JP7606949B2 - Causal Discovery Device - Google Patents

Description

The present disclosure relates to a causal exploration device that performs causal exploration between attributes of target customers. Note that causal exploration between attributes refers to exploring (estimating) causal relationships between attributes, and a causal relationship between attributes refers to a relationship between two attributes in which one is the cause and the other is the effect.

In recent years, it has become possible for customers to carry out the same contract procedures on a web page (hereinafter also referred to as a "WEB page") as they did in a store. However, even if a customer who needs to carry out a contract procedure visits a WEB page, they do not necessarily carry out the contract procedure on the WEB page, and may leave the WEB page and carry out the contract procedure at a real store. In order to determine the reason why a customer leaves a WEB page in this way, it is effective to extract attributes unique to customers who leave a WEB page and carry out the contract procedure at a real store, and analyze the causal relationship between these unique attributes. When analyzing such causal relationships, a technique called causal discovery (also called cause discovery) is used (Patent Document 1).

Patent No. 6154523

However, in the above causal discovery process, if the number of customer attributes to be processed increases, the processing load becomes very high, which can result in inconveniences such as the causal discovery process not completing within the expected time. On the other hand, simply reducing the number of customer attributes to be processed in the causal discovery process raises concerns that the accuracy of the causal discovery process may decrease.

The present disclosure has been made to solve the above-mentioned problems, and aims to reduce the load of causal discovery processing by reducing the number of customer attributes that are the subject of a single causal discovery process while maintaining the accuracy of the causal discovery process.

The causal exploration device according to the present disclosure extracts, from a plurality of customers who have completed a certain procedure, a first group of customers who visited the specific procedure location and completed the procedure at the specific procedure location, and a second group of customers who visited the specific procedure location but completed the procedure at a location other than the specific procedure location, based on information regarding the procedure completion location for each customer and information regarding whether or not the specific procedure location was visited, and stratifies the second group of customers into a plurality of layers determined along the path of movement until the procedure is completed at the specific procedure location, based on the visit history information for each customer; and The system further includes an attribute extraction unit that acquires attribute information of each of the customers in the conversion layer, and extracts, for each layer of the second customer group, correlation attributes that are not correlated with the attributes of customers in the conversion layer but are correlated with the attributes of customers in any layer of the second customer group, based on the feature values for each attribute of each layer of the second customer group and the conversion layer obtained based on the acquired attribute information; and a causal search unit that performs a causal search between the correlation attributes for each layer of the second customer group, based on the correlation attributes for each layer of the second customer group extracted by the attribute extraction unit and the feature values for each attribute of each layer of the second customer group and the conversion layer.

The "conversion layer" mentioned above refers to a layer of customers who have converted from "prospective customers" to "customers" by completing a procedure at a specific processing location, for example.

In the above causal search device, the stratification unit extracts a first group of customers who visited a specific procedure location and completed the procedure at the specific procedure location and a second group of customers who visited a specific procedure location but completed the procedure at a location other than the specific procedure location from among a plurality of customers who completed a certain procedure, based on information on the procedure completion location for each customer and information on whether or not the specific procedure location was visited, and stratifies the second group of customers into a plurality of layers determined along the path of movement until the procedure is completed at the specific procedure location, based on the visit history information for each customer. As a result, the second group of customers is stratified (grouped) into a plurality of layers determined along the path of movement until the procedure is completed at the specific procedure location. Next, the attribute extraction unit acquires attribute information for each layer of the second customer group and for the customers of the conversion layer constituting the first customer group, and extracts correlation attributes for each layer of the second customer group that are not correlated with the attributes of the customers of the conversion layer but are correlated with the attributes of any layer of the second customer group, based on the feature amounts for each attribute of each layer of the second customer group and the conversion layer obtained based on the acquired attribute information. As a result, based on the attribute information of each layer of the second customer group and the conversion layer, correlated attributes (attributes that are not correlated with the attributes of customers in the conversion layer but are correlated with the attributes of customers in any layer of the second customer group) that will be the subject of the subsequent "causal search" are appropriately extracted for each layer of the second customer group. Furthermore, the causal search unit performs a causal search between correlated attributes for each layer of the second customer group based on the extracted correlated attributes for each layer of the second customer group and the feature amounts for each attribute of each layer of the second customer group and the conversion layer. As described above, a causal search is performed for each layer of the second customer group using the correlated attributes appropriately extracted for each layer of the second customer group as the target, so that the accuracy of the causal search process can be maintained at a certain level or higher, and the load of the causal search process can be reduced by reducing the number of customer attributes that are the subject of a single causal search process.

According to the present disclosure, it is possible to reduce the load of causal discovery processing by reducing the number of customer attributes that are the subject of a single causal discovery process while maintaining the accuracy of the causal discovery process.

1 is a functional block diagram of a causal exploration device according to an embodiment of the present invention. (a) is a table showing an example of information regarding the location where each customer completed a procedure, (b) is a table showing an example of web visit history information for each customer, and (c) is a table showing an example of attribute information for each customer. FIG. 4 is a flow chart showing the processing contents executed in the causal exploration device. FIG. 13 is a flow diagram showing a subroutine of a process for extracting attributes correlated with a layer. FIG. 13 is a diagram for explaining extraction of group A (second customer group) and group B (first customer group). FIG. 13 is a diagram for explaining acquisition of WEB visit history information for each group. FIG. 13 is a diagram for explaining determination of stratification coefficients. FIG. 13 is a diagram for explaining the stratification of group A. FIG. 11 is a diagram for explaining generation of a layer information table. FIG. 13 is a diagram for explaining generation of an attribute table with layer information. FIG. 11 is a diagram for explaining derivation of a feature amount for each attribute. FIG. 13 is a diagram for explaining calculation of the proportion of relevant people for each attribute. FIG. 13 is a diagram for explaining extraction of a correlation attribute. FIG. 13 is a diagram for explaining the exclusion of attributes with too few corresponding people. FIG. 13 is a diagram illustrating an example of correlation attributes of each layer. FIG. 13 is a diagram for explaining causal discovery. FIG. 13 is a diagram for explaining generation and output of a causal graph. FIG. 2 is a diagram illustrating an example of a hardware configuration of a causal exploration device.

One embodiment of the causal exploration device will be described with reference to the attached drawings. In the following, the "procedure" will be described using an example of a contract procedure for a certain target service, and the contract procedure by the customer can be completed either online or at an actual store.

As shown in FIG. 1, a causal exploration device 10 according to one embodiment includes a customer data storage unit 11, a stratification unit 12, an attribute extraction unit 13, a causal exploration unit 14, and a graph generation and output unit 15. The functions of each unit are described below.

The customer data storage unit 11 is a database that stores, for each customer, information regarding the location where the procedure was completed, information regarding whether the group of procedure web pages was visited, information regarding which pages were visited among the group of procedure web pages, and customer attribute information, as described below. The above various types of information are acquired in advance from the terminal logs, application logs, subscriber information, etc. of customers who have given consent to the acquisition of such information, and are stored by the customer data storage unit 11.

The above various information may be stored in a variety of forms, but an example will be described using Figures 2(a) to 2(c). As shown in Figure 2(a), information regarding the location where the procedure was completed for each customer and information regarding whether the group of procedure web pages was visited are stored by the customer data storage unit 11 using customer identification information (customer ID) as a key, along with information regarding the time of the event (here, the time the contract procedure was completed). The above "information regarding the location where the procedure was completed" is recorded as "WEB" if the customer completed the contract procedure on the web, for example, and as "REAL" if the customer completed the contract procedure at an actual store. The above "information regarding whether the group of procedure web pages was visited" is recorded as "1" if there was a visit, and as "0" if there was no visit. The "procedure WEB page group" here includes various WEB pages that are expected to be visited in a series of steps when a customer performs a contract procedure for a target service, such as a page related to the price plan of the target service, a page related to comparing price plans, a contract procedure start page, a page during the contract procedure, a contract procedure completion page, and a page related to FAQs (Frequently Asked Questions). The information shown in FIG. 2(a) is collectively referred to as procedure-related information 11A. As shown in FIG. 2(b), the visited page information for each customer is stored by the customer data storage unit 11 with the customer ID as a key together with information on the event occurrence time (here, the time when the corresponding WEB page was visited). The information shown in FIG. 2(b) is collectively referred to as visit-related information 11B. As shown in FIG. 2(c), the attribute information for each customer 11C is stored by the customer data storage unit 11 with the customer ID as a key, in columns for various attributes, and the contents of the attribute or a flag indicating whether the attribute is present or not. For example, for attributes A and B, a flag "1" indicating that the attribute is present or a flag "0" indicating that the attribute is not present is stored, and for attribute C, the contents of attribute C, such as YYY, ZZZ, or XXX, are stored. The "event occurrence time" information shown in Figures 2(a) and (b) is used to determine whether the corresponding contract procedure completion and web page visit were performed as a series of actions. For example, if the time of the web page visit and the time of contract procedure completion are separated by a predetermined time interval or more, it is determined that they were not performed as a series of actions.

Returning to Figure 1, the stratification unit 12 is a functional unit that extracts, from a plurality of customers who have completed the contract procedure for a certain target service, a first customer group (hereinafter referred to as "Group B") consisting of customers who visited the procedure web page group and completed the procedure there, and a second customer group (hereinafter referred to as "Group A") consisting of customers who visited the procedure web page group but completed the procedure outside the procedure web page group, based on information regarding the location where each customer completed the procedure and information regarding whether or not the procedure web page group was visited, and stratifies the customers belonging to Group A (hereinafter referred to as "Group A") into a plurality of layers determined along the path of movement until the procedure is completed on the procedure web page group, based on the visit history information for each customer. Although details will be described later, the stratification unit 12 obtains the movement lines of customers belonging to group B (hereinafter abbreviated as "group B") until they complete the procedures on the group of procedure web pages, and determines multiple layers (hereinafter also referred to as "stratification variables") for stratifying group A based on both the obtained movement lines of group B and a predetermined reference movement line that serves as a standard, and uses these multiple layers to stratify group A. However, it is not essential to use the movement lines of group B as the basis as described above, and the stratification variables may be determined based only on the predetermined reference movement line. Also, the process of determining the stratification variables may be omitted, and multiple layers corresponding to the predetermined reference movement line may be used to stratify group A.

The attribute extraction unit 13 is a functional unit that acquires attribute information of each layer of customers in group A stratified by the stratification unit 12 and customers in the conversion layer (hereinafter also referred to as "layer CV"), which is a layer constituting group B extracted by the stratification unit 12, and extracts correlated attributes for each layer of group A that are not correlated with the attributes of customers in the conversion layer but are correlated with the attributes of customers in any layer of group A based on the feature values for each attribute of each layer of group A and the conversion layer obtained based on the acquired attribute information. As will be described in detail later, the attribute extraction unit 13 counts the number of customers having attributes related to the attribute information of each layer of group A and the conversion layer, and acquires the obtained number of customers (counted value) as the feature value of each attribute. At this time, in addition to using the number of customers obtained by the counting as the feature value as it is, the result of a predetermined calculation such as multiplying the number of customers by a predetermined coefficient may be acquired as the feature value. In addition, the attribute extraction unit 13 excludes attributes for which the number of customers having a certain attribute is below a predetermined lower limit from the correlated attributes. An example of such attribute extraction processing will be described in detail later.

The causal search unit 14 is a functional unit that performs a causal search between correlated attributes for each layer of group A based on the correlated attributes for each layer of group A extracted by the attribute extraction unit 13 and the feature quantities for each attribute of each layer of group A and the conversion layer.

The graph generation and output unit 15 is a functional unit that generates and outputs a directed acyclic graph (DAG) representing the causal search results from the layer-by-layer causal search results obtained by the causal search unit 14.

Next, the processing executed by the causal exploration device 10 will be described in order according to the flow diagram in FIG. 3. The processing in FIG. 3 may be started at a predetermined periodic timing, for example, or may be started in response to a predetermined start operation or the like being performed by the operator of the causal exploration device 10.

The stratification unit 12 obtains the procedure-related information 11A in FIG. 2(a) related to multiple customers who have completed the contract procedure for the target service (hereinafter referred to as "procedure-completed customers") by reading it from the customer data storage unit 11 (step S1), and extracts Group A and Group B from the procedure-completed customers based on the procedure-related information 11A as follows (step S2). That is, as shown in FIG. 5, the stratification unit 12 extracts customers (customers with customer ID 0000001) whose procedure-completed location in the procedure-related information 11A is "REAL (such as a real store other than the WEB)" and whose visit information to the procedure WEB page group is "1 (visited)" as Group A, and extracts customers (customers with customer ID 0000010) whose procedure-completed location is "WEB" and whose visit information to the procedure WEB page group is "1 (visited)" as Group B. In addition, customers other than those mentioned above, i.e., customers whose procedure completion location is "REAL" and whose visit information for the procedure web page group is "0 (no visit)" (customers whose customer ID is 0000002), are excluded from extraction.

Next, the stratification unit 12 reads the visit-related information 11B of FIG. 2(b) for each of the extracted groups A and B from the customer data storage unit 11, thereby obtaining the visit-related information 11B of each of groups A and B shown in FIG. 6 (step S3).

Next, the stratification unit 12 judges whether the stratification variables to be used in the stratification in step S6 described later have been determined (step S4). Here, for example, when the target service of interest repeatedly executes the same causal search process (FIG. 3) continuously, and the stratification variables determined in the first process can be reused in the second and subsequent processes, it is judged that the stratification variables to be used have been determined, and in other cases, it is judged that the stratification variables to be used have not been determined. If it is judged that the stratification variables have been determined in step S4, the process proceeds to step S6 described later, whereas if it is judged that the stratification variables have not been determined, the stratification variables are determined as follows (step S5). As shown in FIG. 7, the WEB visit history information of group B is applied to the main traffic line (framework) determined in advance. Here, a framework of layers A, B, C, D, and CV is predefined as the main flow line, and each layer is determined for each category of web pages, such as layer A being a layer of pages related to price plans, layer B being a layer of pages comparing price plans, layer C being a layer of pages starting the web procedure, layer D being a layer of pages during the web procedure, and layer CV being a layer of pages completing the web procedure. Pages other than the above (e.g., pages related to FAQs) are not included in the main flow line (framework) and are not considered in the following process of determining stratification variables. The left side of FIG. 7 illustrates the visit history information of customers with customer IDs "0000010" and "0000011" belonging to group B, and in the example of visited pages, "aa, aaa" are pages related to price plans, "bbb" is a page comparing price plans, "ccc" is a page starting the web procedure, "dd, ddd" are pages during the web procedure, and "cvpage" is a page completing the web procedure. At this time, as shown in FIG. 7, the flow line of the customer with customer ID "0000010" until the contract is completed is determined to be "Layer A →Layer D →Layer CV," and the flow line of the customer with customer ID "0000011" until the contract is completed is determined to be "Layer A →Layer B →Layer C →Layer D →Layer CV." The flow line "Layer A →Layer B →Layer C →Layer D →Layer CV," which comprehensively includes the flow lines of the customers belonging to group B as described above, is derived as the flow line of group B until the contract is completed, and five stratification variables are determined: "Layer A, Layer B, Layer C, Layer D, Layer CV."

Next, the stratification unit 12 stratifies group A as follows (step S6) using the stratification variables determined in step S5 or already determined. At this time, under the rule that if a customer visits multiple layers, the layer closest to the layer CV is used, as shown in FIG. 8, a customer with a customer ID of "0000001" visits only pages assigned to layer A, and is therefore stratified into layer A. A customer with a customer ID of "0000003" visits pages assigned to layer B, which is the page closest to the layer CV, and is therefore stratified into layer B. Similarly, a customer with a customer ID of "0000004" is stratified into layer C, and a customer with a customer ID of "0000005" is stratified into layer D.

Next, the stratification unit 12 generates a layer information table shown on the right side of Figure 9 in which the stratification results for each customer (stratified layer information) are recorded in association with the customer ID (step S7), and passes it to the attribute extraction unit 13.

Next, the attribute extraction unit 13 obtains the customer attribute information 11C of FIG. 2(c) by reading it from the customer data storage unit 11 (step S8), and generates an attribute table with layer information shown in FIG. 10 by matching the layer information table (FIG. 9) received from the stratification unit 12 with the customer attribute information (step S9). As shown in FIG. 10, the attribute table with layer information stores information about the layer to which the customer belongs and various attributes that the customer possesses, using the customer ID as a key.

Next, the attribute extraction unit 13 executes an extraction process to extract "correlated attributes" for each layer of Group A, which are attributes that are not correlated with the attributes of customers in Layer CV but are correlated with the attributes of customers in any layer of Group A (Step S10). The extraction process in Step S10 is represented by the subroutine in FIG. 4. First, the attribute extraction unit 13 tallies the total number of people belonging to each layer and the number of people having each attribute for each layer, Layers A to D and Layer CV, and sets the number of people having each attribute (tally) as the feature amount of that attribute in that layer (Step S10A). As a result, the total number of people belonging to each layer and the number of people having each attribute (feature amount of that attribute) can be obtained for each layer, as shown in FIG. 11.

Next, the attribute extraction unit 13 calculates the percentage of people with each attribute for each stratum by substituting numerical values into the formula shown in the upper right of FIG. 12 (step S10B). For example, for attribute A in stratum A, the percentage of people is (9000/13800), which is 65.2%. In this way, the percentage of people for each attribute for each stratum is calculated as shown in FIG. 12.

Next, the attribute extraction unit 13 compares the proportion of people with each attribute in layers A to D with the proportion of people with each attribute in the layer CV, and extracts as correlated attributes those attributes in layers A to D whose proportion of people is more than twice the proportion of people in the layer CV (step S10C). For example, as shown in FIG. 13, for "attribute A:1", the proportion of people in layer A (34.8%) and the proportion of people in layer B (27.7%) are more than twice the proportion of people in the layer CV (13.2%), so "attribute A:1" is extracted as a correlated attribute correlated with layers A and B. Similarly, for "attribute B:0", the proportion of people in layers A to D is more than twice the proportion of people in the layer CV (1.3%), so "attribute B:0" is extracted as a correlated attribute correlated with layers A to D. Note that here, the condition used for extracting correlated attributes is "the proportion of people is more than twice the proportion of people in the demographic CV," but this condition is just one example, and other conditions such as "the proportion of people is more than five times the proportion of people in the demographic CV" may also be used.

Furthermore, the attribute extraction unit 13 excludes attributes for which the number of people having the attribute is below a predetermined lower limit from the correlation attributes extracted in step S10C (step S10D). Here, as an example, the attribute extraction unit 13 excludes attributes for which the number of people having the attribute is 1% or less of the total number of people in the layer from the extracted correlation attributes. "Attribute B:0" for layer D shown in FIG. 14 is excluded from the correlation attributes because the number of people having the attribute, 61, is 1% or less of the total number of people in layer D, which is 1149. Note that the condition here, "the number of people having the attribute is 1% or less of the total number of people in the layer" is just an example, and another condition, such as "the number of people having the attribute is 3% or less of the total number of people in the layer", may be used.

By the above process of FIG. 4, among the attributes related to each of layers A to D, the attributes shown in FIG. 15 are extracted as correlated attributes (attributes that are not correlated with the layer CV but are correlated with the layer and have a sufficient number of people who have the attribute). Note that FIG. 15 also shows correlated attributes related to attributes C to G that were not shown in FIGS. 11 to 14.

Next, the causal search unit 14 executes a causal search for each layer (step S11). For example, referring to FIG. 16, the causal search for layer A will be explained. The causal search unit 14 executes a causal search for a total of four attributes, namely, "Layer A: 1", "Layer B: 0", "Layer C: XXX", which are correlation attributes related to layer A, plus "Layer: A". As a result, out of the total of 12 candidate causal relationships between attributes shown in the lower center of FIG. 16, a total of six causal relationships shown in the lower right of FIG. 16 are obtained as a result of the causal search.

Next, the graph generation and output unit 15 generates a directed acyclic graph (DAG) representing the causal search results from the causal search results for each layer obtained by the causal search (step S12) and outputs the DAG (step S13). For example, for layer A, as shown in FIG. 17, a directed acyclic graph representing a total of six causal relationships obtained by the causal search is generated and output.

According to one embodiment of the causal search device 10 described above, causal search is performed for each layer of Group A using correlated attributes appropriately extracted for each layer of Group A based on the features for each attribute of each layer and layer CV of Group A. This makes it possible to maintain the accuracy of the causal search process at a certain level or higher, and also to reduce the load of the causal search process by reducing the number of customer attributes that are the subject of processing in one causal search process.

In addition, since the causal search is performed for each layer of group A, the number of customer attributes that are the subject of a single causal search process can be reduced, which prevents the directed acyclic graph representing the causal relationships between attributes from becoming huge, and makes it possible to output a graph that can determine the causal relationships between attributes.

In the above embodiment, the "procedure" is an example of a contract procedure for a certain target service that can be performed both online and at a physical store. In this case, the results of the causal search can be effectively used to analyze dropout locations and customer attributes in the path leading up to the completion of the procedure for Group A (a group of customers who visited the procedure web pages but completed the procedure on a web page other than the procedure web pages).

In addition, the stratification unit 12 determines multiple layers (stratification variables) for stratifying group A based on a predetermined reference traffic line, which serves as a standard. This prevents, for example, visit histories to "reference information pages that are not directly related to procedures," such as FAQ pages, from being used as the basis, and allows appropriate stratification variables to be set based on the predetermined reference traffic line.

The stratification unit 12 also determines multiple layers (stratification variables) for stratifying group A based on the movement of group B until the procedure is completed on the group of procedure web pages in addition to the predetermined standard movement line, and can determine more appropriate stratification variables that are more in line with the actual situation using the movement line of group B that actually completed the procedure on the group of procedure web pages.

In addition, the attribute extraction unit 13 counts the number of customers who have the attributes related to the attribute information for each layer of group A and the CV layer, and obtains the feature values of each attribute based on the obtained number of customers, making it possible to obtain the feature values of each attribute with relatively simple processing based on the number of customers who have the attribute (counted value).

In addition, the attribute extraction unit 13 can prevent an attribute with an extremely small number of customers from being extracted as a correlated attribute by excluding the attribute with the number of customers having the attribute below a predetermined lower limit from the correlated attributes, thereby optimizing the correlation attribute extraction process.

In addition, in the above embodiment, an example was described in which the causal exploration device 10 is equipped with a customer data storage unit 11 that stores, for each customer, information regarding the procedure completion location, information regarding whether a specific procedure location has been visited, visit history information, and attribute information. In this way, by having the causal exploration device 10 equipped with the customer data storage unit 11 internally, it becomes unnecessary to obtain various pieces of information required when executing the process in FIG. 3 from outside, which contributes to speeding up the process.

(Explanation of terms, explanation of hardware configuration (Fig. 18), etc.)
The block diagrams used to explain the above embodiments and modifications show functional blocks. These functional blocks (components) are realized by any combination of at least one of hardware and software. The method of realizing each functional block is not particularly limited. That is, each functional block may be realized using one device that is physically or logically coupled, or may be realized using two or more devices that are physically or logically separated and directly or indirectly connected (for example, using wires, wirelessly, etc.). The functional block may be realized by combining the one device or the multiple devices with software.

Functions include, but are not limited to, judgement, determination, judgment, calculation, computation, processing, derivation, investigation, search, confirmation, reception, transmission, output, access, resolution, selection, selection, establishment, comparison, assumption, expectation, regarding, broadcasting, notifying, communicating, forwarding, configuring, reconfiguring, allocating, mapping, and assignment. For example, a functional block (component) that performs the transmission function is called a transmitting unit or a transmitter. As mentioned above, there are no particular limitations on the method of realization for either of these.

For example, the causal exploration device in one embodiment of the present disclosure may function as a computer that performs the processing in this embodiment. FIG. 18 is a diagram showing an example of the hardware configuration of a causal exploration device 10 according to one embodiment of the present disclosure. The causal exploration device 10 described above may be physically configured as a computer device including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, etc.

In the following description, the term "apparatus" can be interpreted as a circuit, device, unit, etc. The hardware configuration of the causal exploration device 10 may be configured to include one or more of the devices shown in the figure, or may be configured to exclude some of the devices.

Each function of the causal exploration device 10 is realized by loading a specific software (program) onto hardware such as the processor 1001 and memory 1002, causing the processor 1001 to perform calculations, control communications via the communication device 1004, and control at least one of the reading and writing of data in the memory 1002 and storage 1003.

The processor 1001, for example, runs an operating system to control the entire computer. The processor 1001 may be configured as a central processing unit (CPU) that includes an interface with peripheral devices, a control device, an arithmetic unit, registers, etc.

The processor 1001 also reads out programs (program codes), software modules, data, etc. from at least one of the storage 1003 and the communication device 1004 into the memory 1002, and executes various processes according to these. The programs used are those that cause a computer to execute at least some of the operations described in the above-mentioned embodiments. Although the above-mentioned various processes have been described as being executed by one processor 1001, they may be executed simultaneously or sequentially by two or more processors 1001. The processor 1001 may be implemented by one or more chips. The programs may be transmitted from a network via a telecommunications line.

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

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

The communication device 1004 is hardware (transmitting/receiving device) for communicating between computers via at least one of a wired network and a wireless network, and is also referred to as, for example, a network device, a network controller, a network card, a communication module, etc.

The input device 1005 is an input device (e.g., a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that accepts input from the outside. The output device 1006 is an output device (e.g., a display, a speaker, an LED lamp, etc.) that performs output to the outside. The input device 1005 and the output device 1006 may be integrated (e.g., a touch panel). Each device, such as the processor 1001 and the memory 1002, is connected by a bus 1007 for communicating information. The bus 1007 may be configured using a single bus, or may be configured using different buses between each device.

Each aspect/embodiment described in this disclosure may be used alone, in combination, or switched depending on the execution. In addition, notification of specific information (e.g., notification that "X is the case") is not limited to being done explicitly, but may be done implicitly (e.g., not notifying the specific information).

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

The processing steps, sequences, flow charts, etc. of each aspect/embodiment described in this disclosure may be reordered unless inconsistent. For example, the methods described in this disclosure present elements of various steps using an example order and are not limited to the particular order presented.

The input and output information may be stored in a specific location (e.g., memory) or may be managed using a management table. The input and output information may be overwritten, updated, or added to. The output information may be deleted. The input information may be transmitted to another device.

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

When the terms "include," "including," and variations thereof are used in this disclosure, these terms are intended to be inclusive, similar to the term "comprising." Additionally, the term "or," as used in this disclosure, is not intended to be an exclusive or.

In this disclosure, where articles have been added through translation, such as a, an, and the in English, the disclosure may include that the nouns following these articles are in the plural form.

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

10...causal search device, 11...customer data storage unit, 12...stratification unit, 13...attribute extraction unit, 14...causal search unit, 15...graph generation and output unit, 1001...processor, 1002...memory, 1003...storage, 1004...communication device, 1005...input device, 1006...output device, 1007...bus.

Claims (7)

a stratification unit that extracts, from a plurality of customers who have completed a certain procedure, a first customer group who visited the specific procedure location and completed the procedure at the specific procedure location, and a second customer group who visited the specific procedure location but completed the procedure at a location other than the specific procedure location, based on information regarding the procedure completion location for each customer and information regarding whether the specific procedure location was visited, and stratifies the second customer group into a plurality of layers determined along the path of movement until the procedure is completed at the specific procedure location, based on the visit history information for each customer;
an attribute extraction unit that acquires attribute information of customers in each stratum of the second customer group stratified by the stratification unit and customers in a conversion stratum constituting the first customer group extracted by the stratification unit, and extracts, for each stratum of the second customer group, a correlated attribute that is not correlated with the attributes of customers in the conversion stratum but is correlated with the attributes of customers in any stratum of the second customer group, based on feature amounts for each attribute of each stratum of the second customer group and the conversion stratum obtained based on the acquired attribute information;
a causal search unit that performs a causal search between the correlated attributes for each stratum of the second customer group, based on the correlated attributes for each stratum of the second customer group extracted by the attribute extraction unit, and feature amounts for each attribute of each stratum of the second customer group and the conversion stratum;
A causal discovery device comprising: The layering section is
From a plurality of customers who have completed a certain procedure, a first group of customers who visited the procedure web page group and completed the procedure at the procedure web page group, and a second group of customers who visited the procedure web page group but completed the procedure at a page other than the procedure web page group are extracted based on information regarding the location where the procedure was completed for each customer and information regarding whether or not the procedure web page group was visited, and the second group of customers is stratified into a plurality of layers determined along the path of movement up to the completion of the procedure at the procedure web page group, based on the visit history information for each customer.
The causal exploration device according to claim 1 . the stratification unit determines the plurality of tiers for stratifying the second customer group based on a reference traffic line that is a predetermined reference.
The causal exploration device according to claim 2 . the stratification unit determines the plurality of strata based on a flow line of the customers of the first customer group until the customers complete the procedure on the group of procedure web pages;
The causal exploration device according to claim 3 . the attribute extraction unit counts the number of customers having attributes related to the attribute information of each stratum of the second customer group and the conversion stratum, and acquires feature quantities of each attribute based on the obtained number of customers;
The causal exploration device according to any one of claims 1 to 4. the attribute extraction unit excludes an attribute, the number of customers having the attribute being below a predetermined lower limit, from the correlated attributes;
The causal exploration device according to claim 5 . a customer data storage unit that stores, for each customer, information on the place where the procedure was completed, information on whether or not the specific procedure place was visited, visit history information, and attribute information;
The causal exploration device according to any one of claims 1 to 6, further comprising:

Images