Intervention device and learning device

Abstract

Provided are an intervention device and a learning device with which it is possible to improve the probability of completion by a user.　An intervention device 1 comprises: a user behavior acquisition unit 11 that acquires a user behavior; a completion probability estimation unit 12 that estimates a completion probability until a final behavior on the basis of the user behavior; and an intervention processing unit 13 that executes intervention processing on the basis of the completion probability.

Description

Intervention and learning devices

[0001] The present disclosure relates to interventional and learning devices.

[0002] Generally, when a user purchases a product through online shopping, the user performs a step-by-step process leading up to a final purchasing action that confirms the purchase. For example, the user may perform the step-by-step process by searching for a desired product, selecting the searched product, entering a payment method, entering a delivery method, and finally confirming the purchase of the product. Patent Document 1 describes a customer viability estimation system that analyzes a customer viability based on data on a user's purchasing behavior and pre-purchase behavior (e.g., behavior of adding a product to a cart) for a product, and outputs the analysis results.

[0003] JP 2013-73534 A

[0004] As a user performs actions step by step up to a purchase action (final action), an event may occur that prevents the user from performing the final action. From the perspective of increasing the probability that the user will perform the final action, it is desirable to suppress the occurrence of such an event. The customer viability estimation system described in Patent Literature 1 outputs the analysis results of the customer viability rate, but does not suppress the occurrence of the above-mentioned events to increase the probability that the user will perform the final action (completion probability).

[0005] In order to solve the above problem, the present disclosure aims to provide an intervention device and a learning device that can improve a user's completion probability.

[0006] The intervention device of the present disclosure includes a user behavior acquisition unit that acquires user behavior, a completion probability estimation unit that estimates the completion probability up to the final behavior based on the user behavior, and an intervention processing unit that executes intervention processing based on the completion probability.

[0007] The learning device disclosed herein includes a learning data acquisition unit that acquires user actions, including a predetermined number of user actions up to the time of a predetermined target action and the user's final action, and a model generation unit that generates a completion probability estimation model through machine learning using the predetermined number of user actions up to the time of the target action as explanatory variables and reaching the final action as a target variable.

[0008] According to the present disclosure, the user's completion rate can be improved.

[0009] 1 is a diagram illustrating an intervention device, a learning device, a user terminal, and a user behavior database according to the present embodiment. FIG. 2 is a diagram illustrating an example of a series of user behaviors. FIG. 3 is a flowchart illustrating an example of the flow of user behavior on a shopping site. FIG. 4 is a diagram illustrating an example of a table stored in the user behavior database. FIG. 5 is a block diagram illustrating the functional configuration of the intervention device according to the present embodiment. FIG. 6 is a diagram illustrating an example of a process of calculating a lifting effect. FIG. 7 is a diagram illustrating an example of intervention processing. FIG. 8 is a flowchart illustrating an example of the operation of the intervention device. FIG. 9 is a block diagram illustrating the functional configuration of the learning device according to the present embodiment. FIG. 10 is a diagram illustrating an example of a process of classifying users. FIG. 11 is a diagram illustrating an example of a process of calculating the degree of association between a sub-action and a next action. FIG. 12 is a diagram illustrating an example of a process of acquiring an important action. (a) is a diagram illustrating an example of learning related to a dropout user. (b) is a diagram illustrating an example of learning related to a completed user. A schematic diagram of a completion probability estimation model. A diagram illustrating an example of a process of acquiring completion probability. A diagram illustrating an example of a process of calculating a difference in completion probability. A flowchart illustrating an example of the operation of the learning device according to the present embodiment. FIG. 11 is a diagram illustrating an example of the hardware configuration of the intervention device and the learning device according to the present embodiment.

[0010] The present disclosure will be described with reference to the accompanying drawings. Whenever possible, the same parts are designated by the same reference numerals and redundant description will be omitted.

[0011] FIG. 1 is a diagram showing an intervention device 1, a learning device 2, a user terminal 3, and a user behavior database DB according to this embodiment. The intervention device 1, the learning device 2, the multiple user terminals 3, and the user behavior database DB are communicatively connected via a network NW. The network NW is any wired or wireless communication network. The network NW is, for example, the Internet, a public line network, a dedicated line network, etc. Note that the network NW may be configured to include multiple different communication networks.

[0012] The user terminal 3 is a terminal owned by a user, such as a smartphone, tablet terminal, or personal computer. The user is, for example, a person who purchases a product through online shopping or the like using the user terminal 3. In this embodiment, the user accesses a shopping site S via the user terminal 3. The user performs a final action. The final action refers to a final action to achieve a desired goal. In this embodiment, the final action is an action of confirming the purchase of a product on the shopping site S. The final action is, for example, an action of pressing a purchase confirmation button on the shopping site S.

[0013] A user performs various user actions in stages before executing a final action. In the process of a user performing these user actions, an event may occur that prevents the user from performing the final action. Such an event includes, for example, abandonment. Abandonment refers to a situation in which a user performs a certain user action, but does not perform the next user action and does not reach the final action. From the perspective of improving the probability that a user will perform the final action (completion probability), it is necessary to suppress the occurrence of such an event.

[0014] The intervention device 1 executes a predetermined intervention process to improve the user's completion probability. The intervention process is a process of having the user execute an important action (described later) to improve the user's completion probability. The intervention device 1 communicates with the user terminal 3 and acquires the user action. The intervention device 1 executes the intervention process based on the acquired user action. The user action database DB transmits and receives data to and from the user terminal 3. The user action database DB acquires and stores the user action received from the user terminal 3. The learning device 2 learns a completion probability estimation model used by the intervention device 1 based on the user action recorded in the user action database DB.

[0015] 2 is a diagram showing an example of a series of user actions. The user actions include a final action and actions that the user performs until the final action is performed. In this embodiment, the user actions are a series of actions until the user confirms the purchase of a product on the shopping site S (performs the final action). The user performs the user actions in stages up to the final action.

[0016] As an example, a user may access a product selection page P1 of a shopping site S via a user terminal 3. After selecting a product, the user may access a page P2 for adding the product to the cart and add the product to the cart. Here, some users may access an address input page P3 after adding the product to the cart and input a delivery address. After inputting the address, some users may access a payment method input page P4 and input a payment method. After inputting the payment method, the user may access a final confirmation page P5 to make a final confirmation regarding the purchase of the product. After making the final confirmation, the user may access a product purchase confirmation page P6 to confirm the purchase of the product. In this case, the user has performed a final action.

[0017] Furthermore, other users, different from the aforementioned portion of users, may access the frequently asked questions page P7 after adding a product to their cart to conduct research and consideration regarding purchasing the product. Some of the users who conducted the research and consideration may access the address input page P3. On the other hand, other users, different from the portion of users who conducted the research and consideration, may give up on purchasing the product on the shopping site S and access the store visit reservation page P8 to make a reservation for a visit. A user who has made a reservation for a visit may visit the store and purchase the product. In this case, the user who visited the store will have purchased the product by another method without performing a final action. In other words, the user has dropped out of a series of user actions.

[0018] 3 is a flowchart showing an example of the flow of user actions on a shopping site S. As an example, first, a user selects a product on the shopping site S (step S1). In step S1, the user selects a product. Next, the user adds the product selected in step S1 to a cart (step S2). Next, the user inputs a delivery address for the product (step S3). In step S3, the user inputs, for example, their own address as the delivery address. Next, the user inputs a payment method (step S4). Next, the user performs a final confirmation to purchase the product (step S5). Finally, the user confirms the purchase of the product as a final action (step S6).

[0019] The user terminal 3 transmits user actions performed by the user to the user action database DB. For example, each time the user updates a page on the shopping site S, the user terminal 3 transmits the pages viewed by the user as user actions to the user action database DB. For example, the user terminal 3 may transmit user actions to the user action database DB using a known tool that collects browsing history of a web browser.

[0020] 4 is a diagram showing an example of a table stored in the user behavior database DB. The user behavior database DB stores, for each user, user behavior transmitted from the user terminal 3. The user behavior database DB stores the user name (or ID information that uniquely identifies the user), the time the user behavior was performed, the channel, the type of user behavior, and the content of the user behavior.

[0021] User actions are classified into, for example, major actions and minor actions. Major actions are, for example, preset and can be changed as appropriate. Major actions are, for example, each of the minimum number of user actions required to perform a final action. In this embodiment, major actions are the user actions of steps S1 to S6 shown in FIG. 3. Minor actions are user actions performed between multiple major actions. A minor action is, for example, an action of browsing another page to research and consider a product after selecting and adding it to a cart and before entering the information required to pay the fee.

[0022] Major actions include a target action, a next action, and a final action. The target action is one of the major actions excluding the final action. The target action is, for example, set in advance and can be changed as appropriate. A method for determining the target action will be described later. In this embodiment, an example will be described in which the target action is the action of adding to cart (step S2). The next action is the major action that is executed next after the target action. For example, if the target action is the action of adding to cart (step S2), the next action is the action of entering an address (step S3).

[0023] The minor actions include important actions. Important actions are minor actions that increase the probability that a user who has performed a target action will perform the next action. Important actions are acquired in advance by the learning device 2, which will be described later. In this embodiment, an example will be described in which the important action is the action of viewing a page that summarizes matters that users are likely to have questions about (hereinafter, sometimes referred to as an "FAQ page").

[0024] The user behavior database DB stores a step number for each user. The step number is the number of the major behavior last performed by the user. In this embodiment, the step number is the number of the user behavior performed latest among the major behaviors shown in Figure 3. For example, a user with step number 2 is a user who selected a product and added it to their cart, but did not enter a shipping address for the product. A user with step number 6 is a user who confirmed the purchase of the product and performed the final behavior. In the following, the number of the user behavior corresponding to the final behavior may be referred to as the final step number. In the example of Figure 3, the final step number is 6.

[0025] 5 is a block diagram showing the intervention device 1 according to this embodiment. The intervention device 1 executes a predetermined intervention process. By executing the intervention process, the intervention device 1 suppresses the occurrence of the above-mentioned events. The intervention process will be described in detail later. The intervention device 1 includes, as its functional configuration, a user behavior acquisition unit 11, a completion probability estimation unit 12, and an intervention processing unit 13.

[0026] The user behavior acquisition unit 11 acquires user behavior. In this embodiment, the user behavior acquisition unit 11 communicates with the user terminal 3 via the network NW to acquire the user behavior. The user behavior acquisition unit 11 acquires the time when the user behavior was performed, the type of the user behavior, and the content of the user behavior. The user behavior acquisition unit 11 periodically acquires the user behavior, for example, by periodically communicating with the user terminal 3.

[0027] When acquiring user behavior on the web, the user behavior acquisition unit 11 may acquire the user behavior using, for example, a tool for collecting browsing history of a web browser.When acquiring user behavior during telephone calls, the user behavior acquisition unit 11 may acquire the user behavior using, for example, a response system of a call center or the like.

[0028] The user behavior acquisition unit 11 identifies user behaviors acquired from, for example, the user terminal 3 that correspond to predetermined major behaviors, and identifies user behaviors that do not correspond to major behaviors as minor behaviors.

[0029] The completion probability estimation unit 12 estimates the completion probability up to the final action based on the user action acquired by the user action acquisition unit 11. The completion probability estimation unit 12 is a completion probability estimation model trained by machine learning using user actions up to the time of a preset target action as an explanatory variable and reaching the final action as a target variable. The completion probability estimation model is generated in advance by the learning device 2, which will be described later.

[0030] The completion probability estimation unit 12 determines whether the latest user action is a target action based on the user action acquired by the user action acquisition unit 11. The completion probability estimation unit 12 acquires the user action most recently acquired by the user action acquisition unit 11 as the latest user action. If the latest user action is a target action, the completion probability estimation unit 12 estimates a completion probability up to the final action based on the user actions up to the target action.

[0031] The completion probability estimation unit 12 estimates the completion probability of user actions that do not include important actions. The completion probability estimation unit 12 acquires user actions up to the time of the target action acquired by the user action acquisition unit 11 as user actions that do not include important actions. The completion probability estimation unit 12 inputs the user actions that do not include important actions into a completion probability estimation model, and estimates the completion probability of the user actions that do not include important actions.

[0032] The completion probability estimation unit 12 estimates the completion probability of a user action including an important action. The completion probability estimation unit 12 acquires, as a user action including an important action, a user action obtained by adding a sub-action acquired as an important action to a sub-action after the target action, for the user action up to the target action acquired by the user action acquisition unit 11. The completion probability estimation unit 12 inputs the user action including the important action into a completion probability estimation model, and estimates the completion probability of the user action including the important action.

[0033] The completion probability estimation unit 12 calculates a lifting effect based on the estimated completion probabilities (the completion probability of the user behavior including the important behavior and the completion probability of the user behavior not including the important behavior). The lifting effect is the degree to which the completion probability can be improved by the user performing the important behavior. In this embodiment, the lifting effect is the difference between the completion probability of the user behavior including the important behavior and the completion probability of the user behavior not including the important behavior.

[0034] 6 is a diagram illustrating an example of a process for calculating the lifting effect. In the example of FIG. 6, the completion probability estimation unit 12 outputs a value of X+δ% as the completion probability of a user behavior including an important behavior, and outputs a value of X% as the completion probability of a user behavior not including an important behavior. The completion probability estimation unit 12 calculates δ% as the lifting effect, which is the difference between the above respective completion probabilities.

[0035] The intervention processing unit 13 executes the intervention processing based on the completion probability estimated by the completion probability estimation unit 12. In this embodiment, the intervention processing unit 13 determines whether to execute the intervention processing based on the completion probability of the user behavior not including the important behavior estimated by the completion probability estimation unit 12 and the lifting effect calculated by the completion probability estimation unit 12.

[0036] The intervention processing unit 13 determines whether the completion probability of user actions excluding important actions is equal to or less than a predetermined value. The predetermined value is, for example, set in advance and can be changed as appropriate. The intervention processing unit 13 determines whether the lifting effect is equal to or greater than a predetermined value. The predetermined value is, for example, set in advance and can be changed as appropriate.

[0037] In this embodiment, if the completion probability of user actions that do not include important actions is equal to or less than a predetermined value and the lifting effect is equal to or greater than a predetermined value, the intervention processing unit 13 determines to execute the intervention processing. Otherwise, the intervention processing unit 13 determines not to execute the intervention processing.

[0038] The intervention processing unit 13 executes, as an intervention process, a process of performing an important action that has been performed between the target action and the next action and that has a strong degree of association with the next action. The degree of association is the strength of the association of a sub-action performed between the target action and the next action with the next action. A user who has performed a sub-action with a strong degree of association is more likely to perform the next action than a user who has not performed a sub-action with a strong degree of association. The degree of association of the sub-action is output in advance by the learning device 2, which will be described later.

[0039] 7 is a diagram illustrating an example of the intervention process. In this embodiment, the intervention processing unit 13 executes, as the intervention process, a process of displaying a link to an FAQ page on the add-to-cart page P2. By referring to the FAQ page, the user can resolve their doubts before finalizing the purchase of the product. Therefore, by executing such an intervention process, the intervention processing unit 13 can increase the probability (completion probability) that the user will perform the action (final action) of finalizing the purchase of the product.

[0040] In this embodiment, the intervention processing unit 13 performs the intervention processing by displaying a pop-up P such as "Ask a question via chat" on the add-to-cart page P2. When the user clicks on the pop-up P, the user accesses the above-mentioned FAQ page. The user accesses the fee diagnosis page from the FAQ page, for example. The user can access the fee diagnosis page, for example, to check the monthly fee to be paid if the product is purchased.

[0041] The above-described content of the intervention process is merely an example and can be changed as appropriate. As the intervention process, the intervention processing unit 13 may, for example, send a direct mail to the user terminal 3, or display a link to a page for receiving chat support or telephone support.

[0042] Next, the operation of the intervention device 1 will be described. Fig. 8 is a flowchart showing an example of the operation of the intervention device 1. First, the user behavior acquisition unit 11 acquires user behavior (step S11). In step S11, the user behavior acquisition unit 11 communicates with the user terminal 3 via the network NW and acquires the user behavior.

[0043] In step S11, the user behavior acquisition unit 11 identifies user behaviors acquired, for example, from the user terminal 3 that correspond to pre-set major behaviors, and identifies user behaviors that do not correspond to major behaviors as minor behaviors.

[0044] Next, the completion probability estimation unit 12 determines whether the latest user action is a predetermined target action based on the user action acquired in step S11 (step S12). In step S12, the completion probability estimation unit 12 acquires the user action last acquired in step S11 as the latest user action. The target action in step S12 is predetermined in one step of the operation of the learning device 2, which will be described later.

[0045] If the latest user action is not the target action (step S12: NO), the user action acquisition unit 11 executes step S11 again. If the latest user action is the target action (step S12: YES), the completion probability estimation unit 12 executes step S13.

[0046] In step S13, the completion probability estimation unit 12 estimates the completion probability up to the final action based on the user actions acquired in step S11. The completion probability estimation unit 12 is a completion probability estimation model that has been machine-learned using user actions up to the time of a preset target action as an explanatory variable and reaching the final action as a target variable. The completion probability estimation model used in step S13 is generated in advance in one step of the operation of the learning device 2, which will be described later.

[0047] In step S13, the completion probability estimation unit 12 estimates the completion probability of the user behavior that does not include the important behavior. Also in step S13, the completion probability estimation unit 12 estimates the completion probability of the user behavior that includes the important behavior.

[0048] Next, the completion probability estimation unit 12 calculates a lifting effect based on the respective completion probabilities estimated in step S13 (the completion probability of the user behavior including the important behavior and the completion probability of the user behavior not including the important behavior) (step S14). In step S14, the completion probability estimation unit 12 calculates the lifting effect as the difference between the completion probability of the user behavior including the important behavior and the completion probability of the user behavior not including the important behavior.

[0049] Next, the intervention processing unit 13 determines whether to execute the intervention processing (step S15). In step S15, the intervention processing unit 13 determines whether to execute the intervention processing based on the completion probability of the user behavior not including the important behavior estimated by the completion probability estimation unit 12 and the lifting effect calculated by the completion probability estimation unit 12.

[0050] In step S15, the intervention processing unit 13 determines whether the completion probability of user actions that do not include important actions is equal to or less than a predetermined value. Also, in step S15, the intervention processing unit 13 determines whether the lifting effect is equal to or greater than a predetermined value. In this embodiment, if the completion probability of user actions that do not include important actions is equal to or less than a predetermined value and the lifting effect is equal to or greater than a predetermined value, the intervention processing unit 13 determines to execute the intervention processing. In other cases, the intervention processing unit 13 determines not to execute the intervention processing. If it is determined to execute the intervention processing, the intervention processing unit 13 executes step S16. If it is determined not to execute the intervention processing, the intervention device 1 completes the series of processes.

[0051] In step S16, the intervention processing unit 13 executes the intervention processing (step S16). In step S16, the intervention processing unit 13 executes a process of performing an important action that has a high degree of relevance to the next action and that has been performed between the target action and the next action. The degree of relevance of the small action and the important action are acquired in advance in one step of the operation of the learning device 2, which will be described later. After going through the above steps, the intervention device 1 completes a series of operations.

[0052] Next, a learning device 2 according to this embodiment will be described. The learning device 2 generates the above-described completion probability estimation model. In this embodiment, the learning device 2 acquires important actions to be used in the intervention process executed by the intervention device 1. As shown in FIG. 9 , the learning device 2 includes, as its functional configuration, a learning data acquisition unit 21, a classification unit 22, an output model 23, an extraction unit 24, an important action acquisition unit 25, a model generation unit 26, a completion probability acquisition unit 27, a calculation unit 28, and a determination unit 29.

[0053] The learning data acquisition unit 21 acquires user actions including a predetermined number of user actions up to a preset target action and the user's final action. In this embodiment, the learning data acquisition unit 21 refers to the user action database DB and acquires the type of user action, the content of the user action, and the step number for each user.

[0054] The learning data acquisition unit 21 identifies major actions from the acquired user actions. In this embodiment, the major actions are set in advance. The learning data acquisition unit 21 identifies user actions that correspond to the predetermined major actions from the acquired user actions by referring to the user action database DB.

[0055] In this embodiment, the learning data acquisition unit 21 determines the target behavior based on the step number acquired for each user. The learning data acquisition unit 21 determines the major behavior with the step number with the largest number of users as the target behavior, excluding the major behavior with the final step number. However, the method for determining the target behavior is not particularly limited. The target behavior may be set in advance by, for example, the user of the learning device 2. Below, an example will be described in which the learning data acquisition unit 21 determines the behavior of adding a product to a cart (step S2) as the target behavior.

[0056] In this embodiment, the learning data acquisition unit 21 further acquires one or more small actions that occur between the target action and the next action as user actions. The learning data acquisition unit 21 identifies, from the acquired user actions by referring to the user action database DB, user actions that do not correspond to major actions as small actions. From the identified small actions, the learning data acquisition unit 21 acquires one or more small actions that occur between the target action and the next action.

[0057] 10 is a diagram illustrating an example of a process for classifying users. The classification unit 22 classifies users into either a first user group or a second user group. In this embodiment, the first user group includes users who have reached the final action. The second user group includes users who have not reached the final action.

[0058] The classification unit 22 determines whether the user's step number is the final step number based on the step number acquired for each user by the learning data acquisition unit 21. If the classification unit 22 determines that the user's step number is the final step number, it classifies the user into a first user group. If the classification unit 22 determines that the user's step number is not the final step number, it classifies the user into a second user group.

[0059] The output model 23 outputs the degree of association between one or more sub-actions acquired by the learning data acquisition unit 21 and a next action. The output model 23 outputs the degree of association between the sub-actions and the next action using known means. In this embodiment, the output model 23 analyzes the sequence of actions from the target action to the next action based on the user actions acquired by the learning data acquisition unit 21. As a method for the output model 23 to analyze the sequence of actions, any known method can be adopted as long as it is a method that can analyze the strength of association (contribution) between user actions recorded in a sequence.

[0060] For the above analysis, it is conceivable to apply a technology used in natural language processing, such as BERT (Bidirectional Encoder Representations from Transformers). In this case, the output model 23 may be generated by applying the fact that in natural language processing, the strength of association between tokens defined by words or the like is obtained as attention (degree of association). Each user action is defined as a token, and a model that outputs the magnitude of the degree of association between user actions is trained.

[0061] 11 is a diagram showing an example of a process for calculating the degree of association between a sub-action and a subsequent action. The output model 23 outputs the degree of association between the sub-action and the subsequent action for each sub-action that occurs between the target action and the subsequent action.

[0062] 11 shows an example in which a user performs sub-actions A, B, and C between a target action and a next action. The output model 23 outputs, for example, a value of 0.2 as the degree of association between sub-action A and the next action. The output model 23 outputs, for example, a value of 0.5 as the degree of association between sub-action B and the next action. The output model 23 outputs, for example, a value of 0.1 as the degree of association between sub-action C and the next action.

[0063] 12 is a diagram illustrating an example of a process for acquiring important actions. The extraction unit 24 extracts a predetermined number of small actions of the first user group and a predetermined number of small actions of the second user group based on the degree of association output by the output model 23.

[0064] In this embodiment, the extraction unit 24 extracts a predetermined number of small actions with the highest degree of association from among the small actions performed by users in the first user group. The extraction unit 24 extracts a predetermined number of small actions with the highest degree of association from among the small actions performed by users in the second user group. However, the method by which the extraction unit 24 extracts small actions is not limited to the method described above. For example, the extraction unit 24 may extract a predetermined number of small actions with a degree of association equal to or greater than a predetermined value from among the small actions performed by users in the first user group. For example, the extraction unit 24 may extract a predetermined number of small actions with a degree of association equal to or greater than a predetermined value from among the small actions performed by users in the second user group.

[0065] The predetermined number is set in advance, for example, and can be changed as appropriate. In this embodiment, the predetermined number is 2. In the example of Fig. 12, the extraction unit 24 extracts small actions A and small actions B as the predetermined number of small actions of the first user group. The extraction unit 24 extracts small actions A and small actions C as the predetermined number of small actions of the second user group.

[0066] The important action acquisition unit 25 acquires an important action from one or more small actions based on the degree of association output by the output model 23. In this embodiment, the important action acquisition unit 25 acquires, as an important action, a small action that is present in the predetermined number of small actions of the first user group extracted by the extraction unit 24 but is not present in the predetermined number of small actions of the second user group.

[0067] 12 , the important behavior acquisition unit 25 acquires, as an important behavior, small behavior B, which is a small behavior that is included in small behaviors A and B of the first user group extracted by the extraction unit 24 and is not included in small behaviors A and C of the second user group extracted by the extraction unit 24. The important behavior acquisition unit 25 may acquire multiple important behaviors.

[0068] 13(a) is a diagram showing an example of learning related to dropout users by the model generation unit 26. FIG. 13(b) is a diagram showing an example of learning related to complete users by the model generation unit 26. A dropout user refers to a user who dropped out without performing a final action, and a complete user refers to a user who performed a final action. The model generation unit 26 generates a completion probability estimation model by machine learning, using a predetermined number of user actions up to the time of the target action acquired by the learning data acquisition unit 21 as explanatory variables and reaching the final action as a target variable.

[0069] The model generation unit 26 generates a completion probability estimation model by, for example, setting a model in which user behavior up to the target behavior is used as an explanatory variable and whether the user has reached the final behavior is used as an objective variable, and optimizing the coefficients (parameters) of the model through learning. The model may be a linear model such as logistic regression, or a nonlinear model such as a multilayer neural network. The model may be any model that can set a one-dimensional output for a multidimensional input.

[0070] For example, the explanatory variables may be generated by applying a technology used in natural language processing such as BERT. As an example, first, each user action is defined as a token, and a model is trained that uses a sequence of these as input and outputs a semantic vector for the token at a certain point in time. Then, the user actions up to the time of the target action are input to the model, and the output semantic vector at the time of the target action may be used as the explanatory variable of the completion probability estimation model. The objective variable may be created by setting the value to 1 when a user's step number is the final step number and 0 when a user's step number is not the final step number.

[0071] Fig. 14 is a schematic diagram of a completion probability estimation model. Fig. 15 is a diagram showing an example of a process for acquiring a completion probability. As shown in Fig. 14 and Fig. 15 , the completion probability estimation model generated by the model generation unit 26 estimates the probability that the user's step number is the final step number (completion probability) using user behavior up to the time of a target behavior as input.

[0072] As shown in Figure 15, as an example, the completion probability estimation model receives input of each user action up to the target action. The completion probability estimation model inputs each user action up to the target action into a BERT or the like, and outputs a semantic vector for each user action. The completion probability estimation model inputs the output semantic vector into a model such as logistic regression, and outputs a completion probability. However, the processing content of the completion probability estimation model can be changed as appropriate.

[0073] 16 is a diagram illustrating an example of a process for calculating the difference in completion probability. The completion probability acquisition unit 27 acquires, for each user, a completion probability of reaching a final action of user actions that do not include a significant action, using the completion probability estimation model generated by the model generation unit 26. The completion probability acquisition unit 27 inputs user actions up to the target action into the completion probability estimation model as user actions that do not include a significant action. As a result, the completion probability acquisition unit 27 acquires the completion probability of user actions that do not include a significant action.

[0074] The completion probability acquisition unit 27 acquires, for each user, a completion probability of user actions including an important action using the completion probability estimation model. The completion probability acquisition unit 27 inputs user actions, which are user actions up to the time of the target action and which include an important action acquired by the important action acquisition unit 25 after the time of the target action, into the completion probability estimation model as user actions including an important action. In this way, the completion probability acquisition unit 27 acquires the completion probability of user actions including an important action. The completion probability acquisition unit 27 acquires the completion probability of user actions including an important action for each important action acquired by the important action acquisition unit 25.

[0075] 16 , the completion probability acquisition unit 27 acquires a value of X+δ% as the completion probability of a user behavior including an important behavior. The completion probability acquisition unit 27 acquires a value of X% as the completion probability of a user behavior excluding an important behavior.

[0076] The calculation unit 28 calculates the difference between the completion probability of a user action including the important action acquired by the completion probability acquisition unit 27 and the completion probability of a user action not including the important action acquired by the completion probability acquisition unit 27. The calculation unit 28 calculates the difference for each important action acquired by the important action acquisition unit 25. In the example of FIG. 16 , the calculation unit 28 calculates a value of δ% as the difference.

[0077] The determination unit 29 determines whether or not the important action is effective based on the difference calculated by the calculation unit 28. In this embodiment, the determination unit 29 calculates the ratio of the number of users whose lifting effect is higher than 0% to the number of all users for each important action acquired by the important action acquisition unit 25. The determination unit 29 determines that the important action with the highest ratio is effective. The determination unit 29 determines that the important actions other than the important action with the highest ratio are ineffective. The determination unit 29 may, for example, determine that the important action is effective for two or more important actions in descending order of the ratio among the important actions.

[0078] In this embodiment, the learning device 2 acquires the important behavior determined to be effective by the determination unit 29 as the important behavior to be used in the intervention process. The intervention device 1 executes, as the intervention process, a process of executing the important behavior acquired by the learning device 2.

[0079] Next, the operation of the learning device 2 according to this embodiment will be described. FIG. 17 is a flowchart showing an example of the operation of the learning device 2 according to this embodiment. First, the learning data acquisition unit 21 acquires user actions, including a predetermined number of user actions up to a preset target action and the user's final action (step S21). In step S21, the learning data acquisition unit 21 refers to the user action database DB and acquires the type of user action, the content of the user action, and the step number for each user.

[0080] In step S21, major actions are identified from the acquired user actions by the learning data acquisition unit 21. In step S21, the learning data acquisition unit 21 identifies user actions that correspond to predetermined major actions.

[0081] In step S21, the learning data acquisition unit 21 determines a target action based on the step number acquired for each user. In this embodiment, in step S21, the learning data acquisition unit 21 further acquires one or more small actions performed between the target action and the next action as user actions. In step S21, the learning data acquisition unit 21 identifies, as small actions, user actions that do not correspond to major actions among the user actions acquired by referring to the user action database DB.

[0082] Next, the classification unit 22 classifies the user into either a first user group or a second user group (step S22). In step S22, the classification unit 22 determines whether the user's step number is the final step number based on the step number acquired for each user in step S21. In step S22, if the classification unit 22 determines that the user's step number is the final step number, the user is classified into the first user group. If the classification unit 22 determines that the user's step number is not the final step number, the user is classified into the second user group.

[0083] Next, the output model 23 outputs the degree of association between the one or more sub-actions acquired in step S21 and the next action (step S23). In step S23, the output model 23 uses a known means to output the degree of association between the sub-action and the next action.

[0084] Next, the extraction unit 24 extracts a predetermined number of small actions of the first user group and a predetermined number of small actions of the second user group based on the degrees of association output in step S23 (step S24). In this embodiment, in step S24, the extraction unit 24 extracts a predetermined number of small actions with the highest degree of association from among the small actions performed by users of the first user group. Also, in step S24, the extraction unit 24 extracts a predetermined number of small actions with the highest degree of association from among the small actions performed by users of the second user group.

[0085] Next, the important action acquisition unit 25 acquires an important action from one or more small actions acquired in step S21 based on the degree of association output in step S23 (step S25). In this embodiment, in step S25, the important action acquisition unit 25 acquires, as an important action, a small action that is among the predetermined number of small actions of the first user group extracted in step S24 but is not among the predetermined number of small actions of the second user group.

[0086] Next, the model generation unit 26 generates a completion probability estimation model by machine learning using the predetermined number of user actions up to the target action acquired in step S21 as explanatory variables and the fact that the important action has been reached as a target variable (step S26). The completion probability estimation model generated in step S26 estimates the user's completion probability using the user actions up to the target action as input.

[0087] Next, the completion probability acquisition unit 27 uses the completion probability estimation model generated in step S26 to acquire, for each user, a completion probability of reaching a final action of the user action not including the important action (step S27). In step S27, the completion probability acquisition unit 27 inputs the user action up to the target action into the completion probability estimation model as a user action not including the important action.

[0088] In step S27, the completion probability acquisition unit 27 uses the completion probability estimation model generated in step S26 to acquire, for each user, a completion probability of the user action including the important action reaching a final action. In step S27, the completion probability acquisition unit 27 inputs, to the completion probability estimation model, the user action to which the important action acquired in step S25 after the target action has been added, relative to the user action up to the target action. In step S27, the completion probability of the user action including the important action is acquired for each important action acquired in step S25.

[0089] Next, the calculation unit 28 calculates the difference between the completion probability of the user behavior including the important behavior acquired in step S27 and the completion probability of the user behavior not including the important behavior (step S28). In step S28, the calculation unit 28 calculates the difference for each important behavior acquired in step S25.

[0090] Next, the determination unit 29 determines whether or not the important actions are effective based on the difference calculated in step S28 (step S29). In step S29, the determination unit 29 calculates the ratio of the number of users whose promotion effect was greater than 0% to the total number of users for each important action acquired in step S25. In step S29, the determination unit 29 determines that the important action with the highest ratio is effective. In step S29, the determination unit 29 determines that the important actions other than the important action with the highest ratio are ineffective. After going through the above steps, the learning device 2 completes a series of operations.

[0091] Next, we will explain the effects of the intervention device 1 and the learning device 2 according to this embodiment. The intervention device 1 according to this embodiment includes a user behavior acquisition unit 11 that acquires user behavior, a completion probability estimation unit 12 that estimates the completion probability up to the final behavior based on the user behavior, and an intervention processing unit 13 that executes intervention processing based on the completion probability.

[0092] With this configuration, an intervention process is executed based on the completion probability estimated based on the user's behavior. If the user's completion probability is lower than a predetermined threshold as a result of estimating the completion probability based on the user's behavior, the intervention process can be executed to increase the user's completion probability. This can improve the user's completion probability.

[0093] In this embodiment, the completion probability estimation unit 12 is a completion probability estimation model that has been machine-learned, with user behavior up to the time of a preset target behavior as an explanatory variable and reaching the final behavior as a target variable. With this configuration, by using the machine-learned completion probability estimation model, it is possible to accurately estimate the user's completion probability when user behavior up to the time of the target behavior is input.

[0094] In this embodiment, the intervention processing unit 13 executes, as an intervention process, a process of causing an important action that has a strong correlation with a next action that is performed after a predetermined target action to be performed between the target action and the next action. With this configuration, by executing, as an intervention process, a process of causing an important action that has a strong correlation with the next action, it is possible to increase the probability that the user will perform the next action after performing the target action. This further improves the user's completion probability.

[0095] In this embodiment, the completion probability estimation unit 12 estimates the completion probability of user actions that do not include important actions, and the intervention processing unit 13 determines whether to execute intervention processing based on the completion probability of user actions that do not include important actions. With this configuration, it is possible to determine that intervention processing should be executed when the completion probability of user actions that do not include important actions is equal to or less than a predetermined value. Executing intervention processing can cause the user to perform important actions, thereby further improving the user's completion probability.

[0096] In this embodiment, the completion probability estimation unit 12 estimates the completion probability of user actions including important actions and the completion probability of user actions not including important actions, and calculates a lifting effect based on the respective completion probabilities. The intervention processing unit 13 determines whether to execute an intervention process based on the lifting effect. This configuration makes it possible to calculate the lifting effect, which is the degree to which the completion probability can be improved by executing the important action. If the calculated lifting effect is equal to or greater than a predetermined value, it can be determined that an intervention process should be executed. This makes it possible to execute an intervention process that is expected to have an effect equal to or greater than the predetermined value, thereby further improving the user's completion probability.

[0097] The learning device 2 according to this embodiment includes a learning data acquisition unit 21 that acquires user actions, including a predetermined number of user actions up to the time of a preset target action and the user's final action, and a model generation unit 26 that generates a completion probability estimation model by machine learning using the predetermined number of user actions up to the time of the target action as explanatory variables and reaching the final action as a target variable.

[0098] With this configuration, when a predetermined number of user actions up to the target action are input, a completion probability estimation model can be generated that can accurately estimate the user's completion probability.

[0099] In this embodiment, the learning data acquisition unit 21 further acquires, as user actions, one or more sub-actions that occur between the target action and a subsequent action that follows the target action. The learning device 2 includes an output model 23 that outputs the degree of association between one or more sub-actions and the subsequent action, and a significant action acquisition unit 25 that acquires an important action from the one or more sub-actions based on the degree of association. With this configuration, of the sub-actions that occur between the target action and the subsequent action, a sub-action that has a high degree of association with the subsequent action can be acquired as an important action.

[0100] In this embodiment, the learning device 2 further includes a classification unit 22 that classifies users into either a first user group including users who have reached the final action or a second user group including users who have not reached the final action, and an extraction unit 24 that extracts a predetermined number of small actions from the first user group and a predetermined number of small actions from the second user group based on the degree of association. The important action acquisition unit 25 acquires, as important actions, small actions that are among the predetermined number of small actions from the first user group but are not among the predetermined number of small actions from the second user group. This configuration makes it possible to extract small actions that are highly associated with the next action from the small actions of the first user group and the second user group. Then, it is concluded that small actions that are among the extracted predetermined number of small actions from the first user group but are not among the predetermined number of small actions from the second user group are small actions (important actions) necessary to execute the final action. Therefore, important actions can be acquired with greater accuracy.

[0101] In this embodiment, the learning device 2 further includes a completion probability acquisition unit 27 that acquires, for each user, a completion probability of reaching a final action for each of user actions including an important action and user actions not including an important action using a completion probability estimation model; a calculation unit 28 that calculates the difference between the completion probability of a user action including an important action and the completion probability of a user action not including an important action; and a determination unit 29 that determines whether or not the important action is effective based on the difference. With this configuration, it is possible to determine whether or not the important action is effective based on the difference between the completion probability of a user action including an important action and the completion probability of a user action not including an important action. As a result, it is possible to identify important actions that are effective in encouraging users to perform the final action.

[0102] Next, an intervention device 1 and a learning device 2 according to a modification will be described. Differences from the above embodiment will be mainly described below, and overlapping descriptions will be omitted as appropriate.

[0103] In the above embodiment, withdrawal is exemplified as an event that prevents a user from performing a final action. However, this event may also include a detour. A detour refers to, for example, the execution of a user action that is not necessary for the execution of the final action between one major action and the next major action that is executed after the first major action. For example, after a user accesses a shopping site S, browsing a large number of pages unrelated to the product during a series of user actions corresponds to the above detour.

[0104] In this modification, the classification unit 22 classifies users into either a first user group including users who have reached the next action with a small number of small actions, or a second user group including users who have reached the next action with a large number of small actions. The classification unit 22 determines whether the number of small actions performed between a preset target action and the next action is equal to or greater than a predetermined number, based on the user actions acquired by the learning data acquisition unit 21. The predetermined number is, for example, set in advance and can be changed as appropriate.

[0105] The classification unit 22 classifies the user into a first user group when it determines that the number of small actions performed between the target action and the next action is less than a predetermined number, and classifies the user into a second user group when it determines that the number of small actions performed between the target action and the next action is greater than or equal to a predetermined number.

[0106] In the above embodiment, an example has been described in which the intervention processing unit 13 determines whether to execute the intervention processing based on the completion probability of the user behavior not including the important behavior estimated by the completion probability estimation unit 12 and the lifting effect calculated by the completion probability estimation unit 12. However, the intervention processing unit 13 may determine whether to execute the intervention processing based only on the completion probability of the user behavior not including the important behavior estimated by the completion probability estimation unit 12. Furthermore, the intervention processing unit 13 may determine whether to execute the intervention processing based only on the lifting effect calculated by the completion probability estimation unit 12.

[0107] The intervention device and learning device of the present disclosure have the following configuration.

[0108] [1] An intervention device comprising: a user behavior acquisition unit that acquires user behavior; a completion probability estimation unit that estimates a completion probability up to a final behavior based on the user behavior; and an intervention processing unit that executes intervention processing based on the completion probability.

[0109] [2] The intervention device according to [1], wherein the completion probability estimation unit is a completion probability estimation model that is machine-learned using the user's behavior up to the time of a predetermined target behavior as an explanatory variable and the user's reaching the final behavior as a target variable.

[0110] [3] The intervention device according to [1] or [2], wherein the intervention processing unit executes, as the intervention processing, a process of performing an important action that has a strong relevance to a next action that is performed after a predetermined target action and that is performed between the target action and the next action.

[0111] [4] The intervention device according to [3], wherein the completion probability estimation unit estimates a completion probability of the user behavior that does not include the important behavior, and the intervention processing unit determines whether to execute the intervention processing based on the completion probability of the user behavior that does not include the important behavior.

[0112] [5] The intervention device described in [3] or [4], wherein the completion probability estimation unit estimates a completion probability of the user behavior including the important behavior and a completion probability of the user behavior not including the important behavior, and calculates a lifting effect based on the respective completion probabilities, and the intervention processing unit determines whether to execute the intervention processing based on the lifting effect.

[0113] [6] A learning device comprising: a learning data acquisition unit that acquires user actions including a predetermined number of user actions up to a predetermined target action and a final action of the user; and a model generation unit that generates a completion probability estimation model by machine learning using the predetermined number of user actions up to the target action as explanatory variables and reaching the final action as a target variable.

[0114] [7] The learning device described in [6], wherein the learning data acquisition unit further acquires, as the user action, one or more small actions that occur between the target action and a next action that occurs after the target action; an output model that outputs a degree of association between the one or more small actions and the next action; and an important action acquisition unit that acquires an important action from the one or more small actions based on the degree of association.

[0115] [8] The learning device described in [7], further comprising: a classification unit that classifies the users into either a first user group including the users who have reached the final action or have reached the next action with a small number of the small actions, or a second user group including the users who have not reached the final action or have reached the next action with a large number of the small actions; and an extraction unit that extracts a predetermined number of the small actions of the first user group and a predetermined number of the small actions of the second user group based on the degree of association, wherein the important action acquisition unit acquires, as the important action, a small action that is in the predetermined number of the small actions of the first user group but is not in the predetermined number of the small actions of the second user group.

[0116] [9] The learning device described in [8], further comprising: a completion probability acquisition unit that acquires, for each of the users, a completion probability of reaching the final action for each of the user actions including the important action and the user actions not including the important action using the completion probability estimation model; a calculation unit that calculates the difference between the completion probability of the user action including the important action and the completion probability of the user action not including the important action; and a determination unit that determines whether or not the important action is effective based on the difference.

[0117] The block diagrams used to explain the above embodiments show functional blocks. These functional blocks (components) are 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. The functional block may also be realized by combining the single device or multiple devices with software.

[0118] Functions include, but are not limited to, judgment, determination, discrimination, calculation, computation, processing, derivation, investigation, search, confirmation, reception, transmission, output, access, resolution, selection, selection, establishment, comparison, assumption, expectation, regard, broadcasting, notifying, communicating, forwarding, configuring, reconfiguring, allocating, mapping, and assignment. For example, a functional block (component) that performs transmission is called a transmitting unit or transmitter. As mentioned above, there are no particular limitations on how these functions are implemented.

[0119] 18 is a diagram showing an example of the hardware configuration of the intervention device 1 and the learning device 2 according to this embodiment. The intervention device 1 and the learning device 2 may be physically configured as a computer device including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, etc.

[0120] In the following description, the term "apparatus" can be interpreted as a circuit, a device, a unit, etc. The hardware configuration of the intervention apparatus 1 and the learning apparatus 2 may be configured to include one or more of the apparatuses shown in the drawings, or may be configured to exclude some of the apparatuses.

[0121] Each function of the intervention device 1 and the learning device 2 is realized by loading specified software (programs) onto hardware such as a processor 1001 and a memory 1002, causing the processor 1001 to perform calculations, control communication via the communication device 1004, and control at least one of reading and writing data in the memory 1002 and the storage 1003.

[0122] The processor 1001 controls the entire computer by running, for example, an operating system. The processor 1001 may be configured by a central processing unit (CPU) including an interface with peripheral devices, a control device, an arithmetic unit, a register, etc. For example, the above-mentioned user behavior acquisition unit 11 and the like may be realized by the processor 1001.

[0123] The processor 1001 also reads programs (program codes), software modules, data, etc. from at least one of the storage 1003 and the communication device 1004 into the memory 1002 and executes various processes in accordance with the programs. The programs used are those that cause a computer to execute at least some of the operations described in the above-described embodiments. For example, at least one of the functional units of the intervention device 1 and the learning device 2 may be implemented by a control program stored in the memory 1002 and running on the processor 1001, and similar implementations may be used for other functional blocks. While the above-described various processes have been described as being executed by one processor 1001, they may also be executed simultaneously or sequentially by two or more processors 1001. The processor 1001 may be implemented by one or more chips. The programs may also be transmitted from a network via a telecommunications line.

[0124] The memory 1002 is a computer-readable recording medium and may be configured, for example, by at least one of a read-only memory (ROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a random access memory (RAM), etc. The memory 1002 may also be referred to as a register, a cache, a main memory (primary storage device), etc. The memory 1002 can store executable programs (program codes), software modules, etc. for implementing the accompanying determination method according to one embodiment of the present disclosure.

[0125] Storage 1003 is a computer-readable recording medium, and may be composed of at least one of, for example, an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disk, a digital versatile disk, a Blu-ray (registered trademark) disk), a smart card, a flash memory (e.g., a card, a stick, a key drive), a floppy (registered trademark) disk, a magnetic strip, etc. Storage 1003 may also be referred to as an auxiliary storage device. The above-mentioned storage medium may be, for example, a database, a server, or other appropriate medium including at least one of memory 1002 and storage 1003.

[0126] The communication device 1004 is hardware (transmission / reception device) for communicating between computers via at least one of a wired network and a wireless network, and is also referred to as, for example, a network device, a network controller, a network card, or a communication module. The communication device 1004 may be configured to include a high-frequency switch, a duplexer, a filter, a frequency synthesizer, etc. to realize at least one of frequency division duplex (FDD) and time division duplex (TDD). For example, the above-mentioned user behavior acquisition unit 11 and the like may be realized by the communication device 1004. The communication device 1004 may be implemented with a transmitter and a receiver that are physically or logically separated.

[0127] The input device 1005 is an input device (e.g., a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that receives input from the outside. The output device 1006 is an output device (e.g., a display, a speaker, an LED lamp, etc.) that outputs to the outside. Note that the input device 1005 and the output device 1006 may be integrated into one device (e.g., a touch panel).

[0128] Furthermore, each device, such as the processor 1001 and the memory 1002, is connected by a bus 1007 for communicating information. The bus 1007 may be configured using a single bus, or may be configured using different buses between each device.

[0129] The intervention device 1 and the learning device 2 may be configured to include hardware such as a microprocessor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA), and some or all of the functional blocks may be realized by the hardware. For example, the processor 1001 may be implemented using at least one of these pieces of hardware.

[0130] The notification of information is not limited to the aspects / embodiments described in the present disclosure and may be performed using other methods. For example, the notification of information may be performed by physical layer signaling (e.g., Downlink Control Information (DCI) and Uplink Control Information (UCI)), higher layer signaling (e.g., Radio Resource Control (RRC) signaling, Medium Access Control (MAC) signaling, broadcast information (Master Information Block (MIB) and System Information Block (SIB))), other signals, or a combination thereof. Furthermore, the RRC signaling may be referred to as an RRC message, and may be, for example, an RRC Connection Setup message, an RRC Connection Reconfiguration message, or the like.

[0131] The order of the procedures, sequences, flowcharts, etc. of each aspect / embodiment described in this disclosure may be changed unless it is consistent. 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.

[0132] Input and output information may be stored in a specific location (for example, 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 sent to another device.

[0133] The determination may be made based on a value represented by one bit (0 or 1), a Boolean value (true or false), or a numerical comparison (e.g., comparison with a predetermined value).

[0134] The aspects / embodiments 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).

[0135] 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.

[0136] Software shall be construed broadly 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., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0137] 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.

[0138] The information, signals, etc. described in this disclosure 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.

[0139] Note that terms described in this disclosure and terms necessary for understanding this disclosure may be replaced with terms having the same or similar meanings. For example, at least one of a channel and a symbol may be a signal (signaling). Furthermore, a signal may be a message. Furthermore, a component carrier (CC) may be called a carrier frequency, a cell, a frequency carrier, etc.

[0140] Furthermore, the information, parameters, etc. described in the present disclosure may be expressed using absolute values, may be expressed using relative values ​​from a predetermined value, or may be expressed using other corresponding information. For example, a radio resource may be indicated by an index.

[0141] The names used for the above-described parameters are not intended to be limiting in any way. Furthermore, the mathematical expressions using these parameters may differ from those explicitly disclosed in this disclosure. The various channels (e.g., PUCCH, PDCCH, etc.) and information elements may be identified by any suitable names, and therefore the various names assigned to these various channels and information elements are not intended to be limiting in any way.

[0142] In this disclosure, the terms "Mobile Station (MS)," "user terminal," "User Equipment (UE)," "terminal," and the like may be used interchangeably.

[0143] 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 suitable terminology.

[0144] 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 in a table, database, or other data structure), ascertaining, and the like. "Determining" and "determining" may also include receiving (e.g., receiving information), transmitting (e.g., sending information), input, output, accessing (e.g., accessing data in memory), and the like. 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.

[0145] The terms "connected," "coupled," or any variation 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 physical, logical, 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 one or more wires, cables, and / or 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.

[0146] As used in this disclosure, the phrase "based on" does not mean "based only on," unless expressly stated otherwise. In other words, the phrase "based on" means both "based only on" and "based at least on."

[0147] As used in this disclosure, any reference to an element using a designation such as "first," "second," etc. does not generally limit the quantity or order of those elements. These designations may be used in this disclosure as a convenient method of distinguishing between two or more elements. Thus, a reference to a first and a second element does not imply that only two elements may be employed or that the first element must in some way precede the second element.

[0148] When the terms "include," "including," and variations thereof are used in this disclosure, these terms are intended to be inclusive, similar to the term "comprising." Furthermore, when the term "or" is used in this disclosure, it is not intended to be an exclusive or.

[0149] 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 in the plural form.

[0150] In the present 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 "coupled" may also be interpreted in the same way as "different."

[0151] 1...intervention device, 2...learning device, 11...user behavior acquisition unit, 12...completion probability estimation unit, 13...intervention processing unit, 21...learning data acquisition unit, 22...classification unit, 23...output model, 24...extraction unit, 25...important behavior acquisition unit, 26...model generation unit, 27...completion probability acquisition unit, 28...calculation unit, 29...judgment unit, A, B, C...small actions.

Claims

1. An intervention device comprising: a user behavior acquisition unit that acquires user behavior; a completion probability estimation unit that estimates a completion probability up to a final action based on the user behavior; and an intervention processing unit that executes intervention processing based on the completion probability.

2. The intervention device according to claim 1, wherein the completion probability estimation unit is a completion probability estimation model trained by machine learning using the user's actions up to the time of a preset target action as an explanatory variable and reaching the final action as an objective variable.

3. The intervention device according to claim 1, wherein the intervention processing unit executes, as the intervention processing, a process of performing an important action that is performed between a predetermined target action and a next action and that has a strong relevance to the next action that is performed after the target action.

4. The intervention device according to claim 3, wherein the completion probability estimation unit estimates a completion probability of the user behavior that does not include the important behavior, and the intervention processing unit determines whether or not to execute the intervention processing based on the completion probability of the user behavior that does not include the important behavior.

5. The intervention device described in claim 3, wherein the completion probability estimation unit estimates the completion probability of the user behavior including the important behavior and the completion probability of the user behavior not including the important behavior, and calculates a lifting effect based on the respective completion probabilities, and the intervention processing unit determines whether or not to execute the intervention processing based on the lifting effect.

6. A learning device comprising: a learning data acquisition unit that acquires user actions including a predetermined number of user actions up to a preset target action and a final action of the user; and a model generation unit that generates a completion probability estimation model through machine learning using the predetermined number of user actions up to the target action as explanatory variables and reaching the final action as an objective variable.

7. The learning device described in claim 6, further comprising: an output model that further acquires, as the user action, one or more small actions that are performed between the target action and a next action that is performed after the target action; and an important action acquisition unit that further acquires an important action from the one or more small actions based on the degree of association.

8. The learning device described in claim 7, further comprising: a classification unit that classifies the users into either a first user group including the users who have reached the final action or who have reached the next action with a small number of the small actions, and a second user group including the users who have not reached the final action or who have reached the next action with a large number of the small actions; and an extraction unit that extracts a predetermined number of the small actions of the first user group and a predetermined number of the small actions of the second user group based on the degree of association, wherein the important action acquisition unit acquires, as the important action, a small action that is included in the predetermined number of the small actions of the first user group but is not included in the predetermined number of the small actions of the second user group.

9. The learning device described in claim 8, further comprising: a completion probability acquisition unit that acquires, for each user, a completion probability of reaching the final action for each of the user actions including the important action and the user actions not including the important action using the completion probability estimation model; a calculation unit that calculates the difference between the completion probability of the user action including the important action and the completion probability of the user action not including the important action; and a judgment unit that judges whether or not the important action is effective based on the difference.

Images