JP6596605B1 - Estimation system, estimation method, and computer program - Google Patents

Abstract

A technique for estimating a feature of an object is provided. In an estimation method according to an aspect of the present disclosure, a first feature of each of a plurality of first objects is represented, or a plurality of first objects having a common first feature are represented. First data is acquired (S110). Further, second data representing the second feature of each of the plurality of first objects is acquired (S130). Estimation for estimating the first feature from the second feature based on the first feature and the second feature of each of the plurality of first objects determined from the first data and the second data A model is constructed (S140). Furthermore, object data representing the second feature of the second object is acquired. Based on the target data, the first feature of the second target object is estimated using the estimation model. [Selection] Figure 3

Description

  The present disclosure relates to estimation systems and data combining systems and corresponding methods.

  Analyzing consumer behavior is traditionally performed. For example, a consumer's purchasing behavior is analyzed, a potential purchase layer for a product is estimated, and an advertisement for the product is distributed to the estimated purchase layer. There is also known a technique for providing a user with various services by connecting a plurality of types of data related to the same person using unified identification information (see Patent Document 1).

JP 2018-5536 A

However, with the conventional method of collecting and connecting a plurality of types of data, it is not possible to analyze the characteristics of a person whose data is missing.
Therefore, according to one aspect of the present disclosure, it is desirable to be able to provide a technique for estimating a target feature indirectly from obtainable data.

  An estimation system according to one aspect of the present disclosure includes a first acquisition unit, a second acquisition unit, a model construction unit, a target data acquisition unit, and an estimation unit. The first acquisition unit is configured to acquire first data representing a first feature of each of the plurality of first objects. Alternatively, the first acquisition unit is configured to acquire first data representing a plurality of first objects having the same first characteristic.

  The second acquisition unit is configured to acquire second data representing a second feature of each of the plurality of first objects. The model construction unit obtains the first feature from the second feature based on the first feature and the second feature of each of the plurality of first objects determined from the first data and the second data. It is configured to build an estimation model for estimation.

  The target data acquisition unit is configured to acquire target data representing the second feature of the second target object. The estimation unit is configured to estimate the first feature of the second target object using the estimation model based on the target data.

  According to this estimation system, the second feature that the first subject and the second subject have in common in an environment in which data relating to the first feature cannot be acquired with respect to the second subject. The first feature of the second object can be estimated based on the data representing. Therefore, according to this estimation system, it is possible to indirectly estimate the first feature of the second object that cannot be directly discriminated.

  According to one aspect of the present disclosure, the first feature may be a dynamic feature. Similarly, the second feature may be a dynamic feature. The first data may include information capable of determining a change in the first feature. The second data may include information capable of determining the change in the second feature. The target data may include information that can determine a change in the second feature.

  The model construction unit may be configured to construct an estimation model for estimating the change of the first feature from the change of the second feature based on the change of the first feature and the change of the second feature. Good. The estimation unit may be configured to estimate a change in the first feature of the second target object using the estimation model based on the target data. The data collection load for dynamic features is higher than the data collection for static features. Therefore, it is meaningful to be able to estimate dynamic feature changes.

  According to one aspect of the present disclosure, each of the plurality of first objects and second objects may be an individual or a consumer in a cluster unit. That is, the consumer may be a single individual or a cluster that includes multiple individuals.

  The first characteristic may be a characteristic related to the first behavior of the consumer. The second feature may be a feature related to the second behavior of the consumer. The second action may be an online action.

  According to one aspect of the present disclosure, the second acquisition unit includes, from a database system that stores behavior history data related to online behaviors of a plurality of online users in association with online identifiers of the plurality of online users, As the data, action history data associated with an online identifier corresponding to each of the plurality of first objects may be acquired. For this reason, the first data may include information capable of discriminating an online identifier corresponding to each of the plurality of first objects.

  According to one aspect of the present disclosure, the target data acquisition unit receives, as target data, the second target object from the database system in response to input of data capable of determining an online identifier corresponding to the second target object. It may be configured to acquire action history data associated with a corresponding online identifier.

  According to one aspect of the present disclosure, the first data may be data representing an action history related to the first action of each of the plurality of first objects. The second data may be data representing an action history related to the second action of each of the plurality of first objects. The target data may be data representing an action history related to the second action of the second target object.

  The model construction unit includes an action history related to each first action of the first object determined based on the first data and each of the first object determined based on the second data. An estimation model for estimating a state change related to the first action of the second target object from the target data may be constructed based on the action history related to the second action.

  According to one aspect of the present disclosure, the first feature may be a feature related to whether or not the first action is performed. The model construction unit constructs an estimation model for estimating the possibility of performing the first action from the second feature as the estimation model based on the second feature of each of the plurality of first objects. May be configured. The estimation unit may be configured to estimate the possibility that the second object performs the first action using the estimation model.

  The estimation system may include a third acquisition unit configured to acquire third data representing a third characteristic of each of the plurality of second objects. The target data acquisition unit may be configured to acquire data representing the second feature of each of the plurality of second target objects as the target data. The estimation unit may estimate the first feature of each of the plurality of second target objects using the estimation model based on the target data. The estimation unit may be configured to output data representing the first feature and the third feature of each of the plurality of second objects as extension data for the third data.

  According to one aspect of the present disclosure, an estimation system is configured to acquire first data representing a plurality of first objects having a common characteristic regarding the presence or absence of execution of a first action A first acquisition unit, a second acquisition unit configured to acquire second data representing characteristics related to the second behavior of each of the plurality of first objects determined from the first data; In order to estimate the possibility of performing the first action from the characteristics related to the second action as an estimation model based on the characteristics related to the second action of each of the plurality of first objects determined from the second data Based on the target data, a model construction unit configured to construct an estimation model of the target object, a target data acquisition unit configured to acquire target data representing characteristics related to the second behavior of the second target object, and the target data Using the estimation model When configured estimator so that the second object to estimate the possibility of performing the first action, it may be provided to estimate system comprising a.

  According to one aspect of the present disclosure, a data combination system may be provided. The data providing system may include a first acquisition unit, a second acquisition unit, a third acquisition unit, a fourth acquisition unit, and a data combination unit.

  In the data combination system, the first acquisition unit may be configured to acquire a plurality of first feature data related to a plurality of first objects. Each of the plurality of first feature data may be data representing the first feature of the corresponding first object.

  The second acquisition unit may be configured to acquire a plurality of second feature data related to the plurality of first objects. Each of the second feature data may be data representing the second feature of the corresponding first object.

  The third acquisition unit may be configured to acquire a plurality of third feature data related to the plurality of second objects. Each of the third feature data may be data representing the third feature of the corresponding second object.

  The fourth acquisition unit may be configured to acquire a plurality of fourth feature data related to the plurality of second objects. Each of the fourth feature data may be data representing the second feature of the corresponding second object.

  The data combination unit includes a plurality of second objects determined from the plurality of fourth feature data and a second feature of each of the plurality of first objects determined from the plurality of second feature data. The first feature data of the first object for each combination of the first object and the second object at least similar to the second feature, The database may be configured to generate a combined database representing the first feature and the third feature for each combination in association with the third feature data of the second object.

  According to this data combination system, the first feature and the third feature can be associated from the similarity of the second feature. Therefore, in an environment where the first feature of the second object cannot be directly discriminated from the data, the first feature of the second object can be estimated indirectly from the combined database.

  According to one aspect of the present disclosure, the data combining unit outputs the first feature data of the first object and the third feature data of the second object for each combination. The second feature data and the fourth feature data of the second object may be associated with each other. Accordingly, the data combining unit may be configured to generate a combined database representing the first feature, the second feature, and the third feature for the corresponding combination for each combination as a combined database. .

  Each of the plurality of first objects and second objects in the data combination system may be an individual or a cluster consumer. The second feature may be a feature related to consumer online behavior.

  According to one aspect of the present disclosure, the second acquisition unit includes, as the plurality of second feature data, the online behavior associated with each of the plurality of first identifiers from the database system described above. It may be comprised so that the action history data regarding may be acquired. In this case, the plurality of first feature data may include information capable of determining an online identifier corresponding to each of the plurality of first objects.

  The fourth acquisition unit may be configured to acquire action history data associated with an online identifier corresponding to each of the plurality of second objects as the plurality of fourth feature data from the database system. Good. In this case, the plurality of third feature data may include information capable of determining an online identifier corresponding to each of the plurality of second objects.

  According to one aspect of the present disclosure, a computer program for causing a computer to function as the first acquisition unit, the second acquisition unit, the model construction unit, the target data acquisition unit, and the estimation unit in the estimation system described above is provided. Good.

  According to one aspect of the present disclosure, a computer program for causing a computer to function as the first acquisition unit, the second acquisition unit, the third acquisition unit, the fourth acquisition unit, and the data combination unit in the above-described data combination system is provided. May be. The computer program may be recorded on a computer-readable non-transitory recording medium.

  According to one aspect of the present disclosure, an estimation method executed by a computer may be provided. The estimation method includes obtaining first data representing a first feature of each of a plurality of first objects, or representing a plurality of first objects having a common first feature; Obtaining second data representing a second characteristic of each of the first objects of the first object, and first of each of the plurality of first objects determined from the first data and the second data Constructing an estimation model for estimating the first feature from the second feature based on the feature and the second feature, and obtaining target data representing the second feature of the second subject And estimating the first feature of the second target object using the estimation model based on the target data.

  According to one aspect of the present disclosure, an estimation method executed by a computer is configured to obtain first data representing a plurality of first objects having common characteristics regarding the presence or absence of execution of the first action. And acquiring second data representing characteristics relating to the second behavior of each of the plurality of first objects determined from the first data, and a plurality of first data determined from the second data Constructing an estimation model for estimating the possibility of performing the first action from the characteristics relating to the second action as the estimation model based on the characteristics relating to the second action of each of the objects; Obtaining target data representing characteristics related to the second behavior of the target object, and estimating a possibility that the second target object performs the first behavior using the estimation model based on the target data; The estimation method including Test may be.

  According to one aspect of the present disclosure, a data combining method executed by a computer may be provided. The data combination method acquires a plurality of first feature data related to a plurality of first objects, each of which represents a first feature of the first object corresponding to each of the first feature data. And a plurality of second feature data relating to the plurality of first objects, each of which represents a plurality of second feature data representing the second feature of the corresponding first object. Obtaining a plurality of third feature data relating to a plurality of second objects, each of which represents a third feature of the corresponding second object, and a plurality of third feature data A plurality of fourth feature data relating to the second object of the second object, each of the plurality of fourth feature data representing the second feature of the corresponding second object, Each of the plurality of first objects discriminated from the second feature data Based on the second feature and the second feature of each of the plurality of second objects determined from the plurality of fourth feature data, the first object at least similar to the second feature, For each combination with the second object, the first feature data of the first object and the third feature data of the second object are associated with each other, and the first feature and third for each combination. Generating a combined database representing the features of

It is a figure explaining the estimation method of 1st embodiment. It is a figure explaining the composition and peripheral composition of the information processing system of a first embodiment. It is a flowchart showing the model construction process which a processor performs. It is explanatory drawing regarding construction of an estimation model. It is a flowchart showing the estimation process which a processor performs. It is a flowchart showing the model construction process of 2nd embodiment. 7A and 7B are diagrams showing the configuration of the purchase history database after expansion. It is explanatory drawing regarding discrimination | determination of a state change. It is a flowchart showing the estimation process of 2nd embodiment. FIG. 10A is a diagram illustrating a configuration of a customer database before and after expansion, and FIG. 10B is a diagram illustrating a configuration of another example customer database having customer data for each cluster. It is a figure explaining the structure and peripheral structure of the information processing system of 3rd embodiment. It is a figure explaining the data combination method of 3rd embodiment. It is a flowchart showing the data combination process which a processor performs. It is a figure showing the structure of the 1st action database before and behind expansion. It is a figure showing the structure of the 2nd action database before and behind expansion. It is a figure which illustrates the structure of a joint database.

In the following, exemplary embodiments of the present disclosure will be described with reference to the drawings.
[First embodiment]
As shown in FIG. 1, the information processing system 1 of the present embodiment cannot obtain data relating to the first behavior from the first data D1 relating to the first behavior of the first consumer group C1. The first behavior of the consumer group C2 is configured to be estimated from the second data D2 relating to the second behavior of the first consumer group C1 and the second consumer group C2.

  Specifically, the information processing system 1 determines the first behavior of the first consumer group C1 determined from the first data D1 and the second behavior of the first consumer group C1 determined from the second data D2. Based on the relationship with the behavior, an estimation model M for estimating the first behavior is constructed from the second behavior.

  The information processing system 1 estimates the first behavior of the second consumer group C2 based on the estimation model M and the second behavior of the second consumer group C2 determined from the second data D2. . The estimation here includes estimating past actions and predicting future actions.

  As illustrated in FIG. 2, the information processing system 1 includes a processor 11, a memory 13, a storage 15, and a communication interface 17. The processor 11 is configured to execute processing according to a computer program stored in the storage 15.

  The memory 13 includes a ROM and a RAM. The storage 15 stores a computer program and various data used for processing executed by the processor 11. Examples of the storage 15 include a hard disk drive and a solid state drive.

  The communication interface 17 is configured to be able to communicate with an external device through a network. The information processing system 1 is configured to be able to communicate with a data collection system 20 that collects data related to consumers through a communication interface 17. Furthermore, the information processing system 1 is configured to be able to communicate with a user system 30 that is an information processing system of a user who uses a service provided by the information processing system 1.

  The information processing system 1 acquires first data D1 related to the first behavior of the first consumer group C1 from the user system 30 through communication. The information processing system 1 further acquires second data D2 related to the second behavior of the first consumer group C1 and the second consumer group C2 from the data collection system 20. The information processing system 1 provides the estimation result regarding the first behavior of the second consumer group C2 to the user system 30 based on the first data D1 and the second data D2.

  A specific example of processing executed by the information processing system 1 will be further described using the model construction processing shown in FIG. 3 and the estimation processing shown in FIG. The model construction process and the estimation process are executed by the processor 11.

  In this specific example, the first action is a consumer use contract action for a specific service provided to a consumer by a company corresponding to the user system 30. The second behavior is consumer online behavior, specifically, web page browsing behavior.

  In the specific example illustrated in FIG. 3 and FIG. 4, the information processing system 1 constructs an estimation model M based on the online behavior of a consumer who has a usage contract record, and the online of a consumer who does not have a use contract record A score relating to the possibility of the use contract is calculated from the behavior using the estimation model M.

  In modern times, many consumers access the Internet and browse web pages. For this reason, a lot of consumer data can be collected for online behavior. Furthermore, if cookie technology is used, data relating to online behavior such as web page browsing behavior can be collected relatively easily. For this reason, adopting online behavior as the second behavior helps to build a highly accurate estimation model M related to the first behavior.

  When the processor 11 receives an instruction to execute the model construction process from the user system 30 through the network, the processor 11 starts the model construction process shown in FIG.

  In the model construction process, the processor 11 receives the first consumer list L1 from the user system 30 (S110). The first consumer list L1 represents, as the first consumer group C1, a plurality of consumers who are customers of companies corresponding to the user system 30 and have a track record of using a specific service. For example, the first consumer list L1 has a configuration in which customer IDs of a plurality of consumers are listed. The customer ID is, for example, a customer ID managed by a company that provides the first consumer list L1.

  In subsequent S120, the processor 11 converts each of the customer IDs of the plurality of consumers listed in the first consumer list L1 into a cookie ID with reference to the conversion table (S120). The processor 11 can acquire the conversion table from the data collection system 20. The cookie ID is an identifier issued to an online user (specifically, a web browser) based on cookie technology.

  The association between the customer ID and the cookie ID can be realized, for example, by cooperation of a company that provides the first consumer list L1. The association can be realized by embedding a tag for issuing a cookie managed by the data collection system 20 in a company web page.

  In subsequent S130, the processor 11 acquires the online behavior data D3 of each of the plurality of consumers listed in the first consumer list L1 from the data collection system 20 based on the cookie ID. The data collection system 20 stores the online behavior data D3 of a plurality of online users in association with the corresponding cookie ID.

  The data collection system 20 is configured to generate and store online behavior data D3 of a plurality of online users by issuing cookies and analyzing the behavior of online users using cookies. The data collection system 20 is configured to provide the information processing system 1 with online behavior data D3 associated with the cookie ID specified by the information processing system 1.

  The online behavior data D3 includes web page browsing history data by the corresponding online user, and data indicating the affinity category of the corresponding online user. The browsing history data is behavior history data regarding browsing behavior of the corresponding online user's web page.

  Specifically, the browsing history data includes web page identification information and browsing date information for each web page browsed by the corresponding online user. The affinity category indicates a category in which the corresponding online user is interested and interested.

  In subsequent S140, the processor 11 uses the plurality of online behavior data D3 corresponding to the plurality of consumers in the first consumer list L1, and uses the plurality of online behavior data D3 to determine the possibility of using a specific service. An estimation model M for calculation is constructed.

  Specifically, the processor 11 can generate a plurality of teacher data D4 based on the online behavior data D3 of a plurality of consumers who have the usage contract results obtained in S130. As shown in FIG. 4, each of the teacher data D4 may be a combination of a consumer's web browsing history and affinity category determined from the corresponding online behavior data D3, and a score regarding the possibility of a usage contract.

  For example, in the teacher data D4, the web browsing history can be represented by a numerical value indicating the presence or absence of browsing of each of a plurality of web pages or an array indicating the number of browsing times. The score may be expressed as a usage contract probability (%). All consumers in the first consumer list L1 have a track record of usage contracts. Thus, the score can be set to the value 100.

  The processor 11 can construct an estimation model M that calculates a score related to the use contract probability with respect to the input of the site browsing history and the affinity category by machine learning using such a plurality of teacher data D4.

  In subsequent S150, the processor 11 outputs the constructed estimation model M. Thereafter, the model construction process ends. In S150, the processor 11 can execute an operation of storing the estimated model M in the memory 13 or the storage 15 as an output operation. Alternatively, the processor 11 can execute an operation of transmitting the estimation model M to the user system 30.

  Furthermore, when the processor 11 receives an execution command for the estimation process from the user system 30 through the network, the processor 11 starts the estimation process shown in FIG.

  When the estimation process is started, the processor 11 receives the second consumer list L2 from the user system 30 (S210). The second consumer list L2 represents a plurality of consumers to be estimated as the second consumer group C2 for the possibility of using the specific service. Similarly to the first consumer list L1, the second consumer list L2 has a configuration in which customer IDs of a plurality of consumers are listed.

  In subsequent S220, the processor 11 converts each of the customer IDs of the plurality of consumers listed in the second consumer list L2 into cookie IDs in the same manner as in S120. In subsequent S230, the processor 11 acquires the online behavior data D3 of each of the plurality of consumers listed in the second consumer list L2 from the data collection system 20 based on the cookie ID.

  In subsequent S240, the processor 11 acquires the estimated model M. In S <b> 240, the processor 11 can acquire the estimated model M from the user system 30. Alternatively, the processor 11 can acquire the estimated model M by reading out the estimated model M of the user system 30 constructed and stored in the storage 15 in the model construction process from the storage 15.

  In subsequent S250, the processor 11 estimates the combination of the site browsing history and the affinity category determined from the corresponding consumer online behavior data D3 for each of the plurality of consumers listed in the second consumer list L2. By inputting to the model M, a score related to the usage contract probability of the corresponding consumer is calculated. As a result, the processor 11 estimates the possibility of using a specific service for each of the plurality of consumers corresponding to the second consumer group C2.

  In subsequent S260, the processor 11 transmits the score data D5 to the transmission source of the second consumer list L2. The score data D5 corresponds to data obtained by adding a score corresponding to each consumer in the second consumer list L2 to the second consumer list L2. Thereafter, the processor 11 ends the estimation process.

  In the information processing system 1 according to the present embodiment described above, the first consumer list L1 representing a plurality of consumers having the same characteristics related to the use contract, specifically, a plurality of consumers with a track record of use contract is acquired. (S110). And the online action data D3 which shows the characteristic regarding each online action of several consumers in the 1st consumer list L1 is acquired (S130).

  Furthermore, an estimation model M for estimating the possibility of a usage contract as a feature related to a usage contract from a feature related to the online behavior based on a feature related to the usage contract and a feature related to the online behavior of each consumer in the first consumer list L1. Is constructed (S140).

  Further, a second consumer list L2 representing a plurality of consumers for which the possibility of use contract is estimated is acquired (S210). Online behavior data D3 of each of the plurality of consumers in the second consumer list L2 is acquired (S230).

  Thereafter, for each of the plurality of consumers in the second consumer list L2, the possibility of the corresponding consumer's usage contract is estimated based on the corresponding consumer's online behavior data D3 (S250). In the information processing system 1 of the present embodiment, the possibility of the usage contract of the second consumer group C2 is estimated from the results regarding the usage contract of the first consumer group C1 in this way.

  Therefore, in the information processing system 1 of the present embodiment, it is possible to estimate consumer behavior that cannot be discriminated directly from the data by using the data that can be acquired meaningfully. Therefore, this information processing system 1 is very meaningful.

[Second Embodiment]
The information processing system 1 according to the second embodiment is configured such that the processor 11 executes a model construction process illustrated in FIG. 6 and an estimation process illustrated in FIG. 9 instead of the processes illustrated in FIGS. 3 and 5. Thus, it is different from the first embodiment. Except for this point, the information processing system 1 of the second embodiment is configured similarly to the first embodiment. Accordingly, the model construction process and the estimation process executed by the processor 11 will be selectively described below.

  The processor 11 starts the model construction process shown in FIG. 6 when the predetermined condition is satisfied. The processor 11 can start the model construction process based on a command from an operator of the information processing system 1.

  When the model construction process shown in FIG. 6 is started, the processor 11 acquires the purchase history database B1 related to the purchase behavior of the first consumer group C1 from the purchase data collection system 40 (S310). The purchase history database B1 has purchase history data of each first consumer.

  The purchase history data is associated with a consumer ID, which is identification information of the corresponding consumer, and for each purchase in the online space and offline space of the corresponding consumer, the purchase product identification information, purchase date and time information, Have

  The purchase data collection system 40 may be a system of a company that operates the information processing system 1 or may be a system of another company. The processor 11 can acquire the purchase history database B1 through the network.

  In subsequent S320, the processor 11 determines the cookie ID of each consumer based on the consumer ID included in the purchase history data of each consumer. The determination of the cookie ID can be realized by acquiring a conversion table from the data collection system 20 as in the first embodiment. The association between the consumer ID and the cookie ID can be realized by embedding the tag in a website that can be associated.

  In subsequent S330, the processor 11 acquires the online behavior data D3 corresponding to each of the first consumers from the data collection system 20 based on the cookie ID, and expands the purchase history database B1. The expanded purchase history database B2 is shown in FIG. 7A. The expanded purchase history database B2 has a record representing a purchase history, a web browsing history, and an affinity category for each of the first consumers.

  In subsequent S340, the processor 11 determines, for each of the first consumers, a change in the state of the consumer from the purchase history of the consumer indicated by the corresponding record. Thereafter, the processor 11 expands the purchase history database B2 so as to add data representing a state change to each record of the first consumer. The data representing the state change includes information that can determine the state of the corresponding consumer at each time.

  The expanded purchase history database B3 is shown in FIG. 7B. The expanded purchase history database B3 has a record representing the purchase history, state change, web browsing history, and affinity category of the corresponding consumer for each first consumer.

  In the estimation process illustrated in FIG. 9, when estimating the change probability of the second consumer group C2 to the specific state, in S340, the processor 11 determines the corresponding consumer non-specific state Z1 for each of the first consumers. The change to the specific state Z2 can be determined. For example, the non-specific state corresponds to a state where a specific type of product is not purchased and corresponds to a state where a specific type of product is purchased. As a specific type of product, a product related to childbirth can be cited as an example.

  FIG. 8 conceptually illustrates a state change discrimination method. The example shown in FIG. 8 indicates that the purchase history of the consumer purchases the first product corresponding to the circle mark before the time T1 on the time axis, and corresponds to the square mark after the time T1. Indicates that you are purchasing a second product.

  When the purchase history data indicates such a change in purchase, the processor 11 determines that it is in the non-specific state Z1 in which the second product is not purchased before the time T1, and after the time T1, the processor 11 It is determined that the two products are in the specific state Z2 in which they are purchased. Then, in the purchase history database B3, as data representing the state change, data indicating that the non-specific state Z1 before the time T1 and the specific state Z2 after the time T1 are recorded as data of each consumer. Add to

  In subsequent S350, the processor 11 constructs an estimation model M for estimating a consumer's state change from a change in online behavior. Specifically, the processor 11 creates teacher data representing a web browsing history, an affinity category, and a consumer state change as teacher data corresponding to each record in the purchase history database B3.

  In the teacher data, the web browsing history can be expressed by a numerical value indicating the presence or absence of browsing of each of a plurality of web pages or an array indicating the number of times of browsing for each section defined by dividing the collection period of the web browsing history. In the teacher data, the consumer state change may be an array having values representing the consumer state for each section.

  Based on such a group of teacher data, the processor 11 can construct an estimation model M that estimates a consumer's state change from the consumer's web browsing history and affinity category. Specifically, the estimation model M may be a function that outputs a score related to a state change probability of a consumer to a specific state in response to an input of a web browsing history and an affinity category.

  In subsequent S <b> 360, the processor 11 stores the constructed estimation model M in the storage 15. Thereafter, the model construction process shown in FIG.

  Furthermore, when the processor 11 receives an execution command for the estimation process from the user system 30 through the network, the processor 11 starts the estimation process shown in FIG. When the estimation process is started, the processor 11 receives the customer database B5 related to the second consumer group C2 from the user system 30 (S410).

  The customer database B5 includes customer data of each second consumer who is a customer of the user. Each of the customer data includes a corresponding consumer customer ID and corresponding consumer feature data. The customer ID is an ID managed by a company on the user system 30 side, which is different from the consumer ID of the purchase history database B1.

  In subsequent S420, the processor 11 determines the cookie ID corresponding to each of the second consumers based on the customer ID included in the customer data. In subsequent S430, the processor 11 acquires each online behavior data D3 of the second consumer from the data collection system 20 based on the determined cookie ID.

  In subsequent S440, the processor 11 acquires the estimated model M by reading out the estimated model M constructed and stored by the model construction process from the storage 15. In subsequent S450, the processor 11 estimates each state change of the second consumer based on the acquired estimation model M. Specifically, the processor 11 can calculate a score relating to a state change probability to a specific state as a feature amount of the state change.

  Thereafter, the processor 11 adds the data representing the corresponding second consumer's web browsing history, the affinity category, and the estimated state change to the customer data included in the customer database B5 received in S410. The database B5 is expanded. The expanded customer database B6 is shown in FIG. 10A.

  The expanded customer database B6 can have a score related to the state change probability of the second consumer to the specific state for each of the second consumers. The processor 11 transmits the expanded customer database B6 to the user system 30 (S460).

  According to the information processing system 1 of the present embodiment described above, information on a consumer's customer database B5 based on the customer ID of each consumer indicated by the customer database B5 ( Information on the probability of change to a specific state) can be added.

  As described above, in the present embodiment, the customer ID is converted into the cookie ID, the online behavior data D3 corresponding to the customer is obtained, and the consumption based on the online behavior data D3 and the presumed model M constructed in advance. The state change of the person. The estimation model M is constructed based on the relationship between the state change determined from the purchase history obtained beforehand by data collection and the corresponding online behavior of the consumer.

  As described above, in the information processing system 1 according to the present embodiment, a consumer state change that can be determined from a change in purchase behavior can be estimated from data related to online behavior that is easier to collect than the purchase history. Therefore, the information processing system 1 of this embodiment is very meaningful.

  In the above-described embodiment, personal information is exchanged between the user system 30 and the information processing system 1. However, in order to protect personal information, data in units of clusters, which is a set of individuals, may be exchanged between systems instead of data in units of individuals.

  For example, the customer database B5 provided from the user system 30 to the information processing system 1 may be customer data for each cluster instead of customer data for each individual. In this case, the customer data of each cluster may be data in which features of customers belonging to the cluster are expressed as statistics together with customer ID groups of customers belonging to the cluster. The statistic may be an average value, for example.

  In this case, the processor 11 of the information processing system 1 can acquire the online behavior data associated with the cookie ID group corresponding to the customer ID group indicated by the customer data for each cluster. Furthermore, the processor 11 can calculate a statistic regarding the state change for each cluster based on the online behavior data and the estimation model M.

  As shown in FIG. 10B, the processor 11 has statistical data on each cluster, customer data statistics, web page browsing history statistics, affiliate category statistics, and state change statistics. Customer database B7 may be configured to be transmitted to user system 30 at S460.

[Third embodiment]
Next, the information processing system 1 according to the third embodiment will be described. As shown in FIG. 11, the information processing system 1 of the third embodiment has the same hardware configuration as that of the first embodiment. On the other hand, the information processing system 1 includes the first behavior database B11 regarding the first behavior of the first consumer group C3 obtained from the first behavior data collection system 50 and the second consumption obtained from the second behavior data collection system 70. As shown in FIG. 12, the second behavior database B21 related to the second behavior of the consumer group C4 is configured to combine the online behavior of the first consumer group C3 and the second consumer group C4. It differs from the first embodiment and the second embodiment.

  That is, in the information processing system 1 of the present embodiment, each of the first consumers lacks data related to the second behavior, and each of the second consumers lacks data related to the first behavior. The data about the first behavior of the first consumer and the data about the second behavior of the second consumer based on the similarity of the online behavior in the existing environment, and generate a new combined database B30 Configured to do.

  In the first embodiment and the second embodiment, consumer characteristics are indirectly estimated from online behavior by using the estimation model M. However, in this embodiment, by connecting a plurality of databases based on the similarity of online behavior, it supplements data on consumer behaviors that are mutually lacking among the plurality of databases, and indirectly identifies consumer characteristics. Make wide estimation possible.

  The generation of the combined database B30 is realized by the processor 11 executing the data combining process shown in FIG. Hereinafter, as the description of the information processing system 1 of the present embodiment, the contents of the data combination process executed by the processor 11 will be selectively described according to the computer program.

  The processor 11 can start the data combining process shown in FIG. 13 in accordance with a command from the operator of the information processing system 1. When the data combination process is started, the processor 11 acquires the first behavior database B11 from the first behavior data collection system 50 (S510). The first behavior data collection system 50 is a system that collects data related to the first behavior of the first consumer group C3.

  The first behavior database B11 includes, for each of the first consumers, first behavior data D11 (representing a feature related to the first behavior of the first consumer in association with a consumer ID that is identification information of the corresponding consumer). 14).

  In the following, it is assumed that the first action is a purchase action and the first action data D11 is purchase history data of the first consumer. As with the purchase history data of the second embodiment, the first action data D11 is associated with the consumer ID, and for each purchase of the corresponding first consumer, the purchased product identification information, the purchase date and time information, Have

  In subsequent S520, the processor 11 determines a cookie ID corresponding to the consumer ID indicated by each first action data D11 in the first action database B11. The determination of the cookie ID can be realized by acquiring a conversion table from the online behavior data collection system 60 as in the first embodiment.

  Similar to the data collection system 20 of the first embodiment, the online behavior data collection system 60 generates a conversion table between a consumer ID and a cookie ID by issuing a cookie and analyzing an online user's behavior using the cookie. Configured. The online behavior data collection system 60 is further configured to generate online behavior data D3 for a plurality of online users and store the online behavior data D3 in association with the cookie ID.

  In subsequent S530, the processor 11 acquires the online behavior data D3 corresponding to each of the first consumers from the online behavior data collection system 60 based on the cookie ID, and expands the first behavior database B11.

  The first behavior database B12 after expansion is shown in FIG. The first action database B12 after expansion has a record representing the characteristics of the first action and the online action in association with the corresponding consumer ID for each of the first consumers. According to the example shown in FIG. 14, the first behavior database B <b> 12 has a record representing a purchase history, a web browsing history, and an affinity category. Below, this record is expressed as extended first action data D12.

  In subsequent S540, the processor 11 acquires the second behavior database B21 from the second behavior data collection system 70. The second behavior data collection system 70 is a system that collects data related to the second behavior of the second consumer group C4.

  The second behavior database B <b> 21 has second behavior data D <b> 21 (see FIG. 15) that represents the second behavior characteristics of the second consumer in association with the corresponding consumer ID for each second consumer. It may be understood that the consumer ID included in the second behavior data D21 is an independent ID different from the consumer ID included in the first behavior data D11.

  In the following description, the second behavior is a television broadcast viewing behavior, and the second behavior data D21 is an example of data representing characteristics of a second consumer television broadcast viewing behavior. The second behavior data D21 can have identification information of a program viewed by the second consumer as a viewing history in association with the consumer ID.

  In subsequent S550, the processor 11 determines a cookie ID corresponding to the consumer ID indicated by each second behavior data D21 in the second behavior database B21. The determination of the cookie ID can be realized by acquiring a conversion table from the online behavior data collection system 60, as in the processing in S520.

  In subsequent S560, the processor 11 acquires the online behavior data D3 corresponding to each of the second consumers from the online behavior data collection system 60 based on the cookie ID, and expands the second behavior database B21.

  The expanded second behavior database B22 is shown in FIG. The expanded second behavior database B22 has a record representing the characteristics of the second behavior and online behavior in association with the corresponding consumer ID for each second consumer. According to the example shown in FIG. 15, the second behavior database B <b> 22 has records representing viewing history, web browsing history, and affinity category. Hereinafter, this record is expressed as extended second action data D22.

  By executing the processing from S510 to S560 described above, the processor 11 causes the first behavior database B11 and the second behavior database B21 to store variables related to online behavior, specifically web variables, as common variables related to consumer characteristics. Add browsing history and affinity categories.

  This common variable is used to appropriately combine the first action data D11 and the second action data D21. The present embodiment is characterized in that a common variable related to online behavior is added to each of the first behavior database B1 and the second behavior database B2 in which no common variable exists, using a cookie technology.

  In subsequent S570, the processor 11 determines the first behavior data D11 and the second behavior data D21 of the first consumer and the second consumer that are similar to each other based on the similarity of the common variable, specifically, the similarity of the online behavior. The combined database B30 is generated so as to combine them. The similarities here should be understood in a broad sense including coincidence.

  For example, the processor 11 represents the web browsing history and the affinity category as common variables for each of the first consumer and the second consumer as a vector, and generates a feature vector having the common variable as an element. The processor 11 uses the distance (for example, cosine distance) on the feature space between the feature vector of the first consumer and the feature vector of the second consumer as the evaluation value regarding the similarity, and the first consumer and the second consumer. Calculate for all combinations with.

  The processor 11 combines the first action data D11 and the second action data D21 corresponding to the combination of the first consumer and the second consumer having the shortest distance, that is, the highest similarity, based on the distance for each combination. As a result, the combined database B30 can be generated.

  The combined database B30 may have one record in which the corresponding first behavior data D11 and second behavior data D21 are integrated for each combination of the first consumer and the second consumer to be combined. The combined database B30 may have one record in which the corresponding extended first action data D12 and extended second action data D22 are integrated for each combination.

  Alternatively, as shown in FIG. 16, the combined database B30 adds data D31 representing a combined relationship to the first behavior database B11 or the first behavior database B12 and the second behavior database B21 or the second behavior database B22. It may be configured to have.

  The data D31 representing the coupling relationship is data representing the coupling relationship between the first behavior data D11 or the extended first behavior data D12 and the second behavior data D21 or the expanded second behavior data D22. The combined database B <b> 30 may take other configurations that can cross-reference the data related to the first behavior of the first consumer and the data related to the second behavior of the corresponding second consumer.

  According to the information processing system 1 of the present embodiment described above, data relating to the first action is associated with data relating to the second action, and further, data relating to the online action is associated with the information. A combined database B30 for significant consumers with a high volume can be generated.

  In addition, even if the first action data D11 in the first action database B11 and the second action data D21 in the second action database B21 are data representing characteristics related to consumer actions in units of clusters for personal information protection. Good.

  For example, the first behavior data D11 and the second behavior data D21 may be configured to represent the behaviors of a plurality of individuals belonging to the corresponding cluster as statistics. In this case, the first behavior data D11 and the second behavior data D21 can be configured to include the consumer ID group of the corresponding cluster. Similarly, the extended first action data D12 and the extended second action data D22 for each cluster may be configured to represent features related to the online action as statistics.

[Others]
It goes without saying that the present disclosure is not limited to the above-described embodiments, and can take various forms. For example, consumer characteristics associated with characteristics related to online behavior as a common variable are not limited to the usage contract behavior, purchase behavior, and viewing behavior described above. For example, lifestyle characteristics such as marriage, childbirth, graduation, employment, retirement, and building a house may be combined with common variables as consumer characteristics. Further, as the consumer's state change, a change in the consumer's lifestyle may be determined from the purchase history of the consumer.

  The functions of one constituent element in the above-described embodiment may be distributed among a plurality of constituent elements. Functions of a plurality of components may be integrated into one component. A part of the configuration of the above embodiment may be omitted. At least a part of the configuration of the embodiment may be added to or replaced with the configuration of the other embodiment. Any aspect included in the technical idea specified from the wording of the claims is an embodiment of the present disclosure.

  Finally, the correspondence between terms will be described. The information processing system 1 of the first and second embodiments corresponds to an example of an estimation system, and the information processing system 1 of the third embodiment corresponds to an example of a data combination system.

  The processing of S110; S310; S510 executed by the processor 11 corresponds to an example of processing realized by the first acquisition unit. The processing of S120, S130; S320, S330; S520, S530 executed by the processor 11 corresponds to an example of processing realized by the second acquisition unit.

  The processing of S140; S340, and S350 executed by the processor 11 corresponds to an example of processing realized by the model construction unit. The processing of S220, S230; S420, S430 executed by the processor 11 corresponds to an example of processing realized by the target data acquisition unit. The processing of S250, S260; S450, S460 executed by the processor 11 corresponds to an example of processing realized by the estimation unit.

  The processing of S410 and S540 executed by the processor 11 corresponds to an example of processing realized by the third acquisition unit. The processing of S550 and S560 executed by the processor 11 corresponds to an example of processing realized by the fourth acquisition unit. The process of S570 executed by the processor 11 corresponds to an example of a process realized by the data combination unit.

  DESCRIPTION OF SYMBOLS 1 ... Information processing system, 11 ... Processor, 13 ... Memory, 15 ... Storage, 17 ... Communication interface, 20 ... Data collection system, 30 ... User system, 40 ... Purchase data collection system, 50 ... First action data collection system , 60 ... Online behavior data collection system, 70 ... Second behavior data collection system, B1, B2, B3 ... Purchase history database, B5, B6, B7 ... Customer database, B11, B12 ... First behavior database, B21, B22 ... 2nd action database, B30 ... combined database, D1 ... first data, D2 ... second data, D3 ... online action data, D4 ... teacher data, D5 ... score data, D11 ... first action data, D12 ... extended first Action data, D21 ... second action data, D22 ... extended second action data L1 ... first consumer list, L2 ... the second consumer list, M ... estimation model.

Claims (11)

First configured to obtain first data representing a first feature of each of a plurality of first subjects or representing a plurality of first subjects with which the first features are common. An acquisition unit;
A second acquisition unit configured to acquire second data representing a second feature different from the first feature of each of the plurality of first subjects ;
Based on the first feature and the second feature of each of the plurality of first subjects determined from the first data and the second data, the second feature to the first feature A model builder configured to build an estimation model for estimating features;
A target data acquisition unit configured to acquire target data representing the second characteristic of the second target person ;
An estimation unit configured to estimate the first characteristic of the second subject using the estimation model based on the target data;
Equipped with a,
The second feature is a feature related to an action involving communication,
The second acquisition unit is configured to store, as the second data, the behavior relating to each of the plurality of first subjects from a database system that stores behavior history data relating to the behavior involving each of the plurality of subjects. Get historical data,
The target data acquisition unit is configured to acquire the behavior history data related to the second target person from the database system as the target data,
The action involving the communication is an action in which information related to the action of the corresponding target person is transmitted to the database system by communication, and the action of the corresponding target person is recorded in the action history data based on the related information. der Ru estimation system. An estimation system according to claim 1 , wherein
The first data includes information capable of determining a change in the first feature,
The second data includes information capable of determining a change in the second feature,
The model construction unit estimates the change of the first feature from the change of the second feature based on the change of the first feature and the change of the second feature as the estimation model. Build the model ,
The estimation unit is an estimation system that estimates a change in the first characteristic of the second target person using the estimation model based on the target data. An estimation system according to claim 1 or claim 2 , wherein
Wherein the plurality of respective first subject and the second subject, estimated system is subject individual or cluster units. An estimation system according to any one of claims 1 to 3, wherein
The action involving communication is an online action that is an action on the subject's online space ,
The database system is a database system that stores the behavior history data related to the online behavior of each of the plurality of subjects in association with each online identifier of the plurality of subjects,
The first data includes information capable of determining the online identifier corresponding to each of the plurality of first subjects,
The second acquiring unit, from the database system, as the second data, acquires the action history data associated with the online identifier corresponding to each of the plurality of first subject,
The target data acquiring unit, the accordance online identifier to the input of the data that can be determined corresponding to the second subject, from said database system, as the target data, corresponding to said second subject said estimation system for acquiring the action history data associated with the online identifier. A fourth aspect Symbol placing the estimation system,
The online behavior includes an estimation system including browsing behavior for a plurality of web pages. A estimation system of any one of claims 1 to 5,
Said 1st characteristic is the characteristic regarding the presence or absence of execution of specific action of an object person ,
The model construction unit is an estimation model for estimating a possibility of performing the specific action from the second feature as the estimation model based on the second feature of each of the plurality of first subjects. Build
The estimation unit is an estimation system that estimates the possibility that the second target person performs the specific action using the estimation model. An estimation system according to any one of claims 1 to 6 ,
The second subject includes a plurality of second subjects ,
The estimation system further comprises a third acquisition unit configured to acquire third data representing a third characteristic of each of the plurality of second subjects ,
The target data acquisition unit is configured to acquire the action history data of each of the plurality of second target persons as the target data,
The estimation unit estimates the first feature of each of the plurality of second subjects using the estimation model based on the target data, and uses the plurality of pieces as extended data for the third data. An estimation system that outputs data representing each of the first feature and the third feature of each of the second subjects . A first acquisition unit configured to acquire first data representing a plurality of first subjects having characteristics common to the presence or absence of execution of the specific action;
A second acquisition unit configured to acquire second data representing a feature relating to an action involving communication different from the specific action of each of the plurality of first target persons determined from the first data. When,
For estimating the possibility of performing the specific action from the feature related to the action involving communication based on the feature related to the action involving communication of each of the plurality of first subjects determined from the second data. A model builder configured to build an estimation model;
A target data acquisition unit configured to acquire target data representing characteristics relating to the behavior involving the communication of the second target person ;
An estimation unit configured to estimate the likelihood that the second target person will perform the specific action using the estimation model based on the target data;
Equipped with a,
The second acquisition unit is configured to store, as the second data, action histories related to each of the plurality of first target persons from a database system that stores action history data related to actions involving each of the plurality of target persons. Get the data,
The target data acquisition unit is configured to acquire the behavior history data related to the second target person from the database system as the target data,
The action involving the communication is an action in which information related to the action of the corresponding target person is transmitted to the database system by communication, and the action of the corresponding target person is recorded in the action history data based on the related information. der Ru estimation system. In order for a computer to function as the first acquisition unit, the second acquisition unit, the model construction unit, the target data acquisition unit, and the estimation unit in the estimation system according to any one of claims 1 to 8. Computer program. An estimation method performed by a computer,
Representing a first feature of each of a plurality of first subjects , or obtaining first data representing a plurality of first subjects with a common first feature,
Obtaining second data representing a second feature different from the first feature of each of the plurality of first subjects ;
Based on the first feature and the second feature of each of the plurality of first subjects determined from the first data and the second data, the second feature to the first feature Building an estimation model for estimating features;
And obtaining the target data representing the second feature of the second subject,
Estimating the first feature of the second subject using the estimation model based on the target data;
Only including,
The second feature is a feature related to an action involving communication,
Acquiring the second data includes, from the database system storing action history data related to the actions involving the communication of each of a plurality of subjects, as the second data of the plurality of first subjects. Is to get action history data corresponding to each,
Acquiring the target data is acquiring the behavior history data related to the second target person from the database system as the target data,
The action involving the communication is an action in which information related to the action of the corresponding target person is transmitted to the database system by communication, and the action of the corresponding target person is recorded in the action history data based on the related information. Is an estimation method. An estimation method performed by a computer,
Obtaining first data representing a plurality of first subjects having characteristics common to the presence or absence of execution of a specific action;
Obtaining second data representing characteristics relating to behavior involving communication, which is different from the specific behavior of each of the plurality of first subjects determined from the first data;
For estimating the possibility of performing the specific action from the feature related to the action involving communication based on the feature related to the action involving communication of each of the plurality of first subjects determined from the second data. Building an estimation model;
Obtaining target data representing characteristics relating to the behavior involving the communication of the second target person ;
Based on the target data, using the estimation model, estimating the possibility that the second target person performs the specific action;
Only including,
Acquiring the second data includes, from the database system storing action history data related to the actions involving the communication of each of a plurality of subjects, as the second data of the plurality of first subjects. To get action history data for each,
Acquiring the target data is acquiring the action history data related to the second target person as the target data from the database system;
The action involving the communication is an action in which information related to the action of the corresponding target person is transmitted to the database system by communication, and the action of the corresponding target person is recorded in the action history data based on the related information. Is an estimation method.

Images