street names in a map database, a polygonal area of ​​a map street name that matches the acquired postal street name, using a large-scale language model; Associating the postal code with the selected polygonal area; It is constructed by It is characterized by:

[0009] In addition, in the prediction device according to the above aspect, The user database further records the frequency of use of each of a plurality of services, including the service, by the member user; the prediction unit further predicts the frequency of use of the service by potential users who are member users in the specific area and have not yet used the service after the measures are implemented; It is characterized by:

[0010] In addition, in the prediction device according to the above aspect, the prediction unit outputs the number of users of the service in association with a polygonal area corresponding to the specific region; It is characterized by:

[0011] A prediction method according to a second aspect of the present invention includes: A prediction method using a prediction device, a specific area designation acceptance step for accepting designation of a specific area in which measures related to the provision of services are to be implemented; a precedent region identification step of identifying a precedent region where the measure has been taken in the past; a prediction step of predicting the number of potential users of the service who are member users in the specific area and have not yet used the service after the measures are taken, In the prediction step, a regional similarity, which is a similarity between the precedent region and the specific region, and a user similarity between a member user in the precedent region who has not used the service before the implementation of the measure and a member user in the specific region are input into a machine learning model to predict the number of users of the service; The machine learning model is a machine learning model that is trained by machine learning using, as training data, data that corresponds the number of users of the service in the precedent region, the region similarity between the precedent region and a past region in which the measure was taken before the precedent region, and the user similarity between member users in the past region and member users in the precedent region.

[0012] A program according to a third aspect of the present invention comprises: Computer, a specific area designation reception unit that accepts designations of specific areas where measures related to the provision of services are to be implemented; a precedent region identification unit that identifies precedent regions where the measures have been taken in the past; a prediction unit that predicts the number of potential users of the service who are member users in the specific area and have not yet used the service after the measures are taken; the prediction unit predicts the number of users of the service by inputting into a machine learning model a regional similarity, which is a similarity between the precedent region and the specific region, and a user similarity between a member user in the precedent region who has not used the service before the measure is taken and a member user in the specific region; The machine learning model is a machine learning model that is trained by machine learning using, as training data, data that corresponds the number of users of the service in the precedent region, the region similarity between the precedent region and a past region in which the measure was taken before the precedent region, and the user similarity between member users in the past region and member users in the precedent region. It is characterized by:

[0013] The program may be recorded on a non-transitory recording medium. The non-transitory recording medium can be distributed or sold independently of the computer. Here, a non-transitory recording medium refers to a tangible recording medium. Examples of non-transitory recording media include compact discs, flexible disks, hard disks, magneto-optical disks, digital video disks, magnetic tapes, and semiconductor memories. A transitory recording medium refers to the transmission medium (propagation signal) itself. Examples of transitory recording media include electrical signals, optical signals, and electromagnetic waves. A temporary storage area is an area for temporarily storing data and programs, such as volatile memory such as RAM (Random Access Memory). [Effects of the Invention]

[0014] According to the present invention, it is possible to provide a prediction device, a prediction method, and a program that enable measures to be implemented in areas where the effects of implementation will be high. [Brief explanation of the drawings]

[0015] [Figure 1] FIG. 2 is a diagram illustrating the relationship between a prediction device and a service providing server. [Figure 2] FIG. 1 is a block diagram illustrating an example of a prediction device. [Figure 3] FIG. 3 is an explanatory diagram illustrating an example of a user database. [Figure 4] FIG. 2 is an explanatory diagram illustrating an example of a region database. [Figure 5] FIG. 10 is an explanatory diagram illustrating an example of a regional similarity calculation model. [Figure 6] FIG. 10 is an explanatory diagram illustrating an example of a user similarity calculation model. [Figure 7] FIG. 10 is an explanatory diagram showing an example of a prediction result calculation model. [Figure 8] 10 is a flowchart illustrating an example of a prediction process. [Figure 9] FIG. 10 is an explanatory diagram showing a display example of a prediction result. [Figure 10] FIG. 10 is an explanatory diagram showing a display example in a modified example. DETAILED DESCRIPTION OF THE INVENTION

[0016] (Overall composition) A prediction device, a prediction method, and a program according to an embodiment of the present invention will be described in detail with reference to the drawings. Note that identical or corresponding parts in the drawings are designated by the same reference numerals. As shown in FIG. 1, a prediction device 100 according to an embodiment of the present invention is communicatively connected to a service providing server 200 via a computer communication network 300, such as the Internet. In this embodiment, the service providing server 200 is a server that provides various services, including an online supermarket. In this embodiment, the online supermarket service is referred to as a specific service, and members who have not yet used the specific service but who already use other services are referred to as potential users. Meanwhile, members who already use the specific service are referred to as active users. Furthermore, regions where a specific service is not currently offered but are being considered for future provision are referred to as specific regions. Meanwhile, regions where a policy was implemented and a specific service was offered in the past are referred to as precedent regions. In this embodiment, the specific region and precedent region are regions designated by postal codes. Furthermore, in this embodiment, the following explanation will be given using an example of implementing policy A, which distributes first-time-only coupons to member users in a specific region.

[0017] The prediction device 100 is a computer such as a server or a PC (Personal Computer). The prediction device 100 has a function of predicting the behavior of potential users in a specific area and predicting the number of users of a specific service when a measure is implemented. In other words, the prediction device 100 has a function of predicting whether potential users in the specific area are likely to use the specific service when a measure is implemented, thereby providing information for determining whether to provide the specific service to the specific area.

[0018] The service providing server 200 is a server that provides various services, including specific services, to members (member users). In response to a request from an information terminal (a so-called computer) such as a smartphone, tablet, or PC (Personal Computer) used by the member user, the service providing server 200 provides the service according to the request. Various information about the member, such as identification information and login information that identify the member, is registered in the service providing server 200.

[0019] (Functional configuration of the prediction device) Next, the configuration of the prediction device 100 will be described with reference to FIG.

[0020] As shown in FIG. 2, the prediction device 100 includes a storage unit 110, a control unit 120, an input / output unit 130, a communication unit 140, and a system bus (not shown) that interconnects these units.

[0021] The storage unit 110 includes a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The ROM stores a program 111 executed by the control unit 120, various data (not shown) required in advance for executing the program 111, a user database 112, a region database 113, and a prediction model 114.

[0022] The program 111 is a program for executing a prediction process, which will be described later, and is stored in the storage unit 110 in advance.

[0023] The user database 112 is a database of members and is acquired from the service providing server 200. Specifically, as shown in FIG. 3, the user database 112 stores information such as the postal codes of specific and precedent regions, member information such as the member user's gender and age, usage status of services other than specific services such as Services A to F, and changes in usage of specific services after the implementation of a policy, all of which are associated with each other. The member information includes the member user's username and login information. From the perspective of protecting personal information, only the postal code of the member user's address is registered (or can be referenced), and other personal information such as the member user's real name is anonymized. In the example shown in FIG. 3, the postal codes "362-0025" and "362-0013" are precedent regions, and the postal code "362-0023" is a specific region. The user database 112 may be acquired from the service providing server 200 when a prediction process (described later) is started, or may be acquired and updated periodically, for example, several times a day at a predetermined time.

[0024] The area database 113 is a database in which postal codes are associated with polygons of the areas identified by the postal codes. As shown in FIG. 4, the area database 113 is a database generated by a large-scale language model 113D using, for example, a postal code database 113A and a statistical geographic information system 113B. The postal code database 113A may be information provided by, for example, Japan Post Co., Ltd., and the statistical geographic information system 113B may be information provided by, for example, e-Stat. The large-scale language model 113D is a computer language model configured using an artificial neural network with numerous parameters. In the example area database 113 shown in FIG. 4, "Polygon 1" is associated with the postal code "362-0025," "Polygon 2" with the postal code "362-0013," "Polygon 3" with the postal code "362-0023," and "Polygon 4" with the postal code "362-0024." The polygon "Polygon 3" is the polygon indicated by the arrow in Figure 9, and the polygon "Polygon 1" is the polygon indicated by the arrow in Figure 10. Note that the area database 113 in this embodiment is constructed by, for each postal code, acquiring a postal town name associated with the postal code from, for example, the postal code database 113A, selecting a polygonal area (polygon) that best matches the acquired postal town name from polygonal areas (polygons) associated with map town names in a map database such as the statistical geographic information system 113B using the large-scale language model 113D, and associating each postal code with the selected polygonal area (polygon). Note that the postal code database 113A, the statistical geographic information system 113B, and the large-scale language model 113D are no longer needed once the area database 113 is generated.

[0025] The prediction model 114 is a learning model that predicts the number and ratio of potential users in a specific region who will use a specific service based on the region similarity, which is the similarity between the specific region and the precedent region, and the user similarity, which is the similarity between the situation of users in the specific region and the situation of users in the precedent region. The prediction model 114 includes a region similarity calculation model 114A, a user similarity calculation model 114B, and a prediction result calculation model 114C.

[0026] The regional similarity calculation model 114A is a machine learning model that calculates the similarity (regional similarity) between a specified specific region and a precedent region. The regional similarity calculation model 114A calculates the similarity between the specific region and the precedent region, for example, based on the feature vector of the specific region and the feature vector of the precedent region. The feature vector includes information such as land prices, private car ownership rate, presence or absence of stations and bus stops, presence or absence of major roads, private car ownership rate per capita, and presence or absence of supermarkets, and this information may be acquired from the regional database 113, data provided by local governments or the country, map information, etc. Note that the feature vector may include polygons (polygonal regions). The regional similarity calculation model 114A is a machine learning model that, when the postal code of the specific region and the postal code of the precedent region are input, acquires the feature vectors of each region and calculates the similarity between them.

[0027] The user similarity calculation model 114B is a machine learning model that calculates the similarity (user similarity) between a member user in a specified specific region and a member user in a precedent region. The user similarity calculation model 114B calculates the similarity between the member users in the specific region and the member users in the precedent region, for example, based on a feature vector of the member user attributes of the specific region and a feature vector of the member user attributes of the precedent region. The feature vector of the member user attributes includes information such as gender, age, and whether or not the user uses services other than the specific service such as services A to F, and this information can be obtained from the user database 112.

[0028] The prediction result calculation model 114C is a machine learning model that predicts the number of member users in a specific region who will use a specific service (number of users) and the proportion of member users in a specific region who will use the specific service (usage proportion) assuming that measure A is implemented, based on the regional similarity calculated by the regional similarity calculation model 114A and the user similarity calculated by the user similarity calculation model 114B. The prediction result calculation model 114C is a machine learning model that has been trained on training data that matches the number of users and usage proportion in regions where measure A was implemented in the past with the regional similarity and user similarity between the region where measure A was implemented (past region) and the precedent region.

[0029] 2, the control unit 120 is configured with a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), etc. The control unit 120 operates in accordance with a program 111 stored in the storage unit 110, and executes processing in accordance with the program 111. The control unit 120 includes a specific area designation receiving unit 121, a precedent area identification unit 122, and a prediction unit 123 as main functional units provided by the program 111 stored in the storage unit 110.

[0030] The specific area designation receiving unit 121 is a functional unit that receives designation of a specific area. Specifically, the specific area designation receiving unit 121 is a functional unit that receives, through a selection operation by a user of the prediction device 100, designation of a specific area that is an area where it is desired to consider whether or not to implement the measure A (an area that is a candidate for implementation). The specific area designation receiving unit 121 in this embodiment receives designation of a specific area that is an area where the measure A is planned to be implemented, through input of a postal code by the user of the prediction device 100 or selection of an area represented by a polygon as shown in FIG. 9 . When a postal code is input, the specific area designation receiving unit 121 identifies a polygon corresponding to the postal code based on the area database 113 stored in the storage unit 110. Furthermore, when selection of an area represented by a polygon as shown in FIG. 9 is performed, the specific area designation receiving unit 121 identifies a postal code of the area based on the area database 113 stored in the storage unit 110.

[0031] The precedent region identification unit 122 is a functional unit that identifies a precedent region that has already implemented measure A based on the user database 112 and the region database 113 stored in the storage unit 110. Specifically, the precedent region identification unit 122 identifies the precedent region based on the content of the "Change in usage of specific service after implementation of measure" item contained in the user database 112. For example, in the example shown in FIG. 3, the precedent region identification unit 122 identifies, as the precedent region, a region with a postal code where the "Change in usage of specific service after implementation of measure" item is other than "None" but is "None → Use" or "None → None." The precedent region identification unit 122 then identifies a polygon corresponding to the postal code identified as the precedent region based on the region database 113. That is, the precedent region identification unit 122 identifies the postal code of the precedent region based on the user database 112 and identifies the polygon of the precedent region based on the region database 113.

[0032] 2, the prediction unit 123 is a functional unit that predicts the number and ratio of potential users in a specific area who will use a specific service when the measure A is implemented. Specifically, the prediction unit 123 includes functional units such as a region similarity calculation unit 123A, a user similarity calculation unit 123B, and a prediction result output unit 123C.

[0033] The regional similarity calculation unit 123A is a functional unit that calculates the regional similarity between the specific region accepted by the specific region designation acceptance unit 121 and each precedent region identified by the precedent region identification unit 122. Specifically, the regional similarity calculation unit 123A calculates the regional similarity using the regional similarity calculation model 114A based on the feature vector of the specific region accepted by the specific region designation acceptance unit 121 and the feature vector of each precedent region identified by the precedent region identification unit 122. Note that each feature vector is acquired by the regional similarity calculation model 114A. Therefore, the regional similarity calculation unit 123A calculates the regional similarity by inputting the specific region accepted by the specific region designation acceptance unit 121 and each precedent region identified by the precedent region identification unit 122 into the regional similarity calculation model 114A.

[0034] The user similarity calculation unit 123B is a functional unit that calculates the user similarity between the member users in the specific region accepted by the specific region designation acceptance unit 121 and the member users in each precedent region identified by the precedent region identification unit 122. Specifically, the user similarity calculation unit 123B calculates the user similarity using the user similarity calculation model 114B based on the feature vector of the member user attributes of the specific region accepted by the specific region designation acceptance unit 121 and the feature vector of the member user attributes of each precedent region identified by the precedent region identification unit 122. The user similarity calculation unit 123B calculates the user similarity by inputting the feature vector of the user attributes of the member users in the specific region indicated by the user database 112 and the feature vector of the member user attributes of each precedent region indicated by the user database 112 into the user similarity calculation model 114B. The user similarity calculation unit 123B calculates the similarity between the member users in the precedent region who did not use the specific service before measure A was implemented and users in the specific region.

[0035] The prediction result output unit 123C is a functional unit that outputs, as prediction results, the number of member users in a specific region who will use the specific service (number of users) and the proportion of member users in a specific region who will use the specific service (usage proportion) assuming that the measure A is implemented. Specifically, the prediction result output unit 123C is a functional unit that inputs the regional similarity calculated by the regional similarity calculation model 114A and the user similarity calculated by the user similarity calculation model 114B into the prediction result calculation model 114C to predict the number of member users in a specific region who will use the specific service (number of users) and the proportion of member users in a specific region who will use the specific service (usage proportion) assuming that the measure A is implemented, and outputs the result as a prediction result. Note that the prediction result calculation model 114C in this embodiment is trained so that the higher the regional similarity and user similarity (i.e., the more similar the attributes of the region and member users), the closer the number of users and usage proportion of the specific service in a precedent region with high similarity will be.

[0036] The input / output unit 130 is a device that is configured with a keyboard, a mouse, a camera, a microphone, a liquid crystal display, an organic EL (Electro-Luminescence) display, and the like, and is used to input and output various types of data.

[0037] The communication unit 140 is a device that enables the prediction device 100 to communicate with other information terminals such as the service providing server 200 via the computer communication network 300. The configuration of the prediction device 100 has been described above.

[0038] (operation) Next, a description will be given of the operation of the prediction device 100. Fig. 8 is a flowchart showing an example of a prediction process in the prediction device 100. The prediction process in this embodiment starts when a start operation is performed by a user of the prediction device 100 who operates the prediction device 100.

[0039] When the prediction process shown in FIG. 8 starts, the control unit 120 receives a designation of a specific area using the function of the specific area designation receiving unit 121 (step S11). Specifically, in the process of step S11, the specific area designation receiving unit 121 receives the designation of a specific area based on an input by the user of the prediction device 100 of a postal code of an area where the measure A is planned to be implemented, or a selection operation of a polygon of the area where the measure A is planned to be implemented, as shown in FIG. 9. In this example, in the process of step S11, as shown in FIG. 9, a polygon of an area with a postal code of "362-0023" is selected, and the area of ​​362-0023 is accepted as a specific area. Note that if only a postal code is input, in the process of step S11, the specific area designation receiving unit 121 identifies a polygon corresponding to the postal code based on the area database 113 stored in the storage unit 110. Also, when a selection operation is performed for an area indicated by a polygon as shown in Figure 9, in the processing of step S11, the specific area designation receiving unit 121 identifies the postal code of the area based on the area database 113 stored in the memory unit 110.

[0040] After executing the process of step S11 shown in FIG. 8, the control unit 120 uses the function of the precedent area identification unit 122 to identify precedent areas that have already implemented the measure A (step S12). Specifically, in the process of step S12, the precedent area identification unit 122 identifies the precedent area based on the content of the item "Change in usage of specific service after implementation of measure" included in the user database 112 stored in the storage unit 110, and identifies polygons corresponding to the postal codes of the identified precedent areas based on the area database 113 stored in the storage unit 110. In this example, in the process of step S12, the area with the postal code "362-0025" and the area with the postal code "362-0013" shown in FIG. 3 are identified as precedent areas, and the polygons "Polygon 1" and "Polygon 2" shown in FIG. 4 are identified as polygons of the precedent area.

[0041] 8, the control unit 120 calculates the regional similarity between the specific region received in the processing of step S11 and the precedent region identified in the processing of step S12 (step S13) using the function of the regional similarity calculation unit 123A included in the prediction unit 123. Specifically, in the processing of step S13, the regional similarity calculation unit 123A calculates the regional similarity by inputting the specific region received in the processing of step S11 and each precedent region identified in the processing of step S12 into the regional similarity calculation model 114A. More specifically, in the processing of step S13, the regional similarity calculation unit 123A uses the regional similarity calculation model 114A to acquire feature vectors of the specific region and the precedent region, such as land prices, car ownership rate, presence or absence of stations and bus stops, presence or absence of major roads, car ownership rate per capita, and presence or absence of supermarkets, based on the postal codes of the specific region and the precedent region, and calculates the respective similarities based on the feature vectors of the specific region and the precedent region using the regional similarity calculation model 114A. Note that the polygons of the specific region and the precedent region may also be included in the respective feature vectors to calculate the similarities.

[0042] After executing the process of step S13, the control unit 120 uses the function of the user similarity calculation unit 123B included in the prediction unit 123 to calculate the user similarity between the member users in the specific region received in the process of step S11 and the member users in the precedent region identified in the process of step S12 (step S14). Specifically, in the process of step S14, the user similarity calculation unit 123B calculates the user similarity by inputting the feature vector of the member user attributes of the member users in the specific region received in the process of step S11 and the feature vector of the member user attributes of the member users in each precedent region identified in the process of step S12 into the user similarity calculation model 114B. The feature vector of the member user attributes includes information such as gender, age, and whether or not services other than the specific service such as services A to F are used, and this information may be obtained from the user database 112.

[0043] After executing the process of step S14, the control unit 120, using the function of the prediction result output unit 123C included in the prediction unit 123, predicts the number of member users in the specific region who will use the specific service (number of users) and the proportion of member users in the specific region who will use the specific service (usage proportion) assuming that the measure A is implemented, based on the regional similarity calculated in the process of step S13 and the user similarity calculated in the process of step S14, outputs the prediction results (step S15), and ends the prediction process. Specifically, in the process of step S15, the prediction result output unit 123C inputs the regional similarity calculated in the process of step S13 and the user similarity calculated in the process of step S14 into the prediction result calculation model 114C, thereby predicting the number of member users in the specific region who will use the specific service (number of users) and the proportion of member users in the specific region who will use the specific service (usage proportion) assuming that the measure A is implemented, and outputs the prediction results to the input / output unit 130 as shown in FIG. As mentioned above, the prediction results output in step S15 will be closer to the number of users and usage rate of a specific service in a precedent region with high similarity the higher the regional similarity and user similarity (i.e., the more similar the attributes of the region and member users are).

[0044] In the processing of step S15 in Figure 8, in this embodiment, as shown in Figure 9, the prediction result output unit 123C outputs at least the ``measure to be implemented,'' ``postal code of the specific area,'' ``number of member users,'' ``current number of users of the specific service,'' ``expected number of users of the specific service as a result of implementing the measure,'' and ``expected usage rate of the specific service as a result of implementing the measure'' to the input / output unit 130 as prediction results.

[0045] The above is the operation of the prediction device 100. As described above, the prediction device 100 in this embodiment makes it possible to check the predicted results of users of a specific service in a specific area before implementing a measure, thereby enabling measures to be implemented in areas where the implementation will be most effective. Furthermore, since the prediction results are calculated based on the regional similarity between the specific area and the preceding area and the user similarity between the specific area and the preceding area, highly accurate prediction results can be output. Furthermore, the prediction device 100 in this embodiment can display the prediction results in association with the polygons of the specific area, as shown in FIG. 9, allowing the geographical positional relationship to be easily understood.

[0046] (Variation) It should be noted that the present invention is not limited to the above-described embodiment, and various modifications and applications are possible. For example, the prediction device 100 according to the above-described embodiment does not need to have all of the technical features described above, but may have some of the configurations described in the above-described embodiment so as to solve at least one problem in the prior art. Furthermore, at least a portion of each of the following modifications may be combined.

[0047] In the above embodiment, when the prediction process shown in FIG. 8 is executed, the prediction results for implementing measure A in a specific region are displayed as shown in FIG. 9. In addition to this, for example, when a precedent region is selected, the number of member users in the precedent region and the number of users of a specific service may be displayed as shown in FIG. 10. The illustrated example shows an example in which a precedent region with a postal code of "362-0025" (see FIG. 3) is selected. In this case, unlike the prediction process shown in FIG. 8, a precedent region output process that outputs information about the precedent region may be performed. Specifically, in the precedent region output process, the postal code corresponding to the selected polygon is identified based on the region database 113 shown in FIG. 4, and the number of member users and the number of users of the specific service corresponding to the identified postal code are obtained from the user database 112 shown in FIG. 3 and displayed.

[0048] Furthermore, in the above embodiment, for ease of understanding, an example in which measure A is implemented has been described. However, it is also possible to select which of multiple measures other than measure A to implement. In this case, the user database 112 shown in FIG. 3 may store information corresponding to the item "change in usage of specific service" for each measure. Then, in the prediction process shown in FIG. 8, a prediction result corresponding to the measure to be implemented may be output. In this case, the prediction result calculation model 114C may be a machine learning model that uses, as training data, data in which the number of users and usage rate in a region corresponding to a type of measure implemented in the past correspond to the regional similarity and user similarity between the region where the measure was implemented and the preceding region. This makes it possible to predict which measure will be more effective and to implement the more effective measure. In other words, even for the same specific region, it is possible to predict each measure to be implemented, and more effective measures can be implemented.

[0049] In addition, multiple specific regions may be selected, and multiple prediction processes may be executed simultaneously for multiple specific regions. In this case, multiple prediction results may be output, and the display may be different, such as by using different colors, depending on the prediction results of the number of users and usage rate of the specific service. This makes it easy to see which specific regions are most effective in implementing measures for the specific service.

[0050] In the above embodiment, the user database 112 shown in FIG. 3 has registered information indicating the usage status of services other than the specific service, such as services A to F. However, the usage status may include information on usage frequency in addition to whether or not a service is used. Furthermore, the item for the change in usage of a specific service may also include information on the usage frequency of the specific service for users who have changed from "not used" to "used." The prediction results of the prediction process may include the number of member users in a specific region who use the specific service (number of users) and the proportion of member users in the specific region who use the specific service (usage proportion), as well as the usage frequency. This allows for calculation of user similarity taking into account the usage frequency of each service, thereby enabling more accurate prediction of the effects of implementing a policy in a specific region. Furthermore, the prediction results of the prediction process may output the effects of implementing a policy as a score out of 100, rather than the number of users or usage proportion. Alternatively, the predicted value of the usage amount of the specific service may be output as the prediction result.

[0051] In the above embodiment, an example is shown in which a region where a specific service is not currently provided but where it is desired to determine whether to provide the service in the future is defined as a specific region and the number of users of the specific service in the specific region is predicted, but this is just one example. The prediction device 100 can also be applied to, for example, a case in which a region where a specific service is currently provided is defined as a specific region and the effect of measures to further increase the number of users of the specific service is predicted.

[0052] In addition, in the above embodiment, the specific service is an online supermarket service, but this is just one example. The specific service is not limited to an online supermarket, and may be various other delivery services such as pizza delivery or medicine delivery.

[0053] The prediction device 100 according to the above embodiment can be realized using a normal computer, not a dedicated device. For example, the prediction device 100 that executes the above processes may be configured by installing a program for executing any of the above processes on a computer from a recording medium storing the program. Also, one prediction device 100 may be configured by multiple computers operating in cooperation with each other.

[0054] Furthermore, when the above-mentioned functions are realized by sharing the functions between an OS (Operating System) and an application, or by cooperation between the OS and the application, only the parts other than the OS may be stored on the medium.

[0055] It is also possible to superimpose the program on a carrier wave and distribute it via a communication network. For example, the program may be posted on a bulletin board system (BBS) on the communication network and distributed via the network. These programs may then be started and run under the control of an operating system in the same way as other application programs, thereby enabling the above-mentioned processing to be performed.

[0056] Various aspects of the present disclosure are summarized below as appendices.

[0057] (Appendix 1) a specific area designation reception unit that receives designation of a specific area where measures related to the provision of services are to be implemented; a precedent region identification unit that identifies precedent regions where the measures have been taken in the past; a prediction unit that predicts the number of potential users of the service who are member users in the specific area and have not yet used the service after the measures are taken, the prediction unit predicts the number of users of the service by inputting into a machine learning model a regional similarity, which is a similarity between the precedent region and the specific region, and a user similarity between a member user in the precedent region who has not used the service before the measure is taken and a member user in the specific region; The machine learning model is a machine learning model that is trained by machine learning using, as training data, data that corresponds the number of users of the service in the precedent region, the region similarity between the precedent region and a past region in which the measure was taken before the precedent region, and the user similarity between member users in the past region and member users in the precedent region. A prediction device characterized by:

[0058] (Appendix 2) the prediction unit calculates the user similarity based on a user database that records at least the postal code of the area where the member user resides, the user attributes of the member user, and whether or not the member user has used each of a plurality of services including the service; 2. The prediction device according to claim 1,

[0059] (Appendix 3) The specific area designation reception unit accepting a designation of the specific area by selecting a polygonal area corresponding to the specific area; Based on a region database in which postal codes are associated with polygonal regions of regions corresponding to the postal codes in a map database, the postal code corresponding to the selected polygonal region is determined as the postal code of the specific region; the prediction unit identifies, based on the user database, a member user corresponding to a postal code of the specific region as a member user of the specific region, and calculates the user similarity; The regional database includes: Obtaining a postal town name associated with the postal code from a postal code database; selecting, from polygonal areas corresponding to map street names in a map database, a polygonal area of ​​a map street name that matches the acquired postal street name, using a large-scale language model; Associating the postal code with the selected polygonal area; It is constructed by 3. The prediction device according to claim 2,

[0060] (Appendix 4) The user database further records the frequency of use of each of a plurality of services, including the service, by the member user; the prediction unit further predicts the frequency of use of the service by potential users who are member users in the specific area and have not yet used the service after the measures are taken; 4. The prediction device according to claim 2 or 3.

[0061] (Appendix 5) the prediction unit outputs the number of users of the service in association with a polygonal area corresponding to the specific region; 5. The prediction device according to claim 3 or 4.

[0062] (Appendix 6) A prediction method using a prediction device, a specific area designation acceptance step of accepting designation of a specific area as an area in which measures related to the provision of services are to be implemented; a precedent region identification step of identifying a precedent region where the measure has been taken in the past; a prediction step of predicting the number of potential users of the service who are member users in the specific area and have not yet used the service after the measures are taken, In the prediction step, a regional similarity, which is a similarity between the precedent region and the specific region, and a user similarity between a member user in the precedent region who has not used the service before the implementation of the measure and a member user in the specific region are input into a machine learning model to predict the number of users of the service; The machine learning model is a machine learning model that is trained by machine learning using, as training data, data that corresponds the number of users of the service in the precedent region, the region similarity between the precedent region and a past region in which the measure was taken before the precedent region, and the user similarity between member users in the past region and member users in the precedent region. A prediction method characterized by:

[0063] (Appendix 7) Computer, a specific area designation reception unit that accepts designations of specific areas where measures related to the provision of services are to be implemented; a precedent region identification unit that identifies precedent regions where the measures have been taken in the past; a prediction unit that predicts the number of potential users of the service who are member users in the specific area and have not yet used the service after the measures are taken; the prediction unit predicts the number of users of the service by inputting into a machine learning model a regional similarity, which is a similarity between the precedent region and the specific region, and a user similarity between a member user in the precedent region who has not used the service before the measure is taken and a member user in the specific region; The machine learning model is a machine learning model that is trained by machine learning using, as training data, data that corresponds the number of users of the service in the precedent region, the region similarity between the precedent region and a past region in which the measure was taken before the precedent region, and the user similarity between member users in the past region and member users in the precedent region. A program characterized by: [Industrial Applicability]

[0064] According to the present invention, it is possible to provide a prediction device, a prediction method, and a program that enable measures to be implemented in areas where the effects of implementation will be high. [Explanation of symbols]

[0065] 100 Prediction Device 110 Storage section 111 Program 112 User Database 113 Regional Database 114 Predictive Models 114A Regional Similarity Calculation Model 114B User similarity calculation model 114C prediction result calculation model 120 control section 121 Specific Area Designated Reception Department 122 Precedent Area Identification Department 123 Prediction Department 123A Regional Similarity Calculation Unit 123B User similarity calculation unit 123C Prediction result output section 130 Input / output section 140 Communications Department 200 Service provider server 300 Computer Communication Network

Claims

1. a specific area designation reception unit that receives designation of a specific area where measures related to the provision of services are to be implemented; a precedent region identification unit that identifies precedent regions where the measures have been taken in the past; a prediction unit that predicts the number of potential users of the service who are member users in the specific area and have not yet used the service after the measures are taken, the prediction unit predicts the number of users of the service by inputting into a machine learning model a regional similarity, which is a similarity between the precedent region and the specific region, and a user similarity between a member user in the precedent region who has not used the service before the measure is taken and a member user in the specific region; The machine learning model is a machine learning model that is machine-learned using, as training data, data that corresponds the number of users of the service in the precedent region, the region similarity between the precedent region and a past region in which the measure was taken before the precedent region, and the user similarity between member users in the past region and member users in the precedent region. A prediction device characterized by:

2. the prediction unit calculates the user similarity based on a user database that records at least the postal code of the area where the member user resides, the user attributes of the member user, and whether or not the member user has used each of a plurality of services including the service; The prediction device according to claim 1 .

3. The specific area designation reception unit accepting a designation of the specific area by selecting a polygonal area corresponding to the specific area; Based on a region database in which postal codes are associated with polygonal regions of regions corresponding to the postal codes in a map database, the postal code corresponding to the selected polygonal region is determined as the postal code of the specific region; the prediction unit identifies, based on the user database, a member user corresponding to a postal code of the specific region as a member user of the specific region, and calculates the user similarity; The regional database includes: Obtaining a postal town name associated with the postal code from a postal code database; selecting, from polygonal areas corresponding to map street names in a map database, a polygonal area of ​​a map street name that matches the acquired postal street name, using a large-scale language model; Associating the postal code with the selected polygonal area; It is constructed by The prediction device according to claim 2 .

4. The user database further records the frequency of use of each of a plurality of services, including the service, by the member user; the prediction unit further predicts the frequency of use of the service by potential users who are member users in the specific area and have not yet used the service after the measures are implemented; 4. The prediction device according to claim 2 or 3.

5. the prediction unit outputs the number of users of the service in association with a polygonal area corresponding to the specific region; The prediction device according to claim 3 .

6. A prediction method using a prediction device, a specific area designation acceptance step for accepting designation of a specific area in which measures related to the provision of services are to be implemented; a precedent region identification step of identifying a precedent region where the measure has been taken in the past; a prediction step of predicting the number of potential users of the service who are member users in the specific area and have not yet used the service after the measures are taken, In the prediction step, a regional similarity, which is a similarity between the precedent region and the specific region, and a user similarity between a member user in the precedent region who has not used the service before the implementation of the measure and a member user in the specific region are input into a machine learning model to predict the number of users of the service; The machine learning model is a machine learning model that is machine-learned using, as training data, data that corresponds the number of users of the service in the precedent region, the region similarity between the precedent region and a past region in which the measure was taken before the precedent region, and the user similarity between member users in the past region and member users in the precedent region. A prediction method characterized by:

7. Computer, a specific area designation reception unit that accepts designations of specific areas where measures related to the provision of services are to be implemented; a precedent region identification unit that identifies precedent regions where the measures have been taken in the past; a prediction unit that predicts the number of potential users of the service who are member users in the specific area and have not yet used the service after the measures are taken; the prediction unit predicts the number of users of the service by inputting into a machine learning model a regional similarity, which is a similarity between the precedent region and the specific region, and a user similarity between a member user in the precedent region who has not used the service before the measure is taken and a member user in the specific region; The machine learning model is a machine learning model that is machine-learned using, as training data, data that corresponds the number of users of the service in the precedent region, the region similarity between the precedent region and a past region in which the measure was taken before the precedent region, and the user similarity between member users in the past region and member users in the precedent region. A program characterized by:

Images