JP7419313B2 - Information processing device, information processing method, and information processing program - Google Patents

Description

The present invention relates to an information processing device, an information processing method, and an information processing program.

Conventionally, methods for predicting user behavior on the Internet have been known. For example, in Patent Document 1, with the aim of predicting user behavior with higher accuracy than before, the tendency of search queries of users on the Internet is learned, and the learned tendency of search queries and subsequent user behavior are disclosed. A technique has been proposed in which a model is generated by calculating a relevance score with a user, and this model is used to predict the behavior of other users.

Japanese Patent Application Publication No. 2016-177377

However, with the above-mentioned techniques, it is difficult to obtain certainty as to whether or not the generated information can be trusted. For example, there is a possibility that a model that produces effective results may be included, and even if an AB test is conducted, an inappropriate model may be adopted depending on the content of the test.

The present application has been made in view of the above, and aims to provide an information processing device, an information processing method, and an information processing program that can confirm the reliability of generated information.

The information processing device according to the present application includes a first estimation section, a second estimation section, and an evaluation section. The first estimating unit estimates the probability that a related predetermined action will be performed among a plurality of estimation models prepared in advance for each predetermined action in order to estimate the probability that a predetermined action will be performed by the user. A state transition of a predetermined behavior is estimated using a plurality of first models for the first model. The second estimation unit uses a second model for estimating the probability of performing a predetermined action that has a negative relationship with the state transition of the predetermined action from among the plurality of estimation models. Estimate the probability that a predetermined action with relevance will be performed. The evaluation section evaluates the reliability of the second model based on the accuracy estimated by the second estimation section.

According to one aspect of the embodiment, the reliability of the generated information can be verified.

FIG. 1 is a diagram illustrating an example of an information processing system according to an embodiment. FIG. 2 is a diagram illustrating an example of a user classification method and a second list information evaluation method according to the embodiment. FIG. 3 is a diagram illustrating a configuration example of an information processing device according to an embodiment. FIG. 4 is a diagram showing an overview of first list information according to the embodiment. FIG. 5 is a diagram illustrating an overview of second list information according to the embodiment. FIG. 6 is a flowchart illustrating an example of a processing procedure by the information processing apparatus according to the embodiment. FIG. 7 is a hardware configuration diagram showing an example of a computer that implements the functions of the information processing device according to the embodiment.

DESCRIPTION OF THE PREFERRED EMBODIMENTS An information processing apparatus, an information processing method, and an embodiment of an information processing program (hereinafter referred to as "embodiments") according to the present application will be described in detail below with reference to the drawings. Note that the information processing apparatus, information processing method, and information processing program according to the present application are not limited to the embodiments described below. Furthermore, the embodiments described below can be combined as appropriate within the range that does not conflict with the processing contents. Further, in the embodiments described below, the same parts are given the same reference numerals, and redundant explanations will be omitted.

Furthermore, in the following description, the terminal device 10 may be described as a seed user U1 or a user U2. That is, the seed user U1 or the user U2 can be read as the terminal device 10.

[1. Information processing system]
FIG. 1 is a diagram illustrating an example of an information processing system according to an embodiment. As shown in FIG. 1, the information processing system SYS according to the embodiment includes a terminal device 10 and an information processing device 100. Note that the information processing system SYS is not limited to the example shown in FIG. 1, and may include multiple terminal devices 10 and multiple information processing devices 100.

The terminal device 10 and the information processing device 100 are each connected to a network N (see FIG. 3, for example) by wire or wirelessly. The network N is a communication network such as a LAN (Local Area Network), a WAN (Wide Area Network), a telephone network (mobile phone network, fixed telephone network, etc.), a local IP (Internet Protocol) network, or the Internet. . The network N may include a wired network or a wireless network. The terminal device 10 and the information processing device 100 can communicate with each other through the network N.

The terminal device 10 shown in FIG. 1 is an information processing device used by a user U who uses various services provided by the information processing device 100 and the like. For example, the terminal device 10 is realized by a smartphone, a tablet terminal, a notebook PC (Personal Computer), a desktop PC, a PDA (Personal Digital Assistant), or the like.

The terminal device 10 also supports wireless communication networks such as LTE (Long Term Evolution), 4G (4th Generation), and 5G (5th Generation), Bluetooth (registered trademark), and wireless LAN (Local Area It is possible to connect to the network N via short-range wireless communication such as Network).

Further, the terminal device 10 can display information for using various services using a web browser or an application. Note that when the terminal device 10 receives control information for realizing information display processing from the information processing device 100 or the like, it realizes the display processing according to the control information.

Furthermore, the terminal device 10 provides user information to the information processing device 100 through the use of various services. User information includes search history including search queries entered by each user U, purchase history of each user U, history of transaction objects viewed on the online shopping mall, purchase history of transaction objects purchased, etc. , viewing history, which is the history of content such as news that has been viewed, posting history, which is the history of text, images, etc. posted to SNS (Social Networking Service), and location history. In addition, user information includes, for example, the usage history of payment services realized by reading QR codes (registered trademarks) installed in stores using smartphones owned by each user U, and the usage history of payment services realized using the reading results. It may also include a history of usage of payment services (for example, purchase date and time, purchased products, purchase amount, etc.) that is performed by displaying a QR code on an owned smartphone and reading the QR code with a store terminal. Further, the user information may include the usage history of a credit card, the usage history of a bank account, etc. held by each user U.

Further, the user information may include information estimated based on collected information. For example, user information may be based on purchase history if the average monthly grocery purchase total is 100,000 yen or more, household monthly income or personal monthly income is 300,000 yen or more, or a family of three people. It may also include information estimating the family composition, the composition of cohabitants, etc. Further, the user information may include, for example, information that is estimated by estimating the items owned by each user U, the place of residence or workplace of each user U, etc. from the location history and purchase history.

Further, the user information may include, for example, information obtained by acquiring a service usage history from a server device other than the information processing device 100 that provides various services. Further, the user information may include various information estimated from a service usage history by a server device or a single server device that provides such a service. Further, the user information may include information indicating psychographic attributes such as various geographical attributes and hobbies and preferences registered by each user U or estimated for each user U.

Further, the user information may include sensor information acquired by various sensors installed in the terminal device 10. For example, sensor information includes position information acquired by a position sensor, acceleration information detected by an acceleration sensor, angular velocity information detected by a gyro sensor, operation information acquired by a touch panel, and information acquired by an illuminance sensor. The information may include illuminance information obtained by a barometric pressure sensor, atmospheric pressure information obtained by a barometric pressure sensor, sound information obtained by a microphone, and the like.

The information processing apparatus 100 shown in FIG. 1 is an information processing apparatus that provides various services to each user U who is a service user. The various services provided by the information processing device 100 may include API (Application Programming Interface) services corresponding to various applications and various online services. Online services include Internet connectivity, search services, SNS (Social Networking Services), electronic commerce services, electronic payment services, online games, online banking services, online trading services, accommodation reservation services, etc. Ticket reservation services, video distribution services, music distribution services, news distribution services, map information services, route search services, route guidance services, route information services, operation information services, weather information services, etc. Services may be applicable.

Further, the information processing apparatus 100 is an information processing apparatus that executes information processing according to the embodiment using user information collected through provision of various services. The information processing device 100 is typically a server device, but may also be realized by a mainframe, a workstation, or the like.

When the information processing device 100 is implemented as a server device, it may be implemented as a single server, or may be implemented as a cloud system in which multiple server devices and multiple storage devices operate cooperatively. Note that the information processing device 100 may function as a distribution device that distributes control information to the terminal device 10 used by each user U. Here, the control information is written in, for example, a script language such as JavaScript (registered trademark) or a style sheet language such as CSS (Cascading Style Sheets). Note that the application itself distributed from the information processing device 100 may be regarded as control information.

[2. Information processing]
An example of information processing in the information processing system SYS will be described below. Note that in the following description, the targets of information processing according to the embodiment are each user U who uses various services. Furthermore, in the following description, each user U who uses various services may be simply referred to as "user U".

First, in the example shown in FIG. 1, the information processing device 100 stores user information collected from the user U. Then, the information processing apparatus 100 selects a target user to be subjected to information processing from among the user information, either at random or according to predetermined conditions.

In addition, the information processing device 100 estimates the probability that a related action will be performed among a plurality of estimation models prepared in advance for each predetermined action in order to estimate the probability that a predetermined action will be performed by the user U. Select the first model for

For example, the information processing device 100 can select a plurality of first models from among a plurality of estimated models based on the feature amount of a predetermined behavior that is associated with the estimated model. In the example shown in FIG. 1, the behavior A estimation model and the behavior B estimation model are selected as the first models. More specifically, the information processing device 100 uses a natural language processing method such as Word2Vec to find the feature amount (vector expression) of a character string indicating a predetermined behavior, and based on the similarity of the found feature amount, , estimated models whose features are similar to each other can be selected as the first models corresponding to each of the related actions. Further, the information processing device 100 assumes that the state transition of the target user's behavior follows a simple Markov process, and calculates the conditional probabilities of the next behavior from the current behavior in the state transition of the target user's behavior. , an action for which the calculated conditional probability is greater than or equal to a predefined threshold may be identified as an action that is related to the current action.

An example of a method for creating the above-mentioned estimation model will be described below. For example, the information processing device 100 collects users who meet the conditions (titles) such as "bought ____" and "interested in ____" as seed users, and collects them based on the commonality of search queries input by the seed users. Create a score list with assigned scores. The information processing device 100 uses this score list as an estimation model. Specifically, we collected seed users who say ``wear glasses,'' search query: ``glasses,'' ``70 points,'' search query: ``eyeglass cleaning,'' ``50 points,'' and search query: ``apple.'' A score such as "-50 points" is assigned to the search query, and a score list that is a combination of the search query and the score is created.

The seed user described above can be identified using any method. For example, the information processing apparatus 100 may estimate a user who has performed an assumed action based on various types of information such as various purchase histories, usage histories, and position histories, and may identify the user as a seed user. Specifically, the seed user may be identified by inputting the name of a certain predetermined store as a search query and then estimating the number of visitors who actually visited the predetermined store. As another example, seed users may be identified by inputting the name of an event as a search query and then estimating the participants who actually participated in the event. As yet another example, a seed user may be identified by inputting the name of a certain product as a search query and then estimating the purchaser who actually purchased the product with that name. Further, the information processing device 100 may identify the seed user by estimating the number of users who actually used a predetermined store (for example, a store corresponding to a distributed advertisement). Furthermore, the information processing apparatus 100 may identify seed users by estimating the number of viewers who have viewed content such as a predetermined advertisement, web page, movie, or music.

Further, the information processing apparatus 100 can obtain the score of the search query in the above-mentioned score list from a behavior model generated by machine learning as described below. For example, when a search query input by a seed user is taken as a positive example, and a search query input by a randomly selected user other than the seed user is taken as a negative example, when the information processing apparatus 100 inputs a positive example, The learning model is trained to output a higher value and output a lower value when a negative example is input. At this time, any method can be adopted for learning. For example, if a neural network is used as the model, it can be realized by backpropagation or the like.

Furthermore, the information processing apparatus 100 may learn the model so that higher numerical values are output for search queries that are commonly input by seed users. For example, the model may be trained to output a higher value as the number of seed users inputting positive example search queries increases. Furthermore, the model may be trained to output a smaller value (negative value) as the number of other users who input negative example search queries increases. Then, the information processing device 100 generates a list of pairs of each search query and a score output when the search query is input to such a learning model as a score list. That is, the information processing device 100 calculates each search query input by a seed user who has performed a predetermined action, and the above-mentioned score (score indicating the relationship between the search query and the predetermined action) corresponding to each search query. ) is generated as a score list.

Note that the information processing device 100 can determine the probability that a predetermined action will be performed by using the score list described above. For example, the information processing device 100 obtains a score associated with each of a plurality of search queries input by a user whose behavior is to be estimated from a score list, and calculates a total value of the obtained scores. Then, when the total value of the calculated scores exceeds a predetermined threshold, the information processing device 100 determines that there is a high probability that the predetermined action corresponding to the score list will be performed.

Further, the information processing apparatus 100 may use information estimated based on user information of a seed user who has performed a predetermined action as the feature amount of the predetermined action. For example, the information processing device 100 quantifies the relationship between a predetermined action and each user based on a score list and a search query input by each user including a seed user who has performed a predetermined action. A user list is generated in which each user is ranked based on the value indicating the relationship.

For example, the information processing apparatus 100 selects each user including a seed user, and specifies the score associated with the search query input by each selected user from the score list. Note that the information processing apparatus 100 may select each user at random or may select each user based on predetermined conditions. Then, the information processing apparatus 100 derives a user list indicating the relationship between each user and a predetermined behavior set as a condition for collecting seed users from the total of the identified scores. For example, in the user list, users who have a high relationship with a predetermined behavior have a high score, and users who have a low relationship with a predetermined behavior have a low score.

An example of a method for creating the above-mentioned user list will be described below. For example, the information processing device 100 obtains the scores associated with each of the plurality of search queries input by each user including the seed user from the score list, and calculates the total value of the obtained scores. In this way, the relationship between multiple search queries entered by each user and a predetermined behavior can be expressed numerically, which is more reliable than quantifying a single search query. can be realized.

Next, the information processing device 100 calculates the total score for each user, ranks each user in descending order of score, and generates a user list in which search queries entered by each user are associated in order of ranking. do. This user list functions as a list indicating which search queries the users who have input are strongly related to the predetermined behavior. This user list is used to analyze the relationship between a search query and a predetermined behavior, and to analyze users. This user list includes predetermined actions such as actions set when selecting seed users (for example, "bought ____", "went to ____", "doing ____", etc.) created for each

Then, the information processing apparatus 100 extracts each high-ranked user from the above-mentioned user list, and obtains user information associated with the extracted user. Then, using the acquired user information, the feature amounts of each user can be determined, and the feature amounts corresponding to a predetermined behavior can be derived based on each of the determined feature amounts.

For example, the information processing device 100 divides the sensor information of each user extracted from the user list into predetermined periods, and employs data in which characteristic values in each divided period are arranged in chronological order as the feature quantity. can. The value of the feature amount may be the maximum value detected in a predetermined period, the minimum value detected in a predetermined period, or the average value detected in a predetermined period. It may be a value, or it may be an increase rate or a decrease rate of a function that indicates a time-series change in values detected in a predetermined period, or it may be a characteristic value obtained by Fourier transforming this function. It may be a value obtained by differentiating this function with respect to time, or it may be a value obtained by another algorithm. Then, the information processing device 100 can identify a tendency common to the feature amounts of each user extracted from the user list, and employ the identified tendency as a feature amount corresponding to a predetermined action.

In addition, the information processing device 100 uses a second model for estimating the probability that an action having a negative relationship with the state transition of the target user will be performed among the plurality of estimation models. Estimate the probability that an action with relevance will be performed. For example, the information processing device 100 selects, as the second model, an estimation model that estimates the probability that an action that is logically contradictory to the state transition of the target user will be performed. In the example shown in FIG. 1, the behavior C estimation model is selected as the second model.

For example, the operator of the information processing device 100 registers in advance, as a rule base, actions that are logically contradictory to the state transition of the target user. Then, the information processing device 100 identifies an action that is logically contradictory to the state transition of the target user, and selects a second model that estimates the probability that the selected action will be performed, according to a pre-registered rule base.

Alternatively, the information processing device 100 may perform an action that is logically contradictory to the state transition of the target user by specifying an inappropriate behavior as the next behavior from the current behavior in the state transition of the target user's behavior. An estimation model for estimating the accuracy of the calculation may be selected as the second model. For example, the information processing device 100 assumes that the state transition of the target user's behavior follows a simple Markov process, calculates the conditional probability of the next behavior from the current behavior in the state transition of the target user's behavior, An action for which the calculated conditional probability is less than a predefined threshold is identified as an action that is logically contradictory to the state transition of the target user. The information processing device 100 then selects an estimation model that estimates the probability that the specified action will be performed as the second model. For example, when the current action in the state transition of the target user's actions is "purchasing an engagement ring," the information processing device 100 determines that the conditional probability of "submitting divorce notification" as the next action to occur is determined in advance. may be below the threshold. Further, when the current action in the state transition of the target user's actions is "reserving a wedding hall", the conditional probability of "submitting divorce notification" as the next action may be less than a predefined threshold. At this time, if it is estimated that the target user's actions follow the state transition from "purchasing an engagement ring" to "reserving a wedding venue," the information processing device 100 determines the probability that "submitting divorce notification" will be performed. What is necessary is to select an estimation model for estimating as the second model.

Furthermore, the information processing device 100 evaluates the reliability of the second model based on the estimated accuracy, that is, the likelihood that the target user will perform an action that has a negative association. For example, the information processing device 100 estimates the probability that an action having a negative association will be performed during the period in which the state transition is estimated, and evaluates the reliability of the second model based on the change in the estimated probability. evaluate.

Hereinafter, the procedure of information processing by the information processing apparatus 100 will be organized and explained with reference to FIG. As shown in FIG. 1, the information processing device 100 estimates the state transition from behavior A to behavior B of the target user using the behavior A estimation model and the behavior B estimation model selected from the estimation models. (Step S1).

Next, the information processing device 100 uses the behavior C estimation model selected from the estimation models to estimate the probability that the behavior C that has a negative relationship with the target user's state transition will be performed (step S2).

Then, the information processing device 100 evaluates the reliability of the behavior C estimation model based on the probability that behavior C will be performed (step S3). Hereinafter, an example of a method for evaluating the behavior C estimation model will be described using FIG. 2. FIG. 2 is a diagram illustrating an overview of the estimation model evaluation method according to the embodiment.

In Figure 2, as a specific example of a predetermined action corresponding to the estimation model, action A is "purchasing an engagement ring," action B is "reserving a wedding hall," and action C is "submitting divorce notification." Showing. Further, FIG. 2 shows a case where user Ua is selected as a target user to be subjected to information processing according to the embodiment.

First, the information processing device 100 uses the behavior A estimation model and the user information (past log) of the user Ua to obtain a score indicating the probability that the user Ua will perform the behavior A. For example, the information processing device 100 acquires information on a search query that the user Ua inputted in July, and inputs the acquired search query information into the behavior A estimation model, so that the user Ua performs the behavior A in July. Obtain a score that indicates how accurately you performed the test.

Next, the information processing device 100 obtains a score indicating the probability that the user Ua will perform the action B using the action B estimation model that estimates the action B that is related to the action A. For example, the information processing device 100 acquires information on a search query that user Ua inputted in July, and inputs the acquired search query information into the behavior B estimation model, so that user Ua can perform behavior B in August. Obtain a score that indicates how accurately you performed the test.

As shown in FIG. 2, the information processing device 100 determines that the score of the user Ua obtained from the behavior A estimation model is equal to or higher than a predefined threshold A, and the score of the user Ua obtained from the behavior B estimation model is If it is equal to or greater than the prescribed threshold B, it is determined that there is a possibility that the user Ua's behavior has made a state transition from behavior A to behavior B from July to August.

Next, the information processing device 100 calculates the probability that the user Ua will perform the behavior C using the behavior C estimation model that has a negative relationship with the state transition from behavior A to behavior B and the user information of the user Ua. Find the score shown. For example, the information processing device 100 acquires information on a search query that the user Ua inputted in July, and inputs the acquired search query information into the behavior C estimation model, so that the user Ua can perform the behavior C in July. Obtain a score that indicates how accurately you performed the test. Similarly, the information processing device 100 acquires information on a search query that the user Ua inputted in August, and inputs the acquired search query information into the behavior C estimation model. Obtain a score that indicates the accuracy with which the

As shown in FIG. 2, the information processing device 100 detects that the user Ua's score in July and August, which was obtained from the behavior C estimation model, has been on an upward trend from "50" to "70". It is determined that the possibility that Ua will perform Action C from July to August has increased. In this case, the information processing device 100 changes the state of the behavior of the user Ua, that is, behavior A: "purchasing an engagement ring" → behavior B: "reserving a wedding venue" and behavior C: "filing a divorce" which is logically contradictory. The reliability of the behavior C estimation model that derived the estimation result that the possibility of "submission of" is high is evaluated as low.

Note that when the information processing apparatus 100 determines that the reliability of the behavior C estimation model is low, the information processing apparatus 100 may modify the behavior C estimation model using the behavior A estimation model and the behavior B estimation model. For example, the information processing device 100 adjusts the score of the search query included in the behavior A estimation model or the behavior B estimation model to be lower among the scores associated with the search query in the behavior C estimation model. , the behavior C estimation model can be modified. Alternatively, the information processing device 100 may modify the behavior C estimation model by excluding the search queries included in the behavior A estimation model or the behavior B estimation model from the search queries included in the behavior C estimation model.

Note that the information processing device 100 can provide the operator of the information processing device 100 with information regarding the evaluation of the reliability of the second model. Further, when the second model is evaluated to have low reliability, the information processing apparatus 100 may provide information regarding the evaluation of the reliability of the second model as well as information regarding a method for modifying the second model.

[3. Configuration of information processing device]
The configuration of the information processing device 100 according to the embodiment will be described using FIG. 3. FIG. 3 is a diagram illustrating a configuration example of an information processing device according to an embodiment.

As shown in FIG. 3, the information processing device 100 includes a communication section 110, a storage section 120, and a control section 130. Note that FIG. 3 shows an example of the configuration of the information processing apparatus 100, and the information processing apparatus 100 is not limited to the configuration shown in FIG. 3, and may have a configuration including other functional units other than those shown in FIG.

(Communication Department 110)
The communication unit 110 is connected to the network N by wire or wirelessly, for example, and transmits and receives information to and from other devices such as the terminal device 10 via the network N. The communication unit 110 is realized by, for example, a NIC (Network Interface Card), an antenna, or the like. The network N is a communication network such as a LAN (Local Area Network), a WAN (Wide Area Network), a telephone network (such as a mobile phone network or a fixed telephone network), a local IP (Internet Protocol) network, or the Internet. . The network N may include a wired network or a wireless network.

The communication unit 110 receives information such as captured images from the terminal device 10. Furthermore, the communication unit 110 transmits proposal information to the terminal device 10.

(Storage unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 120 stores programs and data for implementing information processing executed by each unit of the control unit 130.

As shown in FIG. 3, the storage unit 120 includes a user information storage unit 121 and an estimated model storage unit 122.

(User information storage unit 121)
The user information storage unit 121 stores user information provided by each user in an arbitrary format. FIG. 4 is a diagram showing an overview of user information according to the embodiment.

As shown in FIG. 4, the user information includes items such as "user ID", "search history", and "purchase history". These items included in the estimated model information are associated with each other.

The "User ID" item stores identification information that is individually assigned to each user by the information processing apparatus 100 in order to identify users who use various services when providing various services.

The "search history" item stores information regarding search history provided by each user as user information. Further, in the "Purchase History" item, information regarding purchase history provided by each user as user information is stored.

(Estimated model storage unit 122)
The estimated model storage unit 122 stores estimated model information regarding estimated models in an arbitrary format. FIG. 5 is a diagram showing an overview of estimated model information according to the embodiment.

As shown in FIG. 5, the estimated model information includes a "model ID" item, a "behavior" item, and a "data" item. These items included in the estimated model information are associated with each other.

The "Model ID" item stores identification information that is individually assigned to each estimated model in order to identify the estimated model.

"Behavior" stores information indicating the content of the behavior estimated by the associated estimation model.

The "data" item stores data of the associated estimation model. Although Fig. 4 shows an example in which conceptual information such as "Data A" and "Data B" is stored in the "Data" item, in reality, the configuration of the estimation model (network This includes various information constituting the corresponding estimation model, such as information on the configuration (configuration) and information on parameters. For example, the individual information that makes up the estimation model includes nodes in each layer of the network, functions adopted by each node, connection relationships between nodes, and connection coefficients set for connections between nodes. may be included.

(Control unit 130)
The control unit 130 shown in FIG. 3 is a controller that controls the information processing device 100. The control unit 130 uses a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), etc. to execute various programs (for example, information processing programs) stored in a storage device inside the information processing device 100. This is realized by executing using RAM as a work area. Further, the control unit 130 may be realized by, for example, an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

As shown in FIG. 3, the control unit 130 includes a first estimation unit 131, a second estimation unit 132, and an evaluation unit 133. The control unit 130 uses these units to realize or execute information processing functions and operations described below. Note that the internal configuration of the control unit 130 is not limited to the configuration shown in FIG. 3, and may be any other configuration as long as it performs information processing to be described later. Further, the connection relationship between the respective units included in the control unit 130 is not limited to the connection relationship shown in FIG. 3, and may be other connection relationships.

(First estimation unit 131)
The first estimating unit 131 prepares in advance for each predetermined action in order to estimate the probability that the predetermined action will be performed by the target user of the information processing performed by the information processing device 100 (for example, see FIGS. 1 and 2). A state transition of the behavior is estimated using a plurality of first models for estimating the probability that a related behavior will be performed among the plurality of estimated models.

For example, the first estimating unit 131 selects a plurality of first models from among the plurality of estimation models based on the feature amount of a predetermined behavior associated with the estimation model. For example, the first estimating unit 131 uses a natural language processing method such as Word2Vec to obtain a feature amount (vector expression) of a character string indicating a predetermined behavior, and based on the similarity of the obtained feature amount, a plurality of A first model corresponding to each of the related behaviors can be selected from the estimated models.

(Second estimation unit 132)
The second estimation unit 132 uses a second model among the plurality of estimation models for estimating the probability that an action having a negative relationship with the state transition of the action of the target user will be performed. Estimate the probability that an action with a negative association will be performed.

For example, the second estimation unit 132 selects, as the second model, an estimation model that estimates the probability that an action that is logically contradictory to the state transition of the target user's action will be performed. Then, the second estimation unit 132 uses the selected second model to estimate the probability that the target user will perform an action that has a negative association.

For example, the second estimation unit 132 identifies an action that is logically contradictory to the state transition of the target user according to a pre-registered rule base, and determines whether the selected action is logically contradictory to the state transition of the target user. A second model can be selected that estimates the accuracy with which it is performed.

Alternatively, the second estimation unit 132 assumes that the state transition of the target user's behavior follows a simple Markov process, and calculates the conditional probability of the next behavior from the current behavior in the state transition of the target user's behavior. Then, actions for which the calculated conditional probability is less than a predefined threshold are identified as actions that are logically contradictory to the state transition of the target user. Then, the second estimation unit 132 selects an estimation model that estimates the probability that the specified action will be performed as the second model.

Further, the second estimating unit 132 estimates the probability that an action having a negative association will be performed during the period in which the state transition of the target user is estimated. For example, the second estimation unit 132 estimates the probability that an action having a negative relationship with the state transition of the target user will be performed for the same period as the period in which the state transition of the target user is estimated. The second estimation unit 132 then passes the estimated accuracy to the evaluation unit 133.

(Evaluation unit 133)
The evaluation unit 133 evaluates the reliability of the second model based on the accuracy estimated by the second estimation unit 132.

For example, the evaluation unit 133 evaluates the reliability of the second model based on the change in accuracy estimated by the second estimation unit 132. Specifically, the evaluation unit 133 evaluates the reliability of the second model if the accuracy estimated by the second estimation unit 132 is on the rise in the same period as the period in which the state transition of the target user is estimated. The quality is evaluated as low.

[4. Processing procedure]
Hereinafter, a processing procedure by the information processing apparatus 100 according to the embodiment will be described using FIG. 6. FIG. 6 is a flowchart illustrating an example of a processing procedure by the information processing apparatus according to the embodiment. The processing procedure shown in FIG. 6 is executed by the control unit 130 of the information processing device 100. The processing procedure shown in FIG. 6 is repeatedly executed while the information processing device 100 is in operation.

As shown in FIG. 6, the first estimation unit 131 estimates the state transition from behavior A to behavior B of the target user using the behavior A estimation model and the behavior B estimation model selected from the estimation models. (Step S101).

Next, the second estimating unit 132 uses the behavior C estimation model selected from the estimation models to estimate the probability that the behavior C that has a negative relationship with the state transition of the target user will be performed (step S102). .

Subsequently, the evaluation unit 133 evaluates the reliability of the behavior C estimation model based on the probability that behavior C will be performed (step S103), and ends the processing procedure shown in FIG. 6.

[5. Modified example]
The information processing device 100 described above may be implemented in various different forms other than the embodiments described above. Therefore, a modification of the embodiment will be described below.

(5-1. Regarding time series of state transition)
In the embodiment described above, the information processing device 100 may consider the time series of state transitions when estimating state transitions of the target user's behavior. For example, when selecting a plurality of related first models from a plurality of estimation models, if it is natural for the actions to be performed in the order of action X → action Y, the information processing apparatus 100 The score of the first model may be evaluated in the order of , and the state transition may be estimated.

Specifically, the information processing apparatus 100 determines whether the score of behavior The target user's behavior changes from behavior X to behavior Y on the condition that the score of behavior It is assumed that the transition is occurring.

(5-2. Regarding the use of state transition estimation results)
In the above-described embodiment, the information processing apparatus 100 may perform, for example, distributing an advertisement to the target user using the estimation result of the state transition of the target user's behavior. At this time, the information processing device 100 may perform control so as not to distribute advertisements that are inappropriate for the event estimated from the state transition of the target user's behavior. For example, if the information processing device 100 estimates that the target user is likely to be married based on the state transition of the target user's behavior, the information processing device 100 may ask the target user an inappropriate message such as "Would you like to get married?" You can control the delivery of inappropriate advertisements. Further, for example, when the information processing device 100 estimates that there is a high possibility that the target user has purchased an expensive product, the information processing device 100 controls not to distribute an advertisement suggesting the purchase of a similar product.

[6. Hardware configuration]
The information processing apparatus 100 according to the embodiment is realized, for example, by a computer 1000 having a configuration as shown in FIG. FIG. 7 is a hardware configuration diagram showing an example of a computer that implements the functions of the information processing device according to the embodiment.

The computer 1000 is connected to an output device 1010 and an input device 1020, and has an arithmetic device 1030, a primary storage device 1040, a secondary storage device 1050, an output IF (Interface) 1060, an input IF 1070, and a network IF 1080 connected by a bus 1090. has.

The arithmetic unit 1030 operates based on programs stored in the primary storage device 1040 and the secondary storage device 1050, programs read from the input device 1020, and performs various processes. The primary storage device 1040 is a memory device such as a RAM that temporarily stores data used by the arithmetic unit 1030 for various calculations. Further, the secondary storage device 1050 is a storage device in which data used by the arithmetic device 1030 for various calculations and various databases are registered, and is realized by a ROM (Read Only Memory), an HDD, a flash memory, or the like.

The output IF 1060 is an interface for transmitting information to be output to the output device 1010 that outputs various information such as a monitor or printer, and is, for example, a USB (Universal Serial Bus), a DVI (Digital Visual Interface), This is realized using a connector compliant with standards such as HDMI (registered trademark) (High Definition Multimedia Interface). Further, the input IF 1070 is an interface for receiving information from various input devices 1020 such as a mouse, a keyboard, and a scanner, and is realized by, for example, a USB or the like.

Note that the input device 1020 is, for example, an optical recording medium such as a CD (Compact Disc), a DVD (Digital Versatile Disc), or a PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), or a tape. It may be a device that reads information from a medium, a magnetic recording medium, a semiconductor memory, or the like. Furthermore, the input device 1020 may be an external storage medium such as a USB memory.

Network IF 1080 receives data from other devices via network N and sends it to computing device 1030, and also sends data generated by computing device 1030 to other devices via network N.

Arithmetic device 1030 controls output device 1010 and input device 1020 via output IF 1060 and input IF 1070. For example, the arithmetic device 1030 loads a program from the input device 1020 or the secondary storage device 1050 onto the primary storage device 1040, and executes the loaded program.

For example, when the computer 1000 functions as the information processing device 100, the arithmetic device 1030 of the computer 1000 executes a program (for example, an information processing program) loaded on the primary storage device 1040, thereby performing the same operation as the control unit 130. Realize the functions of That is, the arithmetic device 1030 realizes processing by the information processing device 100 according to the embodiment in cooperation with a program (for example, an information processing program) loaded onto the primary storage device 1040.

[7. others]
Among the processes described in the embodiments described above, all or part of the processes described as being performed automatically can be performed manually, or all of the processes described as being performed manually can be performed manually. Alternatively, some of the steps can be performed automatically using known methods. In addition, information including the processing procedures, specific names, and various data and parameters shown in the above documents and drawings may be changed arbitrarily, unless otherwise specified.

In the above-described embodiment, in order to realize the information processing method by the information processing device 100 (for example, see FIG. , and the processing function corresponding to the evaluation unit 133) may be realized as an add-on to the information processing program installed in advance in the information processing device 100, or may be realized as a dedicated provided program using a lightweight programming language or the like. This may be realized by flexible description.

Furthermore, each component of each device shown in the drawings is functionally conceptual, and does not necessarily need to be physically configured as shown in the drawings. In other words, the specific form of distributing and integrating each device is not limited to what is shown in the diagram, and all or part of the devices can be functionally or physically distributed or integrated in arbitrary units depending on various loads and usage conditions. Can be integrated and configured. For example, the first estimation section 131 and the second estimation section 132 of the control section 130 may be functionally integrated.

Further, the above-described embodiments and modified examples can be combined as appropriate within the range that does not conflict with the processing contents.

[8. effect]
The information processing device 100 according to the embodiment described above includes a first estimation section 131, a second estimation section 132, and an evaluation section 133. The first estimation unit 131 estimates the probability that a related predetermined action will be performed from among a plurality of estimation models prepared in advance for each predetermined action in order to estimate the probability that a predetermined action will be performed by the user. A state transition of a predetermined behavior is estimated using a plurality of first models for the purpose of the present invention. The second estimation unit 132 uses a second model for estimating the probability of performing a predetermined action that has a negative relationship with the state transition of the predetermined action from among the plurality of estimation models. Estimate the probability that a predetermined action with negative association will be performed. The evaluation unit 133 evaluates the reliability of the second model based on the accuracy estimated by the second estimation unit 132.

For this reason, the information processing apparatus 100 according to the embodiment can easily confirm the reliability of the generated model by comparing the estimation results of a plurality of models.

In the information processing device 100 according to the embodiment, the first estimation unit 131 selects a plurality of first models from among the plurality of estimation models based on the feature amount of a predetermined behavior associated with the estimation model. do. Thereby, the information processing apparatus 100 according to the embodiment can appropriately select a plurality of models that are related to the associated actions.

Furthermore, in the information processing device 100 according to the embodiment, the second estimation unit 132 selects, as the second model, an estimation model that estimates the probability that a predetermined action that is logically contradictory to a predetermined state transition will be performed. Thereby, the information processing apparatus 100 according to the embodiment can estimate the accuracy of the second model based on the contradiction in the estimation results by the second model.

Furthermore, in the information processing device 100 according to the embodiment, the second estimation unit 132 estimates the probability that a predetermined action having a negative association will be performed during a period in which the predetermined state transition is estimated, and evaluates the probability that a predetermined action having a negative association will be performed. The unit 133 evaluates the reliability of the second model based on the change in accuracy estimated by the second estimation unit 132. Thereby, the information processing apparatus 100 according to the embodiment can evaluate the estimation results of the second model along the time series of state transitions.

As mentioned above, the embodiments of the present application have been described in detail based on several drawings, but these are merely examples, and various modifications and variations can be made based on the knowledge of those skilled in the art, including the embodiments described in the disclosure section of the invention. It is possible to carry out the invention in other forms with modifications.

Further, the above-mentioned "section, module, unit" can be read as "means", "circuit", etc. For example, the control section can be read as a control means or a control circuit.

10 Terminal device 100 Information processing device 110 Communication unit 120 Storage unit 121 User information storage unit 122 Estimated model storage unit 130 Control unit 131 First estimation unit 132 Second estimation unit 133 Evaluation unit

Claims (6)

In order to estimate the probability that a predetermined action will be performed by a user , a search query searched by each user that satisfies a predetermined condition, and a score given to the search query based on the commonality of the search query. A score list consisting of a set of a plurality of first models for estimating the probability that the predetermined action having relevance will be performed among the plurality of estimation models prepared in advance for each of the predetermined actions. A score obtained by inputting a search query entered by the user in a predetermined period for one of the plurality of selected first models, and a score for the other first model. From the predetermined behavior corresponding to the one first model by determining whether the scores obtained by inputting the search query input by the user during the predetermined period are both equal to or greater than a threshold value. a first estimation unit that estimates a state transition to the predetermined behavior corresponding to the other first model ;
For estimating the probability that the predetermined action that has a negative relationship with the predetermined action after the state transition estimated by the first estimator among the plurality of estimation models will be performed. by selecting a second model of , a second estimation unit that estimates the probability that the predetermined action having the negative association will be performed by the user;
An information processing device comprising: an evaluation section that evaluates the reliability of the second model based on the accuracy estimated by the second estimation section. The first estimation unit is
The information processing device according to claim 1, wherein a plurality of the first models are selected from the plurality of estimation models based on feature quantities of the predetermined behavior that are associated with the estimation models. . The second estimation unit is
The information processing according to claim 1 or 2, wherein the estimation model that estimates the probability of performing the action that is logically contradictory to the state transition to the predetermined action is selected as the second model. Device. The second estimation unit is
estimating the probability that the action having the negative association will be performed in the period in which the state transition to the predetermined action is estimated;
The evaluation department is
The information processing device according to any one of claims 1 to 3, wherein the reliability of the second model is evaluated based on a change in the accuracy estimated by the second estimation unit. An information processing method performed by a computer, the method comprising:
In order to estimate the probability that a predetermined action will be performed by a user , a search query searched by each user that satisfies a predetermined condition, and a score given to the search query based on the commonality of the search query. A score list consisting of a set of a plurality of first models for estimating the probability that the predetermined action having relevance will be performed among the plurality of estimation models prepared in advance for each of the predetermined actions. A score obtained by inputting a search query entered by the user in a predetermined period for one of the plurality of selected first models, and a score for the other first model. From the predetermined behavior corresponding to the one first model by determining whether the scores obtained by inputting the search query input by the user during the predetermined period are both equal to or greater than a threshold value. a first estimation step of estimating a state transition to the predetermined behavior corresponding to the other first model ;
For estimating the probability that the predetermined action that has a negative relationship with the predetermined action after the state transition estimated in the first estimation step will be performed among the plurality of estimation models. by selecting a second model of , a second estimation step of estimating the probability that the predetermined action having the negative association will be performed by the user;
and an evaluation step of evaluating reliability of the second model based on the accuracy estimated by the second estimation step. to the computer,
In order to estimate the probability that a predetermined action will be performed by a user , a search query searched by each user that satisfies a predetermined condition, and a score given to the search query based on the commonality of the search query. A score list consisting of a set of a plurality of first models for estimating the probability that the predetermined action having relevance will be performed among the plurality of estimation models prepared in advance for each of the predetermined actions. A score obtained by inputting a search query entered by the user in a predetermined period for one of the plurality of selected first models, and a score for the other first model. From the predetermined behavior corresponding to the one first model by determining whether the scores obtained by inputting the search query input by the user during the predetermined period are both equal to or greater than a threshold value. a first estimation procedure for estimating a state transition to the predetermined behavior corresponding to the other first model ;
to estimate the probability that the predetermined action that has a negative relationship with the predetermined action after the state transition estimated by the first estimation procedure will be performed among the plurality of estimation models; by selecting a second model of , a second estimation procedure of estimating the probability that the predetermined action having the negative association will be performed by the user;
and an evaluation procedure for evaluating reliability of the second model based on the accuracy estimated by the second estimation procedure.

Images