Cancellation probability prediction apparatus

Abstract

This cancellation probability prediction apparatus is provided with: a first acquisition unit that acquires, for a target user, first usage information pertaining to the usage status of a first application; a second acquisition unit that acquires, for the target user, second usage information pertaining to the usage status of a second application; and a prediction unit that predicts the probability that the target user will cancel a first service in the future by inputting the first usage information and the second usage information acquired during a past period to a first learning model that has learned the relationship between the learning usage status of the first application and the learning usage status of the second application, and the learning probability at which the first service will be cancelled in the future. The second usage information includes at least one of the startup count and the startup time of the second application.

Description

Churn probability prediction device

[0001] The present invention relates to a churn probability prediction device.

[0002] A service provider may predict the probability of customers cancelling their service during a predetermined period from the present time in order to estimate how many customers will use their service during that period.

[0003] For example, Patent Document 1 discloses a sales activity support device that predicts the churn probability of a customer and then generates a sales activity plan based on the prediction result. The sales activity support device predicts the churn probability, which is the probability that a customer who currently has a contract will cancel the contract in the near future, based on contract customer information that indicates the performance of contracts from the past to the present and analysis target conditions that indicate which information in the contract customer information will be analyzed.

[0004] Japanese Patent Application Laid-Open No. 2021-064406

[0005] However, conventional methods for predicting customer churn rates do not take into account the usage of competing services.

[0006] Therefore, an object of the present invention is to provide a churn probability prediction device that predicts the churn probability of customers for a company's own service, taking into account the usage status of competing services.

[0007] The cancellation probability prediction device of the present invention comprises a first acquisition unit that acquires, for a target user, first usage information regarding the usage status of a first application used to provide a first service; a second acquisition unit that acquires, for the target user, second usage information regarding the usage status of a second application used to provide a second service of the same type as the first service; and a prediction unit that predicts the probability that the target user will cancel the first service in the future by inputting the first usage information and the second usage information acquired over a past period into a first learning model that has learned the relationship between the learning usage status of the first application and the learning usage status of the second application, and the learning probability that the first service will be canceled in the future, and the second usage information includes at least one of the number of times the second application is launched and the launch time.

[0008] According to the present invention, it is possible to predict the probability of customers cancelling their own company's services, taking into account the usage status of competing services.

[0009] A block diagram showing the overall configuration of a churn probability prediction system 1. A block diagram showing an example configuration of a churn probability prediction device 10. A block diagram showing an example configuration of a user database UDB1. A block diagram showing an example configuration of a message database MDB. A block diagram showing an example configuration of a probability table TTB. A block diagram showing an example configuration of a terminal device 20_k. A flowchart showing the operation of the churn probability prediction device 10. A block diagram showing an example configuration of a churn probability prediction device 10A. A block diagram showing an example configuration of a user database UDB2.

[0010] 1: First Embodiment Hereinafter, a churn probability prediction system 1 according to a first embodiment will be described with reference to the drawings.

[0011] 1-1: Configuration of First Embodiment 1-1-1: Overall Configuration FIG. 1 is a block diagram showing the overall configuration of a churn probability prediction system 1 according to this embodiment. As shown in FIG. 1, the churn probability prediction system 1 includes a churn probability prediction device 10 and n terminal devices 20_1 to 20_n. n is an integer equal to or greater than 1. Terminal devices 20_1 to 20_n include terminal device 20_k. k is an integer equal to or greater than 1 and equal to or less than n. For simplicity of explanation, terminal device 20_k may be described below as a representative example of terminal devices 20_1 to 20_n. Note that in FIG. 1, the number n of terminal devices 20_1 to 20_n is merely an example. The churn probability prediction system 1 may include any number of terminal devices 20.

[0012] Each of the terminal devices 20_1 to 20_n is a device used by a user U to use an application installed in the terminal device 20_1 to 20_n. Here, it is assumed that a user Uk uses the terminal device 20_k. The user Uk is an example of a "target user." Furthermore, the terminal devices 20_1 to 20_n are, for example, smartphones or tablets.

[0013] As will be described later, a first application used to provide a first service and a second application used to provide a second service are installed in each of the terminal devices 20_1 to 20_n. The first service and the second service are the same type of service. For example, a payment service provided by one provider and a payment service provided by another provider are the same type of service. Furthermore, a payment service provided by one provider and a nursing care service provided by another provider are different types of services. Services of the same type are in a competitive relationship. Therefore, a first provider that provides a first service and a second provider that provides a second service are in a competitive relationship.

[0014] The churn probability prediction device 10 predicts the probability that a user Uk will cancel a first service in the future. More specifically, the churn probability prediction device 10 predicts the probability that a user Uk will cancel a first service in the future, based on the usage status of each of a first application and a second application installed on a terminal device 20_k by the user Uk.

[0015] As an example, a first provider that provides a first service uses the churn probability prediction device 10. In other words, to the user of the churn probability prediction device 10, the first service is the provider's own service. Also, to the user, the second service is a competing service.

[0016] 1-1-2: Configuration of the churn probability prediction device 10 Fig. 2 is a block diagram showing an example configuration of the churn probability prediction device 10. As shown in Fig. 2, the churn probability prediction device 10 includes a processing device 11, a storage device 12, a communication device 13, a display device 14, and an input device 15. The elements of the churn probability prediction device 10 are connected to each other by one or more buses for communicating information. Note that the term "device" in this specification may be replaced with other terms such as circuit, device, or unit.

[0017] The processing device 11 is a processor that controls the entire churn probability prediction device 10. The processing device 11 is configured using, for example, one or more chips. The processing device 11 is also configured using, for example, a central processing unit (CPU) that includes an interface with peripheral devices, an arithmetic unit, a register, etc. Some or all of the functions of the processing device 11 may be realized by hardware such as a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), and an FPGA (Field Programmable Gate Array). The processing device 11 executes various processes in parallel or sequentially.

[0018] The storage device 12 is a recording medium that can be read and written by the processing device 11. The storage device 12 includes, for example, a nonvolatile memory and a volatile memory. The nonvolatile memory is, for example, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), and an electrically erasable programmable read-only memory (EEPROM). The volatile memory is, for example, a random access memory (RAM). The storage device 12 also stores a control program PR1 executed by the processing device 11. The storage device 12 also functions as a work area for the processing device 11.

[0019] The storage device 12 also stores a user database UDB1, a message database MDB, a probability table TTB, and a first learning model LM1.

[0020] The user database UDB1 stores information about each of users U1 to Un. FIG. 3 shows an example of the configuration of the user database UDB1. In the example shown in FIG. 3, if the current time is month M of year Y, the user database UDB1 has a dataset DS(M) that is a collection of data for month M of year Y, a dataset DS(M-1) that is a collection of data for month (M-1) of year Y, a dataset DS(M-2) that is a collection of data for month (M-2) of year Y, and a dataset DS(M-3) that is a collection of data for month (M-3) of year Y. Note that in the example shown in FIG. 3, the user database UDB1 has datasets DS(M) to DS(M-3) for four months, from month M of year Y to month (M-3) of year Y, three months prior, but this is merely an example. The user database UDB1 may have datasets DS(ML) to DS(M), which are datasets going back an arbitrary number of months, L months, from month M of year Y.

[0021] The data set DS(M) has as its items user ID, age, sex, item x11 to item x16, and item x21 to item x22.

[0022] "User ID" is an identifier for identifying user U who uses the first application. As an example, the user ID of user U1 is "001". The user ID of user U2 is "002". The user ID of user Uk is "00k". The user ID of user Un is "00n". "Age" is the age of each of users U1 to Un as of month M of year Y. In the dataset DS(M), as an example, it is shown that user U1 corresponding to user ID="001" is 38 years old. "Gender" is the gender of each of users U1 to Un. In the example shown in FIG. 3, "P" indicates male. "Q" indicates female. In the dataset DS(M), as an example, it is shown that user U1 corresponding to user ID="001" is male.

[0023] Items x11 to x16 are items related to the first service or the first application. Item x11 indicates the contract period from when each of users U1 to Un starts using the first service. As an example, the data set DS(M) indicates that the contract period for user U1, corresponding to user ID="001", is 183 months.

[0024] Item x12 indicates the number of times that each of users U1 to Un launched the first application during month M of year Y. As an example, data set DS(M) indicates that user U1, corresponding to user ID="001", launched the first application 48 times during month M of year Y.

[0025] Item x13 indicates the length of time that each of users U1 to Un ran the first application during month M of year Y. In the data set DS(M), as an example, the length of time that user U1, corresponding to user ID="001", ran the first application during month M of year Y is shown to be 265 minutes.

[0026] Item x14 indicates the date and time when each of users U1 to Un last used the first application at the current time. As an example, in dataset DS(M), the date and time when user U1, corresponding to user ID="001", last used the first application is shown to be December 5, 2022. Alternatively, item x14 may be the date and time when each of users U1 to Un last used the first application at the end of month M of year Y.

[0027] Item x15 indicates the number of times that each of users U1 to Un visited the site of the provider that provides the first service during month M of year Y. In dataset DS(M), as an example, it is indicated that user U1, corresponding to user ID="001", visited the site of the provider that provides the first service once during month M of year Y.

[0028] Item x16 indicates the total amount of usage that each of users U1 to Un has consumed as a result of using the first service by using the first application during month M of year Y. As an example, data set DS(M) indicates that the total amount of usage that user U1, corresponding to user ID="001", has consumed during month M of year Y is 2,000 yen.

[0029] Items x21 to x22 are items related to the second service or the second application. Item x21 indicates the number of times that each of users U1 to Un launched the second application during month M of year Y. As an example, the dataset DS(M) indicates that user U1, corresponding to user ID="001", launched the second application 16 times during month M of year Y.

[0030] Item x22 indicates the length of time that each of users U1 to Un ran the second application during month M of year Y. In the data set DS(M), as an example, the length of time that user U1, corresponding to user ID="001", ran the first application during month M of year Y is shown to be 86 minutes.

[0031] In the above example, the dataset DS(M) has six items, item x11 to item x16, related to the first service or first application. However, the dataset DS(M) may have any number of items related to the first service or first application. Furthermore, the items related to the first service or first application included in the dataset DS(M) are not limited to the contract period since the start of use of the first service, the number of times the first application was launched, the time the first application was launched, the date and time the first application was last used, the number of times the site of the provider providing the first service was visited, and the total amount spent. The dataset DS(M) may also have other types of items related to the first service or first application.

[0032] Furthermore, in the above example, the dataset DS(M) has two items, item x21 and item x22, as items related to the second service or second application. However, the dataset DS(M) may have any number of items related to the second service or second application. Furthermore, the items related to the second service or second application included in the dataset DS(M) are not limited to the number of times the second application has been launched and the time the second application has been launched. The dataset DS(M) may also have other types of items related to the second service or second application.

[0033] As described above, the dataset DS(M) includes first usage information related to the usage status of the first application and second usage information related to the usage status of the second application. Here, the first usage information includes attribute information related to the attributes of user U. In the example shown in FIG. 3 , the attribute information is information related to the age and gender of user U. Furthermore, the first usage information includes at least one of the date and time when user U last used the first service as of month M of year Y, the number of times user U used the first service during month M of year Y, and the amount of usage that user U spent to use the first service in month M of year Y.

[0034] Moreover, the data sets DS(M-1) to DS(M-3) have the same configuration as the data set DS(M).

[0035] 2, the message database MDB stores messages sent from the churn probability prediction device 10 to the terminal devices 20_1 to 20_n. These messages encourage the users U1 to Un to use the first service.

[0036] 4 shows an example of the structure of the message database MDB. The message database MDB has items such as message IDs and messages.

[0037] "Message ID" is an identifier for identifying a message. "Message" is the message body. The message body may consist of text only. Alternatively, the message body may consist of text and an image.

[0038] In the message database MDB illustrated in FIG. 4, a message with a message ID of "A" notifies that if a product or service is purchased at Store S using Application A, a 30% discount will be applied to the purchase price. A message with a message ID of "B" notifies that if a product or service is purchased at Store S using Application A, a 50% discount will be applied to the purchase price. "Application A" here is an example of the first application. Preferably, "Store S" is a store where both the first and second services are available.

[0039] 4, the message database MDB stores two messages with message IDs "A" and "B." However, the number of messages stored in the message database MDB may be any number.

[0040] 2, the probability table TTB defines the correspondence between the probability predicted by the churn probability prediction device 10 and the message ID of the message notified to the terminal device 20_1 to the terminal device 20_n from the churn probability prediction device 10. Fig. 5 shows an example of the configuration of the probability table TTB. The probability table TTB has the fields of probability and message ID.

[0041] 5 , it is defined that if the probability P that user U will cancel the first service in the future is less than t0, the churn probability prediction device 10 will not notify the terminal device 20 used by user U of a message. Also, it is defined in the probability table TTB that if the probability P is equal to or greater than t0 and less than t1, the churn probability prediction device 10 will notify the terminal device 20 used by user U of a message with a message ID of "A". Also, it is defined in the probability table TTB that if the probability P is equal to or greater than t1 and less than t2, the churn probability prediction device 10 will notify the terminal device 20 used by user U of a message with a message ID of "B".

[0042] The first learning model LM1 is a learning model used by the prediction unit 113 (described later) when predicting the probability that the user U will cancel the first service in the future. The first learning model LM1 is a learning model that has learned the relationship between the learning usage status of the first application and the learning usage status of the second application, and the learning probability that the first service will be canceled in the future. The learning usage status of the first application and the learning usage status of the second application are, for example, at least a portion of the information contained in the data sets DS(M-L) to DS(M-1) from month Y (M-L) to month Y (M-1) in the user database UDB1. Here, the first application is Aown, the second application is Acomp, and the feature amount related to the learning usage status of application A for month t of user Uk is the feature amount x ｋ，ｔ (A), the feature quantity X relating to the learning usage status of the first application and the learning usage status of the second application included in the training data of the first learning model LM1 is ｋ，Ｍ (Aown, Acomp) is expressed by the following formula:

[0043] Feature x ｋ，ｔ(Aown) is, for example, the numerical values ​​of items x11 to x16 corresponding to user Uk in datasets DS(M-L) to DS(M-1), and at least a portion of the attribute information of user Uk. These features are clearly causally related to churn from the first service. Furthermore, these features are considered to have no clear correlation with the usage status of the second service. Generally, adding features causally related to the objective variable as input features to the learning model is expected to improve prediction accuracy, as long as there is no correlation between the features. Therefore, adding the numerical values ​​of items x11 to x16 and at least a portion of the attribute information of users U1 to Un as features to the training data of the first learning model LM1 improves the accuracy of churn prediction.

[0044] In addition, the feature quantity x ｋ，ｔ (Acomp) is, for example, at least a part of the numerical values ​​of the items x21 to x22 corresponding to the user Uk in the data sets DS(ML) to DS(M-1). ｋ，ｔ In addition to (Aown), the feature value x of the competing service ｋ，ｔ By adding at least some of the numerical values ​​of items x21 to x22, which are (Acomp), the accuracy of the dropout prediction is further improved.

[0045] Furthermore, the learning probability of the first service being canceled in the future is the probability indicating how many users U, among multiple users U linked to the learning usage status of the same first application and the learning usage status of the same second application, will cancel in the future. Here, "future" refers to, for example, the most recent month. Furthermore, "multiple users U linked to the learning usage status of the same first application and the learning usage status of the same second application" here refers to, for example, multiple users U belonging to the same group when the items indicating the learning usage status of the first application and the learning usage status of the second application are grouped according to the magnitude of the numerical values.

[0046] The training data used to generate the first learning model LM1 includes multiple pairs of learning usage status of the first application and learning usage status of the second application, and learning probability that the first service will be canceled in the future.

[0047] The first learning model LM1 is generated outside the churn probability prediction device 10. Preferably, the first learning model LM1 is generated in a server (not shown). In this case, the churn probability prediction device 10 acquires the first learning model LM1 from the server (not shown) via the communication network NET.

[0048] The communication device 13 is hardware serving as a transmitting / receiving device for communicating with other devices. The communication device 13 is also referred to as, for example, a network device, a network controller, a network card, or a communication module. The communication device 13 may include a connector for wired connection and an interface circuit corresponding to the connector. The communication device 13 may also include a wireless communication interface. Examples of the connector and interface circuit for wired connection include products that comply with wired LAN, IEEE 1394, and USB. Examples of the wireless communication interface include products that comply with wireless LAN, Bluetooth (registered trademark), etc.

[0049] The display device 14 is a device that displays images and text information. The display device 14 displays various images under the control of the processing device 11. For example, various display panels such as a liquid crystal display panel and an organic EL display panel are suitably used as the display device 14.

[0050] The input device 15 is a device that accepts operations from an administrator of the churn probability prediction device 10. For example, the input device 15 includes a keyboard, a touchpad, a touch panel, or a pointing device such as a mouse. Here, if the input device 15 includes a touch panel, it may also serve as the display device 14.

[0051] The processing device 11 functions as a first acquisition unit 111, a second acquisition unit 112, a prediction unit 113, a notification unit 114, and a notification control unit 115, for example, by reading and executing a control program PR1 from the storage device 12.

[0052] The first acquisition unit 111 acquires first usage information regarding the usage status of a first application used to provide a first service for users U1 to Un, who are target users, from terminal devices 20_1 to 20_n via the communication device 13. The first acquisition unit 111 also stores the acquired first usage information in a user database UDB1. The first usage information regarding the usage status of users U1 to Un acquired by the first acquisition unit 111 includes first usage information regarding the usage status of user Uk.

[0053] The second acquisition unit 112 acquires second usage information regarding the usage status of the second application used to provide the second service for users U1 to Un, who are target users, from terminal devices 20_1 to 20_n via the communication device 13. The second acquisition unit 112 also stores the acquired second usage information in the user database UDB1. The second usage information regarding the usage status of users U1 to Un acquired by the second acquisition unit 112 includes second usage information regarding the usage status of user Uk. The second usage information also includes at least one of the number of times the second application has been launched and the launch time.

[0054] The prediction unit 113 predicts the probability Pk that user Uk will cancel the first service in the future by inputting the first usage information and second usage information of user Uk acquired over a past period into the first learning model LM1. The training data used to generate the first learning model LM1 is, for example, a set of learning usage statuses of the first application and the second application of users U1 to Un from month (M-L) of year Y to month (M-1) of year Y, and a learning probability that the first service will be canceled within one month from the date corresponding to the learning usage status. Furthermore, the first usage information and second usage information of user Uk input into the first learning model LM1 by the prediction unit 113 are, for example, the first usage information and second usage information for month M of year Y. Furthermore, the probability Pk predicted by the prediction unit 113 is, for example, the probability that user Uk will cancel the first service between month M of year Y and month (M+1) of year Y. Furthermore, it is preferable that the items of the first usage information and second usage information of user Uk that the prediction unit 113 inputs into the first learning model LM1 are the same as the usage status of the first application and the items of usage information of the second application included in the above-mentioned training data.

[0055] Furthermore, the prediction unit 113 may predict the probability P of canceling the first service not only for user Uk, but also for one or more users U other than user Uk among users U1 to Un.

[0056] The notification unit 114 notifies the terminal device 20_k used by the user Uk of a message promoting use of the first service. The message is a message stored in the message database MDB. As described above, the content of the message preferably indicates a benefit that promotes use of the first service at a store where the first service and the second service are available.

[0057] The notification control unit 115 references the probability table TTB to control at least one of whether or not to notify the notification unit 114 and the content of the message, depending on the probability P predicted by the prediction unit 113. Referring to the probability table TTB illustrated in FIG. 5 , when the probability P is less than t0, the notification control unit 115 controls the notification unit 114 not to notify any message. When the probability P is equal to or greater than t0 and less than t1, the notification control unit 115 controls the notification unit 114 to notify a message corresponding to message ID = "A". When the probability P is equal to or greater than t1 and less than t2, the notification control unit 115 controls the notification unit 114 to notify a message corresponding to message ID = "B". Referring to FIG. 4 , the percentage of deduction from the purchase price differs between the message corresponding to message ID = "A" and the message corresponding to message ID = "B". That is, the notification control unit 115 controls the benefit depending on the probability P predicted by the prediction unit 113. In this example, the higher the probability P, the more advantageous the benefit is given to the target user.

[0058] 6 is a block diagram showing an example of the configuration of the terminal device 20_k. As shown in Fig. 6, the terminal device 20_k includes a processing device 21, a storage device 22, a communication device 23, a display device 24, and an input device 25. The elements of the terminal device 20_k are connected to each other by one or more buses for communicating information.

[0059] The processing device 21 is a processor that controls the entire terminal device 20_k. The processing device 21 is configured, for example, using one or more chips. The processing device 21 is also configured, for example, using a central processing unit (CPU) that includes an interface with peripheral devices, an arithmetic unit, a register, etc. Note that some or all of the functions of the processing device 21 may be realized by hardware such as a DSP, an ASIC, a PLD, and an FPGA. The processing device 21 executes various processes in parallel or sequentially.

[0060] The storage device 22 is a recording medium that can be read and written by the processing device 21. The storage device 22 includes, for example, a nonvolatile memory and a volatile memory. The nonvolatile memory is, for example, a ROM, an EPROM, and an EEPROM. The volatile memory is, for example, a RAM. The storage device 22 also stores a control program PR2 executed by the processing device 21. The storage device 22 also functions as a work area for the processing device 21.

[0061] The storage device 22 also stores a first application program AP1 and a second application program AP2. The first application program AP1 is a program that runs a first application used by the user Uk. The second application program AP2 is a program that runs a second application used by the user Uk.

[0062] The communication device 23 is hardware serving as a transmitting / receiving device for communicating with other devices. The communication device 23 is also referred to as, for example, a network device, a network controller, a network card, or a communication module. The communication device 23 may include a connector for wired connection and an interface circuit corresponding to the connector. The communication device 23 may also include a wireless communication interface. Examples of the connector and interface circuit for wired connection include products that comply with wired LAN, IEEE 1394, and USB. Examples of the wireless communication interface include products that comply with wireless LAN, Bluetooth (registered trademark), etc.

[0063] The display device 24 is a device that displays images and text information. The display device 24 displays various images under the control of the processing device 21. For example, various display panels such as a liquid crystal display panel and an organic EL display panel are suitably used as the display device 24.

[0064] The input device 25 is a device that accepts operations from the user Uk. For example, the input device 25 includes a keyboard, a touchpad, a touch panel, or a pointing device such as a mouse. Here, if the input device 25 includes a touch panel, it may also serve as the display device 24.

[0065] The processing device 21 functions as an acquisition unit 211, a display control unit 212, an execution unit 213, and a communication control unit 214, for example, by reading and executing a control program PR2 from the storage device 22.

[0066] The acquisition unit 211 acquires a message notified from the churn probability prediction device 10 via the communication device 23. As described above, the message is a message encouraging the use of the first service.

[0067] The display control unit 212 causes the display device 24 to display the message acquired by the acquisition unit 211 .

[0068] The execution unit 213 executes the first application program AP1 and the second application program AP2 stored in the storage device 22 based on an input to the input device 25 by the user Uk.

[0069] The communication control unit 214 causes the communication device 23 to transmit first usage information regarding the usage status of the user Uk for the first application to the churn probability prediction device 10. The communication control unit 214 also causes the communication device 23 to transmit second usage information regarding the usage status of the user Uk for the second application to the churn probability prediction device 10. When the terminal device 20_k is a smartphone or a tablet, the second usage information that the communication control unit 214 causes to be transmitted is acquired from the event log of the smartphone or tablet.

[0070] 1-2: Operation of the First Embodiment FIG. 7 is a flowchart showing the operation of the churn probability prediction device 10. As shown in FIG.

[0071] In step S1, the processing device 11 functions as a first acquisition unit 111. The processing device 11 acquires first usage information regarding the usage status of a first application used to provide a first service for users U1 to Un who are target users.

[0072] In step S2, the processing device 11 functions as the second acquisition unit 112. The processing device 11 acquires second usage information regarding the usage status of a second application used to provide the second service for users U1 to Un who are target users.

[0073] In step S3, the processing device 11 functions as the prediction unit 113. The processing device 11 inputs the first usage information and the second usage information acquired in the past period into the first learning model LM1, thereby predicting the probabilities P1 to Pn that users U1 to Un will cancel the first service in the future.

[0074] 1-3: Effects of the First Embodiment The churn probability prediction device 10 according to this embodiment includes a first acquisition unit 111, a second acquisition unit 112, and a prediction unit 113. The first acquisition unit 111 acquires first usage information regarding the usage status of a first application used to provide a first service for users U1 to Un, who are target users. The second acquisition unit 112 acquires second usage information regarding the usage status of a second application used to provide a second service of the same type as the first service for users U1 to Un, who are target users. The prediction unit 113 predicts the probability Pk that user Uk will cancel the first service in the future by inputting the first usage information and the second usage information acquired over a past period into a first learning model LM1 that has learned the relationship between the learning usage status of the first application, the learning usage status of the second application, and the learning probability that the first service will be canceled in the future. The second usage information includes at least one of the number of times the second application is launched and the launch time.

[0075] With the above configuration, the churn probability prediction device 10 can predict the churn probability of customers for the company's own service, taking into account the usage status of competing services. Furthermore, by adding the feature quantity xk,t(Acomp) of the competing service in addition to the feature quantity xk,t(Aown) of the company's own service to the training data of the first learning model LM1, the accuracy of churn prediction is improved.

[0076] Furthermore, in the churn probability prediction device 10, the first usage information includes attribute information relating to the attributes of users U1 to Un.

[0077] The attribute information includes, for example, age and gender. Generally, age and gender correlate with the degree to which a user is willing to switch from the service they are currently using to another service. Therefore, by having the above configuration, the churn probability prediction device 10 adds at least a portion of the attribute information of users U1 to Un as features to the training data of the first learning model LM1, thereby improving the accuracy of churn prediction.

[0078] In addition, in the cancellation probability prediction device 10, the first usage information includes at least one of the date and time when user U1 to user Un last used the first service, the number of times user U1 to user Un used the first service within the past period, and the amount spent by user U1 to user Un to use the first service within the past period.

[0079] In general, there is a correlation between the number of days since the last use of the first service and the degree to which users are willing to switch from the current service to another service. There is also a correlation between the number of times users U1 to Un have used the first service within a past period and the degree to which users are willing to switch from the current service to another service. There is also a correlation between the amount of money spent by users U1 to Un to use the first service within a past period and the degree to which users are willing to switch from the current service to another service. The churn probability prediction device 10, having the above configuration, improves the accuracy of churn prediction by adding at least one of the date and time when users U1 to Un last used the first service, the number of times users U1 to Un used the first service within a past period, and the amount of money spent by users U1 to Un to use the first service within a past period to the training data of the first learning model LM1.

[0080] The churn probability prediction device 10 further includes a notification unit 114 and a notification control unit 115. The notification unit 114 notifies the user Uk of a message encouraging the user Uk to use the first service. The notification control unit 115 controls at least one of whether to notify the user Uk of the message and the content of the message, depending on the probability Pk predicted by the prediction unit 113.

[0081] By having the above-described configuration, the churn probability prediction device 10 is expected to improve the utilization rate of the first service by the user Uk.

[0082] In addition, in the churn probability prediction device 10, the content of the message indicates a benefit that encourages the use of the first service at a store where the first service and the second service are available. The notification control unit 115 controls the benefit according to the probability Pk predicted by the prediction unit 113.

[0083] Providing a benefit to encourage use of the first service in a store where both the first and second services are available can increase the usage rate of the first service, which is the company's own service, compared to the usage rate of the second service, which is a competing service, rather than providing a benefit to encourage use of the first service in a store where only the first service is available.By having the above-mentioned configuration, the churn probability prediction device 10 can reduce the possibility that the user Uk will churn the first service.In addition, the churn probability prediction device 10 can finely set the content of the benefit depending on the probability P predicted by the prediction unit 113.

[0084] 2: Second Embodiment Hereinafter, a churn probability prediction system 1A according to a second embodiment will be described with reference to the drawings. For simplicity of explanation, the following mainly describes the differences between the churn probability prediction system 1A and the churn probability prediction system 1. Furthermore, among the components included in the churn probability prediction system 1A, the same components as those included in the churn probability prediction system 1 will be designated by the same reference numerals, and explanations of their functions may be omitted.

[0085] 2-1: Configuration of Second Embodiment 2-1-1: Overall Configuration Compared to the churn probability prediction system 1 according to the first embodiment, the overall configuration of the churn probability prediction system 1A according to this embodiment includes a churn probability prediction device 10A instead of the churn probability prediction device 10. In other respects, the overall configuration of the churn probability prediction system 1A is the same as the overall configuration of the churn probability prediction system 1, and therefore is not shown in the figures.

[0086] 2-1-2: Configuration of the churn probability prediction device 10A Fig. 8 is a block diagram showing an example configuration of the churn probability prediction device 10A. The churn probability prediction device 10A differs from the churn probability prediction device 10 in that it includes a processing device 11A instead of the processing device 11 and a storage device 12A instead of the storage device 12.

[0087] The storage device 12A differs from the storage device 12 in that it includes a control program PR1A instead of the control program PR1 and a user database UDB2 instead of the user database UDB1. The storage device 12A also does not include the probability table TTB that the storage device 12 includes. On the other hand, the storage device 12A includes a second learning model LM2 that the storage device 12 does not include.

[0088] FIG. 9 shows an example of the configuration of user database UDB2. Comparing user database UDB2 with user database UDB1, dataset DS(M) included in user database UDB2 includes an item m in addition to the items included in dataset DS(M) included in user database UDB1. Item m is the message ID of the message notified to each of users U1 to Un by the churn probability prediction device 10A. In dataset DS(M), the message ID of the message notified to user U1 corresponding to user ID="001" is "A." In other words, the message notified to user U1 is the message corresponding to message ID="A" in the message database MDB illustrated in FIG. 4. The first row of dataset DS(M) indicates that user U1 used the first service for 2,000 yen as a result of the churn probability prediction device 10A notifying user U1 of the message corresponding to message ID="A."

[0089] Moreover, the data sets DS(M-1) to DS(M-3) have the same configuration as the data set DS(M).

[0090] In FIG. 8 , the second learning model LM2 is a learning model used by the notification control unit 115A (described later) to acquire the content of a message by inputting first usage information and second usage information acquired over a past period. The second learning model LM2 is a learning model trained to maximize the learning usage amount using training data indicating the learning usage status of the first application, the learning usage status of the second application, the learning content of the message, and the learning usage amount of the first service. The learning usage status of the first application and the learning usage status of the second application are, for example, at least a portion of the information contained in datasets DS(M-L) to DS(M-1) from month Y (M-L) to month Y (M-1) in the user database UDB2. The content of the message is the content of the message associated with the message ID of item m contained in datasets DS(M-L) to DS(M-1). The learning usage amount for the first service is the usage amount for 16 items included in datasets DS(M-L) to DS(M-1). The training data used to generate the second learning model LM2 includes multiple pairs of the learning usage status of the first application, the learning usage status of the second application, the learning content of the message, and the learning usage amount for the first service.

[0091] As an example, if the content of the message is a, the second learning model LM2 learns the following correspondence between X and Y for each user U1 to Un.

[0092] Alternatively, the second learning model LM2 may be a learning model trained to maximize the number of learning launches using training data indicating the learning usage status of the first application, the learning usage status of the second application, the learning content of the messages, and the learning launch count of the first application. Alternatively, the second learning model LM2 may be a learning model trained to maximize the learning conversion rate using training data indicating the learning usage status of the first application, the learning usage status of the second application, the learning content of the messages, and the learning conversion rate within the first application.

[0093] The processing device 11A functions as a first acquisition unit 111, a second acquisition unit 112, a prediction unit 113, a notification unit 114, and a notification control unit 115A, for example, by reading and executing a control program PR1A from the storage device 12A.

[0094] The notification control unit 115A acquires the content of the message by inputting the first usage information and the second usage information acquired in the past period into the second learning model LM2. The notification control unit 115A also sets the content of the message notified by the notification unit 114 to the content of the acquired message.

[0095] 2-2: Operation of Second Embodiment The operation of the churn probability prediction device 10A is similar to the operation of the churn probability prediction device 10 shown in FIG. 7, and therefore illustration and description thereof will be omitted.

[0096] 2-3: Advantages of the Second Embodiment In the churn probability prediction device 10A according to this embodiment, the notification control unit 115A acquires the content of an output message that is output by inputting the first usage information and the second usage information acquired in a past period into a second learning model LM2 that has been trained to maximize the learning usage amount using training data indicating the learning usage status of the first application, the learning usage status of the second application, the learning content of the message, and the learning usage amount of the first service. Furthermore, the notification control unit 115A sets the content of the message notified by the notification unit 114 to the content of the acquired output message.

[0097] With the above configuration, the churn probability prediction device 10A can notify the user Uk of a message with the aim of increasing the usage amount of the first service. As a result, the churn probability prediction device 10A can reduce the possibility that the user Uk will churn the first service.

[0098] Furthermore, in the churn probability prediction device 10A according to this embodiment, the notification control unit 115A acquires the content of an output message that is output by inputting the first usage information and the second usage information acquired in a past period into a second learning model LM2 that has been trained to maximize the number of startups for learning, using training data indicating the usage status for learning of the first application, the usage status for learning of the second application, the learning content of the message, and the number of startups for learning of the first application. Furthermore, the notification control unit 115A sets the content of the message notified by the notification unit 114 to the content of the acquired output message.

[0099] With the above configuration, the churn probability prediction device 10A can notify the user Uk of a message with the aim of increasing the number of times the first application is launched. As a result, the churn probability prediction device 10A can reduce the possibility that the user Uk will churn the first service.

[0100] Furthermore, in the churn probability prediction device 10A according to this embodiment, the notification control unit 115A acquires the content of an output message that is output by inputting the first usage information and the second usage information acquired in a past period into a second learning model LM2 that has been trained to maximize the learning conversion rate, using training data indicating the learning usage status of the first application, the learning usage status of the second application, the learning content of the message, and the learning conversion rate within the first application. Furthermore, the notification control unit 115A sets the content of the message notified by the notification unit 114 to the content of the acquired output message.

[0101] With the above configuration, the churn probability prediction device 10A can notify the user Uk of a message with the aim of increasing the conversion rate in the first application. As a result, the churn probability prediction device 10A can reduce the possibility that the user Uk will churn the first service.

[0102] 3: Modifications The present disclosure is not limited to the above-described exemplary embodiments. Specific modifications are exemplified below. Two or more modifications selected from the following examples may be combined.

[0103] 3-1: Modification 1 In the above embodiment, as an example, the first service was a company service for the user of the churn probability prediction device 10 and the churn probability prediction device 10A. However, the first service does not have to be a company service for the user.

[0104] 4: Others (1) In the above-described embodiment, storage device 12, storage device 12A, and storage device 22 are exemplified by ROM and RAM, but they may also be flexible disks, magneto-optical disks (e.g., compact disks, digital versatile disks, Blu-ray (registered trademark) disks), smart cards, flash memory devices (e.g., cards, sticks, key drives), CD-ROMs (Compact Disc-ROMs), registers, removable disks, hard disks, floppy (registered trademark) disks, magnetic strips, databases, servers, or other suitable storage media. The program may also be transmitted from a network via a telecommunications line. The program may also be transmitted from a communications network (NET) via a telecommunications line.

[0105] (2) In the above-described embodiments, the described information, signals, etc. may be represented using any of a variety of different technologies. For example, data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or magnetic particles, optical fields or photons, or any combination thereof.

[0106] (3) In the above-described embodiment, input and output information may be stored in a specific location (for example, a memory) or may be managed using a management table. Input and output information may be overwritten, updated, or added to. Output information may be deleted. Input information may be transmitted to another device.

[0107] (4) In the above-described embodiment, the determination may be made based on a value (0 or 1) represented using one bit, a Boolean value (true or false), or a comparison of numerical values ​​(e.g., comparison with a predetermined value).

[0108] (5) The order of the exemplary procedures, sequences, flowcharts, etc. illustrated in the above-described embodiments may be rearranged unless inconsistent. For example, the methods described in this disclosure present elements of various steps using an example order, and are not limited to the particular order presented.

[0109] (6) Each function illustrated in Figures 1 to 9 is realized by any combination of hardware and / or software. Furthermore, the method for realizing each functional block is not particularly limited. That is, each functional block may be realized using a single device that is physically or logically coupled, or may be realized using two or more physically or logically separated devices that are connected directly or indirectly (e.g., via wire, wirelessly, etc.) and these multiple devices. A functional block may be realized by combining software with the single device or the multiple devices.

[0110] (7) The programs exemplified in the above-described embodiments should be broadly construed to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, threads of execution, procedures, functions, etc., regardless of whether they are called software, firmware, middleware, microcode, hardware description language, or by other names.

[0111] Software, instructions, information, etc. may also be transmitted or received over a transmission medium. For example, if software is transmitted from a website, server, or other remote source using wired technologies (such as coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL)), and / or wireless technologies (such as infrared, microwave), then these wired and / or wireless technologies are included within the definition of transmission media.

[0112] (8) In each of the foregoing embodiments, the terms "system" and "network" are used interchangeably.

[0113] (9) The information, parameters, etc. described in this disclosure may be expressed using absolute values, relative values ​​from a predetermined value, or corresponding other information.

[0114] (10) In the above-described embodiments, the terminal devices 20_1 to 20_n may be mobile stations (MS). A mobile station may also be referred to by those skilled in the art as a subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, or some other appropriate term. In the present disclosure, terms such as "mobile station," "user terminal," "user equipment (UE)," and "terminal" may be used interchangeably.

[0115] (11) In the above-described embodiments, the terms "connected," "coupled," or any variations thereof refer to any direct or indirect connection or coupling between two or more elements, and may include the presence of one or more intermediate elements between two elements that are "connected" or "coupled" to each other. The coupling or connection between elements may be a physical coupling or connection, a logical coupling or connection, or a combination thereof. For example, "connected" may be read as "access." As used in this disclosure, two elements may be considered to be "connected" or "coupled" to each other using at least one of one or more wires, cables, and printed electrical connections, as well as electromagnetic energy having wavelengths in the radio frequency range, microwave range, and optical (both visible and invisible) range, as some non-limiting and non-exhaustive examples.

[0116] (12) In the above embodiments, the phrase "based on" does not mean "based only on," unless otherwise specified. In other words, the phrase "based on" means both "based only on" and "based at least on."

[0117] (13) As used in this disclosure, the terms "determining" and "determining" may encompass a wide variety of actions. "Determining" and "determining" may include, for example, judging, calculating, computing, processing, deriving, investigating, looking up, searching, inquiring (e.g., searching a table, database, or other data structure), and ascertaining something that is considered to be a "determining." Also, "determining" and "determining" may include receiving (e.g., receiving information), transmitting (e.g., sending information), input, output, accessing (e.g., accessing data in memory), and so on. Furthermore, "judgment" and "decision" can include regarding resolving, selecting, choosing, establishing, comparing, etc. as having been "judged" or "decided." In other words, "judgment" and "decision" can include regarding some action as having been "judged" or "decided." Furthermore, "judgment (decision)" can be interpreted as "assuming," "expecting," "considering," etc.

[0118] (14) In the above embodiments, when the terms "include," "including," and variations thereof are used, these terms are intended to be inclusive, similar to the term "comprising." Furthermore, the term "or," as used in this disclosure, is not intended to be an exclusive or.

[0119] (15) In this disclosure, where articles are added by translation, such as a, an, and the in English, the disclosure may include that the nouns following these articles are plural.

[0120] (16) In this disclosure, the term "A and B are different" may mean "A and B are different from each other." The term may also mean "A and B are each different from C." Terms such as "separate" and "combined" may also be interpreted in the same way as "different."

[0121] (17) Each aspect / embodiment described in this disclosure may be used alone, in combination, or switched depending on the implementation. Notification of predetermined information (e.g., notification that "X is true") is not limited to explicit notification, but may be implicit (e.g., not notifying the predetermined information).

[0122] Although the present disclosure has been described in detail above, it is clear to those skilled in the art that the present disclosure is not limited to the embodiments described herein. The present disclosure can be implemented in modified and altered forms without departing from the spirit and scope of the present disclosure as defined by the claims. Therefore, the description of the present disclosure is intended to be illustrative and does not have any limiting meaning on the present disclosure.

[0123] 1: churn probability prediction system, 1A: churn probability prediction system, 10: churn probability prediction device, 10A: churn probability prediction device, 11: processing device, 11A: processing device, 12: storage device, 12A: storage device, 13: communication device, 14: display device, 15: input device, 20: terminal device, 20_1: terminal device, 20_k: terminal device, 20_n: terminal device, 21: processing device, 22: storage device, 23: communication device, 24: display device, 25: input device, 111: first acquisition unit, 112: second acquisition unit, 113: prediction unit, 114: notification unit, 115: notification control unit, 115A: notification control unit, 211: acquisition unit, 212: display Control unit, 213: Execution unit, 214: Communication control unit, A: Application, AP1: First application program, AP2: Second application program, DS: Data set, LM1: First learning model, LM2: Second learning model, MDB: Message database, NET: Communication network, P: Probability, P1: Probability, PR1: Control program, PR1A: Control program, PR2: Control program, Pk: Probability, TTB: Probability table, U: User, U1: User, U2: User, UDB1: User database, UDB2: User database, Uk: User, Un: User

Claims

1. A cancellation probability prediction device comprising: a first acquisition unit that acquires, for a target user, first usage information regarding the usage status of a first application used to provide a first service; a second acquisition unit that acquires, for the target user, second usage information regarding the usage status of a second application used to provide a second service of the same type as the first service; and a prediction unit that predicts the probability that the target user will cancel the first service in the future by inputting the first usage information and the second usage information acquired during a past period into a first learning model that has learned the relationship between the learning usage status of the first application and the learning usage status of the second application, and the learning probability that the first service will be canceled in the future, wherein the second usage information includes at least one of the number of times and the startup time of the second application.

2. The cancellation probability prediction device according to claim 1, wherein the first usage information includes attribute information relating to attributes of the target user.

3. The cancellation probability prediction device of claim 1, wherein the first usage information includes at least one of the date and time when the target user last used the first service, the number of times the target user used the first service within the past period, and the amount of money spent by the target user to use the first service within the past period.

4. The cancellation probability prediction device of claim 1, further comprising: a notification unit that notifies the target user of a message encouraging the target user to use the first service; and a notification control unit that controls at least one of whether or not to notify the message and the content of the message depending on the probability predicted by the prediction unit.

5. The cancellation probability prediction device of claim 4, wherein the content of the message indicates a benefit encouraging use of the first service at a store where the first service and the second service are available, and the notification control unit controls the benefit according to the probability predicted by the prediction unit.

6. The cancellation probability prediction device of claim 4, wherein the notification control unit acquires the content of an output message output by inputting the first usage information and the second usage information acquired in a past period into a second learning model that has been trained to maximize the learning usage amount using training data indicating the learning usage status of the first application, the learning usage status of the second application, the learning content of the message, and the learning usage amount of the first service, and the content of the message notified by the notification unit is the content of the acquired output message.

7. The cancellation probability prediction device of claim 4, wherein the notification control unit uses training data indicating the usage status of the first application for learning, the usage status of the second application for learning, the learning content of the message, and the number of times the first application has been launched for learning to obtain the content of an output message that is output by inputting the first usage information and the second usage information acquired during a past period into a second learning model that has been trained to maximize the number of times the first application has been launched for learning, and the content of the message notified by the notification unit is the content of the acquired output message.

8. The cancellation probability prediction device of claim 4, wherein the notification control unit acquires the content of an output message output by inputting the first usage information and the second usage information acquired in a past period into a second learning model that has been trained to maximize the learning conversion rate using training data indicating the learning usage status of the first application, the learning usage status of the second application, the learning content of the message, and a learning conversion rate within the first application, and the content of the message notified by the notification unit is the content of the acquired output message.

Images