JP6751960B1 - Information processing system and information processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent a decrease in accuracy of an index value regarding the derivation of an index value relating to the similarity of user behavior, and to suppress an increase in the amount of information regarding the user behavior used for derivation. An information processing server 1 converts information indicating a user's action defined as a series of actions into an action character string in which characters assigned to each action are arranged in the order in which the actions are performed. The conversion unit 11 to be converted, and the derivation unit 12 for deriving the editing distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user. Using information consisting of character strings, which is not missing a part of the information, about what each action was performed by the user and the chronological order in which each action was performed. The editing distance, which is an index value for similarity, is derived. [Selection diagram] Fig. 1

Description

The present invention relates to an information processing system and an information processing method by the information processing system, and is particularly suitable for use in an information processing system and an information processing method that execute processing on information indicating a user's behavior.

Conventionally, the behavior when a user browses a web page on a certain site (called browsing behavior, which is how the browsed web page changes), the behavior taken by the user at a certain store, etc. Systems that analyze user behavior are known. For example, Patent Document 1 describes a system that records a user's browsing behavior based on an access log, analyzes the browsing behavior, and executes various predictions.

Japanese Unexamined Patent Publication No. 2006-323629

When analyzing the behaviors of a plurality of users, it is very effective to derive index values related to similarity for the behaviors of different users. This is because it is possible to determine the similarity between two different behaviors in the analysis, which makes it possible to group and analyze similar behaviors or to categorize and analyze a plurality of behaviors. Then, the derivation of the index value regarding the similarity of the behaviors of different users is performed using the information regarding the behavior of each user, but the amount of information regarding the behavior of the user used when deriving the index value is large. There is a need to prevent a situation in which the accuracy of the index value is lowered due to suppressing the increase in the amount of information. This is because if the amount of information related to the user's behavior becomes large, the processing load increases, and adverse effects such as overflow may occur. Further, it is possible to reduce the amount of information by reducing a part of the information, but in this case, the accuracy of the index value naturally decreases, and such a situation is overwhelming.

The present invention has been made to solve such a problem, and is used for deriving the index value regarding the similarity of user behavior while preventing the accuracy of the index value from being lowered. The purpose is to prevent the amount of information about user behavior from becoming large.

In order to solve the above-mentioned problems, in the present invention, an action in which information indicating a user's action defined as a series of actions is a character string in which characters given for each action are arranged in the order in which the actions are performed. Convert to a string. Then, the editing distance between the action character string for the action of one user and the action character string for the action of another user is derived.

In the present invention configured as described above, the converted action character string corresponds to information on the user's behavior used for deriving the editing distance, which is an index value regarding the similarity of the user's behavior. This action string is information that expresses the user's actions without missing a part of the information about what each action was performed by the user and the chronological order in which each action was performed. I can say. Then, in the present invention, since the index value (editing distance) relating to the similarity of the user's behavior is derived by using such an action character string, the index value (editing distance) due to the lack of a part of the information is derived. It is possible to prevent a decrease in the accuracy of. Further, the action character string is information composed of characters representing actions performed by the user arranged side by side, information representing each action is very small, and information related to an action not performed by the user. The amount of information is small because it does not include. That is, according to the present invention, it is possible to suppress an increase in the amount of information regarding the user's behavior.

It is a figure which shows the functional configuration example of the information processing server which concerns on one Embodiment of this invention, and the configuration example of the control system including the information processing server. It is a figure which shows the content of various data stored in a storage part. It is a figure which shows an example of a network. The horizontal axis is the order non-viewing ratio, and the vertical axis is the figure in which the web page is plotted in the two-dimensional space of the ordered viewing ratio. It is a figure used to explain the action of the action modeling part. It is a figure which shows an example of the information provision page. It is a flowchart which shows the operation example of the information processing server which concerns on one Embodiment of this invention.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing a system configuration of a control system 2 including an information processing server 1 and a plurality of related devices according to the present embodiment together with a block diagram showing a functional configuration example of the information processing server 1. The information processing server 1 corresponds to the "information processing system" in the claims. As shown in FIG. 1, the control system 2 includes an information processing server 1, a web server 3, and a terminal 4, and these devices are connected via a network N.

The web server 3 is a server on which a website has been opened by the website creator. Hereinafter, the website established on the web server 3 is particularly referred to as a "dedicated website". In the present embodiment, for convenience of explanation, the website creator is a company that sells the product I as a business, and the dedicated website advertises the product I and orders the product I online. It is assumed that the site can be used. That is, the dedicated website is a site that induces the user to perform a reaction of ordering the product I. A dedicated website is composed of a plurality of web pages. In the present embodiment, the dedicated website is configured to include at least seven web pages of pages H1 to H7. In the present embodiment, for convenience of explanation, the processing executed by the information processing server 1 for one website (dedicated website) will be described, but in reality, the information processing server 1 has a plurality of information processing servers 1. Perform the process described below for the website.

The terminal 4 is a computer equipped with a web browser. The terminal 4 may be in any form as long as it is a computer equipped with a web browser. The terminal 4 is, for example, a smartphone, a tablet computer, a notebook PC, or a desktop PC. FIG. 1 schematically shows one of the innumerable terminals 4. In the following description, a person who browses a dedicated website by the terminal 4 is referred to as a "user".

As shown in FIG. 1, the information processing server 1 includes a communication unit 10, a conversion unit 11, a derivation unit 12, a cluster formation unit 13, a behavior modeling unit 14, and an information providing unit 15 as functional configurations. Each of the above functional blocks 10 to 15 can be configured by any of hardware, DSP (Digital Signal Processor), and software. For example, when configured by software, each of the functional blocks 10 to 15 is actually configured to include a computer CPU, RAM, ROM, etc., and is a program stored in a recording medium such as RAM, ROM, hard disk, or semiconductor memory. Is realized by the operation of. Further, the information processing server 1 includes a storage unit 17 as a storage means. The contents of the data stored in the storage unit 17 will be described later. Although the information processing server 1 is represented by one block in FIG. 1, this does not mean that the information processing server 1 is composed of a single server device. The information processing server 1 may be composed of a plurality of devices or may be a part of a predetermined system.

The communication unit 10 communicates with a device connected to the network N according to a predetermined communication standard.

The information processing server 1 analyzes the browsing behavior of the dedicated website by the user based on the data stored in the storage unit 17, and further provides useful information on the browsing behavior to the website creator based on the analysis result. .. Hereinafter, the operation of the information processing server 1 from the analysis of browsing behavior to the provision of information will be described in detail.

Prior to the explanation of the operation of the information processing server 1, the contents of the browsing history data 20 and the series browsing behavior data 21 will be described first. FIG. 2A is a diagram showing an example of the contents of the browsing history data 20. The browsing history data 20 is relational database type data having one or more records, and as shown in FIG. 2A, each record stores an action ID field in which an action ID is stored and a page ID. It has a page ID field to be displayed, a browsing order field in which browsing order information indicating the browsing order is stored, and a browsing time field in which browsing time information indicating the browsing time is stored.

The action ID is an ID uniquely assigned to a series of browsing actions. A series of browsing behavior means that the user starts browsing one of the web pages in the dedicated website and then transitions the web page in the dedicated website (however, the transition of the web page is performed even once). It is not necessary), it is a series of browsing actions until the end of browsing the web pages in the dedicated website, and it is grasped by the order in which each web page was browsed and the browsing time when each web page was browsed. .. When the browsing of the web page in the dedicated website ends, for example, the transition to the web page of the website other than the dedicated website, or the web browser is closed and the display of the web page of the dedicated website ends. Means that.

In the present embodiment, in a series of browsing actions, a continuous action from the start of browsing a certain page to the end of browsing the one page is defined as a "single browsing action". A series of browsing behaviors consists of one or more single browsing behaviors. In addition, a series of browsing behaviors may include a plurality of single browsing behaviors related to browsing the same web page. For example, when a user transitions a web page from page H1 to page H2 to page H1, browsing of the first page H1, browsing of the second page H2, and browsing of the third page H1 are performed. Each corresponds to a "single browsing action", and a series of browsing actions is composed of three single browsing actions.

The page ID is identification information given to each of the web pages of the dedicated website, and in the present embodiment, the page IDs of h1 to h7 are given to the pages H1 to H7, respectively. The browsing order means the order (number number) of the corresponding single browsing actions in the series of browsing actions. The browsing time means the length of time that the corresponding web page is browsed in the corresponding single browsing behavior.

In the browsing history data 20, a group of records having a common action ID value indicates a series of browsing actions by the user. In addition, each record shows a single browsing behavior. For example, pay attention to the five records from the top of the browsing history data 20 illustrated in FIG. 2A, and since these five records have the same action ID value, it may indicate a series of browsing actions. I understand. Then, from the value of the page ID field and the value of the browsing order field of each record, it can be seen that the web page has transitioned in the order of page H1 → page H2 → page H1 → page H2 → page H3, and further, browsing of each record. The value of the time field indicates that the user has viewed pages H1, H2, H1, H2, and H3 in this order for 20, 30, 40, 50, and 60 minutes.

FIG. 2B is a diagram showing an example of the contents of the series browsing behavior data 21. The series browsing action data 21 is relational database type data having one or more records, and as shown in FIG. 2B, each record has an action ID field in which the action ID is stored and order presence / absence information. It has an order presence / absence field in which is stored, an action character string field in which an action character string (described later) is stored, and a cluster ID field in which a cluster ID (described later) is stored. The order presence / absence information is information indicating whether or not an order for the product I has been placed in the corresponding series of browsing actions. For example, the first record of FIG. 2B shows that the order for the product I was placed in the series of browsing actions of the action ID: A1. Before the conversion process is performed by the conversion unit 11, the value of the action character string field and the value of the cluster ID field of each record are null values.

The information processing server 1 appropriately generates records of the browsing history data 20 and the series browsing behavior data. The information processing server 1 may generate records in these data by any method. As an example, a script (so-called tag) is incorporated in an HTML file in which the web server 3 responds to an HTTP request from the web browser of the terminal 4. Then, by the function of this script, the information necessary for generating the records of the browsing history data 20 and the series browsing behavior data 21 is appropriately transmitted from the web browser to the information processing server 1, and the information processing server 1 sends the information processing server 1 to the web. Generate records for this data based on the information received from the browser. Further, in the present embodiment, the information processing server 1 is configured to generate the records of the browsing history data 20 and the series browsing behavior data 21, but the information processing server is configured to perform this by a server other than the information processing server 1. 1 may be configured to receive data provided from this server.

The conversion unit 11 executes a conversion process for each record group (hereinafter referred to as “series record group”) having a common action ID value for the browsing history data 20 to generate an action character string, and the continuous browsing action data 21 Register with. Hereinafter, a series of records to be processed is referred to as a process target record group, and a conversion process executed by the conversion unit 11 for the process target record group will be described. In the example of FIG. 2A, the record group G1 composed of the five records RA1 to RA5 is one of a series of records.

Here, one letter of the alphabet is assigned in advance to each of the web pages included in the dedicated website. Note that the value of one alphapet character assigned to each web page is different from each other, and the alphabet of the value common to different web pages is not assigned. In the present embodiment, the letters "A", "B", "C", "D", "E", "F", and "G" appear on pages H1, H2, H3, H4, H5, H6, and H7, respectively. Is assigned.

In the conversion process, the conversion unit 11 generates an action character string in which the characters assigned to the web pages browsed in each single browsing action are arranged according to the browsing order for the series of browsing actions indicated by the processing target record group. At that time, the conversion unit 11 makes the characters assigned to the web page browsed in the single browsing action continuous so that the longer the browsing time of the single browsing action is, the larger the number of consecutive times is. In the present embodiment, when the browsing time exceeds "0 minutes" and is "10 minutes" or less, the number of consecutive times is "1 time", and when the viewing time exceeds "10 minutes" and is "20 minutes" or less, the number of consecutive times is "2 times". , "20 minutes" and "30 minutes" or less, the number of consecutive times is "3 times", and so on. When "10 (k-1) <viewing time ≤ 10k", the number of consecutive times is " It is said to be "k times".

Hereinafter, the conversion process will be described in more detail assuming that the record group to be processed is the record group G1 of FIG. 2 (A) as appropriate. First, the conversion unit 11 specifies the character to be converted and the number of repetitions for each single browsing behavior indicated by the processing target record group. For example, referring to FIG. 2A, since the page ID of the single browsing action indicated by the record RA1 is "h1", the conversion unit 11 sets the character to be converted to "A" and further, the browsing time is "20". Since it is "minutes", the number of consecutive times is set to "2 times". Similarly, the conversion unit 11 sets the character to be converted to "B" and the number of consecutive times to "3 times" for the second single-unit browsing behavior indicated by the record RA2, and also sets the third single-unit browsing behavior indicated by the record RA3. Regarding the browsing behavior, the character to be converted is "A" and the number of consecutive times is "4 times", and for the fourth single browsing behavior indicated by the record RA4, the character to be converted is "B" and the number of consecutive times is "5 times". In addition, regarding the fifth single browsing action indicated by the record RA5, the character to be converted is set to "C" and the number of consecutive times is set to "6 times".

Next, the conversion unit 11 generates a character string in which the characters determined to be converted are arranged in the viewing order for each of the single browsing actions, and the continuous character strings (which may be one character) for the determined continuous number of times. The character string generated in this way is the "action character string". The action character string for the record group G1 is "AABBBAAAABBBBBCCCCCC".

The action string is information about what the web page was viewed in the single browsing action (= what was each of the actions performed by the user), and the chronological order in which each single browsing action was performed. It can be said that it is information that expresses the user's behavior without missing a part of. In particular, the action character string according to the present embodiment can be said to be information that reflects the browsing time for each of the single browsing actions. Needless to say, the specific method for generating the action character string is not limited to the illustrated method.

The conversion unit 11 performs conversion processing on each of a series of records (a group of records having a common action ID value) included in the browsing history data 20 to generate an action character string. Next, the conversion unit 11 stores each of the corresponding action character strings in the action character string field of each record of the series browsing action data 21.

The derivation unit 12 executes the following processing after the generation of the action character string by the conversion unit 11 and the registration in the list browsing action data 21 are completed. That is, the derivation unit 12 executes the first derivation process for deriving the editing distance between the action character string and the action character string for each combination of the two records of the series browsing action data 21. Therefore, if there are four records of records RX1 to RX4 in the series browsing behavior data 21, the conversion unit 11 will perform RX1 and RX2, RX1 and RX3, RX1 and RX4, RX2 and RX3, RX2 in the first derivation process. And for each of the six combinations of RX4, RX3, and RX4, the edit distance between the action string and the action string is derived.

In the present embodiment, the derivation unit 12 derives a standardized Levenshtein distance as an editing distance between one action character string and another action character string. As is well known, the Levenshtein distance between one character string and another character string is the minimum edit when one character string is edited (inserted / deleted / replaced) and converted to another character string. It is the distance. The derivation and standardization of the Levenshtein distance is carried out by existing algorithms, with the cost of the editing process properly defined. Since the edit distance is derived as described above, the shorter the edit distance derived for one action character string and the other action character string, the higher the similarity between these action character strings, and the longer the edit distance. , The similarity of these action strings is low.

After deriving the edit distance for each combination of all the combinations of the two records, the derivation unit 12 generates and stores the edit distance data 22 recorded in association with the combination of the action IDs for each of the edit distances. Store in part 17. FIG. 2C is a diagram showing an example of the contents of the edit distance data 22. The edit distance data 22 is relational database type data having one or more records, and as shown in FIG. 2C, each record stores one action ID of a combination of two action IDs. It has a first action ID field, a second action ID field in which the other action ID is stored, and an edit distance field in which the edit distance is stored. In the first record of the edit distance data 22 illustrated in FIG. 2C, the edit distance between the action character string corresponding to the action ID: A1 and the action character string corresponding to the action ID: A2 is "0. It shows that it is "5".

After the edit distance data 22 is generated by the derivation unit 12, the cluster formation unit 13 divides each of the series of browsing actions into groups based on the edit distance data 22, and performs a first cluster formation process for forming a plurality of clusters. Execute. More specifically, first, based on the edit distance data 22, the cluster forming unit 13 establishes a network in which each of the series of browsing actions exists as a node and nodes having an edit distance of T1 or less are connected by a link. Derived. FIG. 3 is a diagram schematically showing an example of a network in a mode suitable for explanation. In FIG. 3, white circles indicate nodes, and the information in the white circles indicates action IDs. In the network illustrated in FIG. 3, the node of action ID: A1 and the node of action ID: A2 are connected by a link (thus, the action character string of action ID: A1 and the action character of action ID: A2). The edit distance to the column is less than or equal to the threshold T1), the node of action ID: A1 and the node of action ID: A5 are not connected by a link (hence, the action character string of action ID: A1 and the action ID: The editing distance of A5 from the action character string exceeds the threshold T1).

Next, the cluster forming unit 13 groups the nodes (= series browsing behavior) so as to maximize the modularity, and forms one or more clusters. As is well known, modularity is defined as "the sum of the ratios in a cluster to the total number of links in a network" to "when the points are randomly connected without changing the number of edges coming out of the node, the nodes are connected to the same cluster. It is defined as the value obtained by subtracting the "expected value". By grouping the nodes so as to maximize the modularity, the nodes belonging to a certain cluster (series browsing behavior) have a short editing distance (= high similarity), while the nodes belonging to a certain cluster (= high similarity). A plurality of clusters are formed so that the editing distance between the continuous browsing behavior) and the nodes belonging to the clusters other than the cluster (serial browsing behavior) is long (= low similarity).

After forming the cluster, the cluster forming unit 13 assigns a unique value cluster ID to each of the clusters according to the rules. Then, the cluster forming unit 13 stores the corresponding cluster ID in the cluster ID field of each record of the series browsing behavior data 21. In the present embodiment, the cluster forming unit 13 derives a network having no link orientation or link weight for the network that is a prerequisite for deriving the modularity, but of course, the network in which the orientation and weight are taken into consideration. May be configured to derive. In addition, the method of dividing into clusters is not limited to the method using modularity, and while the series browsing behaviors belonging to a certain cluster have a short editing distance (= high similarity), the series browsing belonging to a certain cluster The cluster may be formed so that the editing distance between the action and the series of browsing actions belonging to clusters other than the cluster is long (= low similarity).

After a plurality of clusters are formed by the cluster forming unit 13, the conversion unit 11 executes the following processing for each of the clusters. Hereinafter, the cluster to be processed by the cluster forming unit 13 is referred to as a “processed cluster”, and the processing of the conversion unit 11 will be described in detail.

The conversion unit 11 analyzes the series of browsing behaviors belonging to the processing target cluster based on the series of browsing behavior data 21, and for each web page, "a series of browsing behaviors in which an order for product I is made (hereinafter," ordered behavior "". The ratio of "the number of series of browsing behaviors in which a web page was browsed" to "the number of browsing behaviors with orders" (hereinafter referred to as "the percentage of browsing with orders") and "the series of browsing behaviors in which product I was not ordered" The ratio of "the number of series of browsing behaviors in which a web page is browsed" to the "number of (hereinafter referred to as" order no behavior ")" (hereinafter referred to as "order no browsing ratio") is derived. It should be noted that whether or not a certain web page is browsed in a certain series of browsing behavior can be determined by whether or not the characters assigned to the web page are included in the character string indicating the series of browsing behavior.

For example, for the processing target cluster, the number of series browsing actions belonging to the processing target cluster is "150", of which the number of series browsing actions for which the product I is ordered is "100", and the product I It is assumed that the number of series browsing actions for which no order has been placed is "50". In this case, assuming that among the series of browsing actions in which the product I is ordered, the series of browsing actions in which the web page H1 is browsed is "80", the conversion unit 11 describes the web page H1. "80%" (= 80/100) is derived as the ordered browsing ratio of. Further, assuming that among the series of browsing actions in which the product I is not ordered, the series of browsing actions in which the web page H1 is browsed is "5", the conversion unit 11 describes the web page H1. "10%" (= 5/50) is derived as the order non-viewing ratio. That is, the ordered browsing ratio for the web page H1 means the ratio of the number of actions for which the web page H1 is browsed to the number of actions for which the product I was ordered. Further, the order non-browsing ratio for the web page H1 means the ratio of the number of actions for which the web page H1 is browsed to the number of actions for which the product I has not been ordered.

FIG. 4 is a diagram showing an example of plotting each of the web pages in a two-dimensional space in which the horizontal axis is the order non-viewing ratio and the vertical axis is the ordered viewing ratio. In FIG. 4, for example, the web page H1 indicated by the point of reference numeral h1 indicates that both the non-viewing ratio of orders and the viewing ratio of ordered items are about "90%", and the web page H2 indicated by the point of reference numeral H2 is It shows that the non-viewing ratio with orders is about "60%" and the viewing ratio with orders is about "20%".

In FIG. 4, in the web page plotted on the straight line S1 having the slope "1" passing through the origin or in the vicinity of the straight line S1, the value of the non-viewing ratio of orders and the value of the browsing ratio with orders are equal to or very close to each other. It is a web page. Hereinafter, such a web page is referred to as an "antagonist page". The following can be said about this antagonistic page. That is, the rate of browsing the antagonistic page in the series of browsing actions that led to the order of product I and the percentage of browsing in the series of browsing actions that did not lead to the order of product I are almost the same (or the same). When the user performs the action (reaction) of "ordering the product I", it can be said that the page that is viewed or not viewed does not have much influence (at least, the influence cannot be determined).

Further, in FIG. 4, in the web page plotted in the region AR1 below the straight line S1 and separated from the straight line S1, the value of the order non-viewing ratio is considerably higher than the value of the ordered viewing ratio. It's a large web page. Hereinafter, such a web page is referred to as a "non-order tendency page". The following can be said about this non-order trend page. That is, the non-order tendency page has a considerably larger proportion of browsing in the series of browsing behaviors that did not lead to the order of product I than the percentage of browsing in the series of browsing behaviors that led to the order of product I. "Not being viewed" has a positive effect on the behavior (reaction) of "ordering I", and "being viewed" has a negative effect on the behavior of "ordering product I". It can be said that this is a page that can be estimated.

Further, in FIG. 4, in the web page plotted in the region AR2 above the straight line S2 and separated from the straight line S1, the value of the ordered browsing ratio is considerably larger than the value of the ordered non-viewing ratio. It is a web page. Hereinafter, such a web page is referred to as an "order tendency page". The following can be said about this order trend page. That is, the order tendency page has a considerably larger proportion of browsing in the series of browsing behaviors that led to the order of product I than the percentage of browsing in the series of browsing behaviors that did not lead to the order of product I. "Being viewed" has a positive effect on the behavior (reaction) of "ordering", and "not being viewed" has a negative effect on the behavior of "ordering product I". Can be said to be a page that can be estimated.

From the above, it is presumed that the non-ordering tendency page and the ordering tendency page have an influence on the reaction of ordering the product I, and the influence is the ratio of the ordered browsing ratio and the order non-viewing ratio. It can be said that the larger the difference, in other words, the larger the plotted points are from the straight line S1 in the two-dimensional space of FIG. Based on this, the conversion unit 11 derives the ordered browsing ratio and the non-ordered browsing ratio for each of the web pages for the series of browsing behaviors belonging to the processing target cluster, and then puts each on the two-dimensional space of FIG. The web pages are plotted, and the distance between each web page and the straight line S1 (the distance of the perpendicular line drawn from the point indicating the web page to the straight line S1; hereinafter referred to as "separation distance") is derived. Then, the conversion unit 11 derives the number of additions according to the separation distance for each web page (the usage mode of the number of additions will be described later). In the present embodiment, a plurality of threshold values are predetermined, and the conversion unit 11 sets the number of additions to "0" when the separation distance is equal to or less than the smallest threshold value, and thereafter, the number of additions each time the threshold value is exceeded. Is increased one by one. As a result, the number of additions of the web page plotted in the vicinity of the straight line S1 in the graph of FIG. 4 becomes "0 times", and the farther away from the straight line S1, the greater the number of additions.

Here, in the present embodiment, the separation distance is derived for each cluster based on the ordered browsing ratio and the non-browsing ratio of each web page, and the number of additions for each web page is derived (hence, a certain common). For the web page to be added, the number of additions may differ between one cluster and another cluster). This is for the following reason. That is, each of the series of browsing behaviors belonging to a certain cluster is basically highly similar to each other. This is because, as mentioned above, each cluster is formed to have the highest modularity. Then, it is decisive whether or not the order tendency page and the non-order tendency page are viewed depending on the behavior in which the product I is ordered and the behavior in which the product I is not ordered in the similar behavior group. It can be strongly estimated that the web pages are different from each other and influence the reaction of ordering the product I. Such a strong estimation cannot be obtained when a disordered behavior group is analyzed in terms of approximation of behavior. Based on this, by deriving the separation distance for each cluster, it is possible to derive a separation distance with high validity and eligibility in that it has a large influence on the behavior of "ordering product I". The validity and eligibility of the number of additions derived based on the above can also be enhanced.

After deriving the number of additions for each of the web pages, the conversion unit 11 accesses the series browsing action data 21 and corrects each action character string of the record corresponding to the series browsing action belonging to the processing target cluster. Give and update the value. Hereinafter, the action character string to be corrected is expressed as a “process target action character string”, and the correction process applied to the process target action character string will be described in detail.

In the correction process, the conversion unit 11 indicates a web page (= single character) indicated by the characters constituting the character group for each of the character groups (which may be one character) indicating the single browsing action in the processing target action character string. The same characters are continued for the number of additional times derived for the web page viewed in the browsing behavior). For example, it is assumed that the action character string before the correction process is "AABBBAAAABBBBBCCCCCC". The first 2 characters "AA", the next 3 characters "BBBB", the next 4 characters "AAAA", the next 5 characters "BBBBBB", and the next 6 characters "CCCCCC" are in the character group indicating the single browsing behavior. Equivalent to. Further, the number of additions derived for the web page H1 indicated by the character "A" is "1 time", the number of additions derived for the web page H2 indicated by the character "B" is "2 times", and the character "C" indicates. It is assumed that the number of additions derived for the web page H3 is "0". In this case, the conversion unit 11 adds a new character string "AA'A'BBB'BB' AAAA'A'BBBBBB'BB'CCCCCC" (however, "for readability" by performing correction processing. To generate). Hereinafter, the action character string that has been corrected is referred to as a “corrected action character string”. When the action character string of all the records belonging to the processing target cluster is corrected, the value stored in the action character string field of all the records becomes the corrected action character string.

Here, as described above, the action character string to which the correction processing is performed is information that reflects the time-series order in which each single browsing action is performed and the browsing time for each of the single browsing actions. .. Then, in addition to these, the correction action character string generated by the correction processing reflects the strength of the influence on the reaction of "purchasing product I" of the web page browsed in the single browsing action. It can be called information.

As described above, the conversion unit 11 derives the number of additions for each of the web pages for each cluster, and corrects the action character string based on the number of additions to generate the correction action character string. The action character string registered in the series browsing action data 21 is updated by the corrected action character string.

The derivation unit 12 corrects all the records of the series browsing action data 21 by the conversion unit 11, and then corrects the correction action character string and the correction action character for all the combinations of the two records of the series browsing action data 21. The second derivation process for deriving the edit distance from the column is executed. The value of the edit distance field of each record of the edit distance data 22 is updated with the newly derived edit distance. Hereinafter, the editing distance between the corrected action character string and the corrected action character string is referred to as "corrected editing distance".

After the correction distance of each record of the edit distance data 22 is updated to the correction edit distance by the derivation unit 12, the cluster formation unit 13 groups each of the series browsing actions based on the edit distance data 22, and a plurality of them. The second cluster formation process for forming a cluster is executed. Hereinafter, the cluster formed by the second cluster formation process is particularly referred to as a "new cluster". The formation of the new cluster is performed by the same method (method using modularity) as the first cluster formation process described above. Next, the cluster forming unit 13 assigns a new cluster ID to each of the new clusters according to the rule, accesses the series browsing behavior data 21, and newly assigns the value of the cluster ID field of each record to the new cluster ID. Update by.

The following can be said for the new cluster formed as described above. That is, the corrected action character string includes the time-series order in which each single browsing action was performed, the browsing time for each single browsing action, and each web page in response to the reaction of "purchasing product I". This is information that reflects the strength of the influence. Therefore, the correction edit distance derived by the second derivation process of the derivation unit 12 has a stronger influence on the reaction of "purchasing product I" than the edit distance (before correction) derived by the first derivation process. The browsing mode of the web page is an index value that has a stronger influence on the value, and in this respect, it is highly valid as an index value for measuring the similarity of behavior. Also, regarding the cluster formed based on the corrected editing distance, which is highly valid in that the browsing mode of the web page, which has a strong influence on the reaction of "purchasing product I", has more influence. It can be said that the validity is high.

The behavior modeling unit 14 executes the behavior modeling process for each of the new clusters after the second cluster formation process is executed by the cluster formation unit 13. Hereinafter, the new cluster to be executed for the behavior modeling process will be referred to as a “new cluster to be processed”, and the behavior modeling process will be described in detail.

In the behavior modeling process, first, the behavior modeling unit 14 refers to the series browsing behavior data 21 and acquires the record corresponding to the ordered behavior among the records of the new cluster to be processed. FIG. 5A shows an example of the action ID and the corrected action character string in the record acquired in this way. Next, the behavior modeling unit 14 derives the maximum common subsequence (so-called LCS) of the correction action character string and the correction action character string for all the combinations of the two records for the acquired record. Derivation of the longest common subsequence is properly performed by existing algorithms. FIG. 5B is a diagram showing the maximum common subsequence derived for each combination of two records for the four records illustrated in FIG. 5A. In FIG. 5B, for example, the longest common subsequence derived for the corrected action character string “ABCDEFG” having the action ID: X1 and the corrected action character string “ABDGC” having the action ID: X2 is “ABDG”. Is shown.

Next, the behavior modeling unit 14 generates a character string by combining each of the derived maximum common subsequences so that the order of the characters in each maximum common subsequence is maintained. Hereinafter, the character string generated in this way is referred to as a "common action character string". FIG. 5C, together with the common action character string “ABCDCEFG” generated based on the six maximum common subsequences shown in FIG. 5B, constitutes the six maximum common subsequences in this common action character string. Indicates where the characters to be used are placed. From FIG. 5C, it can be seen that the common action character string is generated by combining each of the six maximum common character strings while maintaining the order of the characters in each maximum common character string. .. The common action character string generated in this way can be said to be information that abstracts each of the ordered actions among the series of browsing actions belonging to the new cluster to be processed as one action.

In the behavior modeling process, the behavior modeling unit 14 further refers to the series of browsing behavior data 21 and acquires the record corresponding to the order no action among the records of the new cluster to be processed. Next, the behavior modeling unit 14 generates a common behavior character string by the method described above based on the acquired record. As described above, the behavior modeling unit 14 derives the common action character string for the ordered behavior belonging to the new cluster to be processed and the common action character string for the ordered non-action in the behavior modeling process. The behavior modeling unit 14 executes the behavior modeling process for each of the new clusters, and generates a common behavior character string for the ordered behavior and a common behavior character string for the ordered non-action for each of the new clusters.

After generating the common action string for the ordered behavior and the common action string for the ordered non-action for each of the new clusters, the behavior modeling unit 14 determines the new cluster ID of the new cluster and the ordered behavior. The behavior modeling data 23 registered in association with the common behavior character string and the common behavior character string for order non-action is generated and stored in the storage unit 17. FIG. 2D shows an example of the contents of the behavior modeling data 23. As shown in FIG. 2D, one record of the behavior modeling data 23 includes a new cluster ID field in which the new cluster ID is stored and an ordered behavior in which a common action character string for the ordered behavior is stored. It has a field and an order no action field in which a common action string for order no action is stored.

The information providing unit 15 provides useful information regarding browsing of the dedicated website based on the behavior modeling data 23 in response to the request from the site creator. More specifically, the site creator (in the case where the site creator is an organization, the person who belongs to the site creator as an organization) starts the web browser of the predetermined device and activates the predetermined URL of the information processing server 1. To access (HTTP request). The predetermined URL is notified to the site creator in advance, and although details are omitted, when the predetermined URL is accessed, authentication is performed by an appropriate method.

When the predetermined URL of the information processing server 1 is accessed, the information providing unit 15 refers to the behavior modeling data 23 and generates an HTML file for displaying the information providing page 24 (FIG. 6). Next, the information providing unit 15 responds to the device of the access source with the generated HTML file. The web browser of the device displays the information providing page 24 on the display device based on the received HTML file.

FIG. 6 is a diagram showing an example of the information providing page 24. The information providing page 24 illustrated in FIG. 6 has five major items K1 to K5 that can be selectively selected. The major items correspond to each of the new clusters. Therefore, in the example of FIG. 6, there are five new clusters, but if there are ten new clusters, there will be ten major items. In the example of FIG. 6, it is assumed that the major items K1 to K5 correspond to the new clusters whose new cluster IDs are Y1 to Y5, respectively.

Each of the major items is subordinate to two tabs, an order tab 25 and a non-order tab 26, which can be selectively selected. When the order tab 25 of the major item corresponding to a certain cluster is selected, the actions represented by the common action character string for the ordered actions belonging to the cluster are displayed in a diagrammatic manner. Further, when the non-order tab 26 of the large item corresponding to a certain cluster is selected, the action represented by the common action character string for the order no action belonging to the cluster is displayed in a diagrammatic manner. FIG. 6 shows how the order tab 25 of the major item K1 corresponding to the cluster of the new cluster ID: Y1 is selected. Therefore, in FIG. 6, the characters represented by the common action character string for the ordered action belonging to the new cluster with the new cluster ID: Y1 are displayed in a diagrammatic manner. The common action character string for the ordered action belonging to the cluster of the identification information Y1 is "ABCD".

Explaining the display contents of FIG. 6 in more detail, the rod-shaped rod object 28 extending in the left-right direction in FIG. 6 indicates the period during which the action represented by the common action character string is performed, from the left in the figure. Time has passed to the right. The black circle action point AP drawn on the bar object 28 indicates the start of a single browsing action among the list browsing actions indicated by the common action character string, and a balloon 30 is added to each action point AP. In each balloon 30, the title of the web page browsed by the single browsing behavior is described.

In the bar object 28 of FIG. 6, 1 is set at the start position (= left end of the bar object 28) of the first single browsing action (action represented by the first character "A") of the common action character string "ABCD". The second action point AP1 is drawn, and the title of the web page H1 to which the character "A" is assigned is described in the balloon 30 added to the first action point AP1. Similarly, the second action point AP2 is drawn at the start position of the second single browsing action, and the title of the web page H2 to which the character "B" is assigned to the corresponding balloon 30 is described. .. In the balloon 30, if the corresponding web page is an order tendency page, that fact is described, and if the corresponding web page is a non-order tendency page, that fact is described.

By referring to the status of the major items on the information provision page 24, the site creator can see how many clusters exist for the dedicated website he / she opens, that is, how many when similar actions are grouped. Can recognize whether it can be divided into clusters. Furthermore, the site creator typically selects one of the major items, selects the order tab 25 or the non-order tab 26, and confirms the displayed contents, so that the user who ordered the product for each of the clusters is typical. You can see what the behavior is and what the user who did not order the product is typically doing. In particular, the site operator can recognize how the ordering tendency page or the non-ordering tendency page is viewed in a typical behavior.

As described above, in the present embodiment, the action character string (including the corrected action character string) corresponds to "information about the user's behavior" used for deriving the editing distance which is an index value regarding the similarity of the user's behavior. .. This action string lacks some information about what each of the single browsing actions (actions) performed by the user was, and the chronological order in which each single browsing action (action) was performed. It can be said that it is information that expresses the user's behavior without any problem. Then, in the present embodiment, since the index value (editing distance) relating to the similarity of the user's behavior is derived using such an action character string, the index value (editing distance) due to the lack of a part of the information is derived. ) Can be prevented from decreasing in accuracy. Further, in the present embodiment, since the edit distance between the action character strings is derived as an index value regarding the similarity of the user's behavior, it is effective that the information about the user's behavior is composed of a combination of characters. It is possible to derive an index value related to the similarity of user behavior with high accuracy. Further, the action character string is information composed of characters representing a single browsing action (action) performed by the user, and the information representing each single browsing action (action) is very small. The amount of information is small because it does not include information related to single browsing actions (actions) that the user has not performed. That is, according to the present embodiment, it is possible to suppress an increase in the amount of information regarding the user's behavior.

Next, an operation example of the information processing server 1 will be described with reference to the flowchart of FIG. As shown in FIG. 7, the conversion unit 11 executes conversion processing for each record group having a common action ID value for the browsing history data 20 to generate an action character string, and registers the browsing history data 20 in the series browsing action data 21. (Step SA1). Next, the out-licensing unit 12 executes the first out-licensing process (step SA2). As described above, in the first derivation process of step SA2, the derivation unit 12 derives the edit distance between the action character string and the action character string for each combination of all the combinations of the two records of the series browsing action data 21. To do. Next, the cluster forming unit 13 executes the first cluster forming process (step SA3). As described above, in the first cluster formation process, the cluster formation unit 13 groups a series of browsing actions to form a cluster.

Next, the conversion unit 11 derives the separation distance for each cluster based on the ordered browsing ratio and the non-browsing ratio of each web page, and derives the number of additions for each web page (step SA4). Next, the conversion unit 11 accesses the series of browsing action data 21, performs correction processing based on the number of additions for each derived web page for each cluster, and updates the value of the action character string (step SA5). Next, the out-licensing unit 12 executes the second out-licensing process (step SA6). In the second derivation process of step SA6, the derivation unit 12 derives the edit distance between the correction action character string and the correction action character string for all the combinations of the two records of the series browsing action data 21. Next, the cluster forming unit 13 executes the second cluster forming process to form a new cluster (step SA7). Next, the behavior modeling unit 14 derives a common behavior character string for orderless behavior and a common behavior character string for ordered behavior for each new cluster, and generates behavior modeling data 23 in which these are registered. (Step SA8). As described above, the behavior modeling data 23 generated in step SA8 is used when the information providing unit 15 provides information.

Although one embodiment of the present invention has been described above, the above-described embodiment is merely an example of embodiment of the present invention, whereby the technical scope of the present invention is limitedly interpreted. It should not be done. That is, the present invention can be implemented in various forms without departing from its gist or its main features.

For example, the editing distance is not limited to the Levenshtein distance. That is, the edit distance may be an index related to the similarity between different character strings and may be a value converted by the amount of editing processing.

Further, in the above embodiment, browsing of a certain web page is defined as a single browsing behavior, but in this regard, browsing of a specific area of a scrollable web page is defined as one single browsing behavior. You may do so. To explain in more detail using a simplified example, it is assumed that there is now a vertically long web page, and this web page is basically supposed to be scrolled and viewed. In this case, the web page is divided into a plurality of areas in the vertical direction, characters are assigned to each area, and characters in the action character string are configured as a single browsing action for browsing each area. It may be.

In the above embodiment, the character assigned to the web page is one character of the alphabet, but the character is not limited to one character of the alphabet. That is, it is sufficient that the character string can be composed of characters (for example, symbols) and the edit distance can be derived based on the characters composed of the characters.

Further, in the above embodiment, the reaction of the user induced by the target website (dedicated website) is the purchase of the product I, but the reaction of the user induced by the website is naturally in the above embodiment. It is not limited to the example. As an example, membership registration for some service, request for a pamphlet, transition to a link destination, participation in an event, etc. may be performed.

Further, in the above embodiment, the contents of various data have been described with reference to the figures as appropriate, but the contents of various data are not limited to the illustrated contents.

Further, in the above embodiment, the editing distance (including the corrected editing distance) derived by the derivation unit 12 is used to determine the similarity of actions in deriving the network for forming the cluster. More specifically, the edit distance was used as an index value for determining whether or not to form a link between nodes. However, the method of using the editing distance is not limited to the example. For example, it may be used to determine if any two selected series of browsing behaviors are similar, or it may be used to statistically analyze the distribution of the edit distance itself.

Further, in the above embodiment, regarding the cluster formation by the cluster forming unit 13, both the first cluster forming process based on the editing distance and the second cluster forming process based on the corrected editing distance are executed. A configuration may be used in which only the first cluster formation process is performed to form a cluster. Further, in the above embodiment, the conversion unit 11 generates the action character string reflecting the browsing time, but may generate the action character string that does not reflect the browsing time. That is, when generating the action character string, the process of increasing the number of consecutive characters as the action is continuously performed is not essential, and such a process may not be performed.

Further, in the above embodiment, the action of browsing a web page in the same website is regarded as a list browsing action, and this list browsing action is a target to be converted into an action character string. However, the target action to be converted into the action character string (hereinafter referred to as “target action”) is not limited to the list browsing action illustrated in the above embodiment. For example, not only browsing of web pages in the same website but also browsing of multiple websites may be included in the target behavior. Further, for example, in addition to browsing the website, other actions may be included in the target action. Other actions are, for example, telephone calls (eg, telephone calls for inquiries). In this case, a character is also assigned to the action of calling, for example, "viewing the web page p" (assuming that the character "P" is assigned) → "viewing the web page q" (character "Q"). Is assigned) → "Call to a predetermined phone number" (assuming that the letter "R" is assigned), the conversion unit 11 is assigned to the user. The action is converted into the action character string "PQR".

Further, for example, the behavior of a customer who visits a certain store may be the target behavior. For example, first, the floor of a store is divided into a plurality of areas, and characters are assigned to each area in advance. Then, for each of the plurality of customers, the movement route on the floor of the store is traced. For example, for a certain customer, "area l" (assuming that the character "L" is assigned) → "area m" (character). Assuming that "M" is assigned) → "area n" (assuming that the character "N" is assigned), the conversion unit 11 transfers the user's action to the action character string. Convert to "LMN".

Further, in the above embodiment, an external device capable of communicating with the information processing server 1 may execute all or part of the processing that the functional block of the information processing server 1 is supposed to execute. In this case, the information processing server 1 and the external device cooperate to function as an "information processing system". The external device is, for example, a cloud server that executes processing in response to a request from the information processing server 1. As an example, the external device may execute all or part of the processing of the conversion unit 11, or the external device may execute all or part of the processing of the derivation unit 12.

1 Information processing server (information processing system)
11 Conversion part 12 Derivation part 13 Cluster formation part 14 Behavior modeling part

Claims (15)

A conversion unit that converts information indicating user actions defined as a series of actions into an action character string, which is a character string in which the characters assigned to each action are arranged in the order in which the actions are performed.
It is provided with a derivation unit for deriving an edit distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user.
Each action is continuous for a set amount of time,
When converting information indicating a user's action into the action character string, the conversion unit converts characters assigned to each action so that the longer the action is continuously performed, the greater the number of consecutive actions. information processing system that is characterized in that is continuous. Actions include browsing a particular area of a web page or web page contained within a given website.
When converting a user's action into the action character string, the conversion unit makes the assigned characters continuous for each action related to browsing a web page so that the longer the browsing time, the greater the number of consecutive actions. The information processing system according to claim 1. A conversion unit that converts information indicating user actions defined as a series of actions into an action character string, which is a character string in which the characters assigned to each action are arranged in the order in which the actions are performed.
It is provided with a derivation unit for deriving an edit distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user.
The conversion unit is characterized in that when converting information indicating a user's action into the action character string, the assigned characters are made continuous so that the more important the action, the greater the number of consecutive actions. information processing system that. A conversion unit that converts information indicating user actions defined as a series of actions into an action character string, which is a character string in which the characters assigned to each action are arranged in the order in which the actions are performed.
A derivation unit for deriving the editing distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user.
It includes a cluster forming unit that groups the actions of two or more users based on the editing distance derived by the derivation unit and forms a plurality of clusters.
The cluster forming unit uses each of the user's actions as a node, and derives the modularity for a network in which two nodes whose editing distance between the action strings indicating the user's actions is equal to or less than a certain distance are connected by a link, and the modularity is derived. By forming the cluster so as to maximize, the edit distance between the action strings indicating the actions of each user belonging to one cluster is shortened, and the action strings indicating the actions of the users belonging to a certain cluster are shortened. the information processing system that is characterized in that edit distance between the action string indicating the behavior of users belonging to each cluster other than the cluster to form the cluster to be longer. The information processing system according to claim 4, wherein the action includes browsing one web page included in a predetermined website or a specific area on one web page. A conversion unit that converts information indicating user actions defined as a series of actions into an action character string, which is a character string in which the characters assigned to each action are arranged in the order in which the actions are performed.
A derivation unit for deriving the editing distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user.
A cluster forming unit that forms a plurality of clusters by grouping the actions of two or more users based on the editing distance derived by the derivation unit.
For each of the clusters formed by the cluster forming unit, the maximum common substring was derived for each of the two combinations of the action character strings of the actions of the users belonging to the cluster, and was derived for each of the two combinations of the action character strings. the maximum intersection of each column, information processing systems that anda behavior modeling unit to derive a new string and coupled in the order of the character is maintained in each of the largest common string .. The information processing system according to claim 6, wherein the action includes browsing one web page included in a predetermined website or a specific area on one web page. A conversion unit that converts information indicating user actions defined as a series of actions into an action character string, which is a character string in which the characters assigned to each action are arranged in the order in which the actions are performed.
It is provided with a derivation unit for deriving an edit distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user.
Actions include browsing a particular area of a web page or web page contained within a given website.
The predetermined website is intended to induce the user to perform a predetermined reaction.
When the conversion unit converts the user's action into the action character string, the degree of influence on the user's reaction to the predetermined reaction or the user's failure to perform the predetermined reaction for each of the actions related to browsing the web page. as successive times higher action increases, information processing system that is characterized in that it is continuously assigned characters. A conversion unit that converts information indicating user actions defined as a series of actions into an action character string, which is a character string in which the characters assigned to each action are arranged in the order in which the actions are performed.
It is provided with a derivation unit for deriving an edit distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user.
Actions include browsing a particular area of a web page or web page contained within a given website.
The predetermined website is intended to induce the user to perform a predetermined reaction.
The behavior of two or more users is grouped based on the editing distance derived by the derivation unit, and a cluster forming unit for forming a plurality of clusters is further provided.
The conversion unit converts each of the user's actions into the action character string.
The derivation unit executes a first derivation process for deriving the edit distance between the action character strings for each of the two combinations of user actions.
The cluster forming unit groups the actions of two or more users based on the editing distance derived by the first derivation process of the derivation unit, and executes the first cluster forming process of forming a plurality of clusters. And
The conversion unit has a tendency to be viewed when the predetermined reaction is performed for each of the web pages for each cluster formed by the first cluster formation process of the cluster formation unit, or the predetermined value. The strength of the tendency that was viewed when the reaction was not made is derived, and each of the action character strings indicating the user's action is browsed when the predetermined reaction is made in the cluster to which the user's action belongs. Allocate so that the action related to browsing a web page with a strong tendency to be viewed, or the action related to browsing a web page with a strong tendency to be browsed when the predetermined reaction is not made, the number of consecutive times increases. Executes the correction process to make the characters continuous,
After the correction process is executed by the conversion unit, the derivation unit executes a second derivation process for deriving the edit distance between the action character strings for each of the two combinations of the actions of the two users.
The cluster forming unit, information processing system that is characterized in that performing a second cluster forming process for forming a plurality of clusters based on the edit distance derived by the second derivation of the derived portions. The conversion unit of the information processing system converts the information indicating the user's action defined as a series of actions into an action character string in which the characters given for each action are arranged in the order in which the actions are performed. The first step and
The derivation unit of the information processing system determines the editing distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user. Including the second step to derive
Each action is continuous for a set amount of time,
In the first step, when the conversion unit converts the information indicating the user's action into the action character string, the number of consecutive actions increases as the time during which the action is continuously performed increases for each action. , An information processing method characterized by making the assigned characters continuous. The conversion unit of the information processing system converts the information indicating the user's action defined as a series of actions into an action character string in which the characters given for each action are arranged in the order in which the actions are performed. The first step and
The derivation unit of the information processing system determines the editing distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user. Including the second step to derive
In the first step, when converting information indicating a user's action into the action character string, the conversion unit converts characters assigned to each action so that the more important the action, the greater the number of consecutive times. An information processing method characterized by being continuous. The conversion unit of the information processing system converts the information indicating the user's action defined as a series of actions into an action character string in which the characters given for each action are arranged in the order in which the actions are performed. The first step and
The derivation unit of the information processing system determines the editing distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user. The second step to derive and
The cluster forming unit of the information processing system includes a third step of grouping the actions of two or more users based on the editing distance derived by the derivation unit to form a plurality of clusters.
In the third step, the cluster forming unit has each of the user's actions as a node, and the modularity of the network in which two nodes indicating the user's action and the edit distance between the action strings is less than a certain value are connected by a link. By deriving and forming a cluster so as to maximize the modularity, the edit distance between the action strings indicating the actions of each user belonging to one cluster is shortened, and the actions of users belonging to a certain cluster are shortened. An information processing method characterized in that a cluster is formed so that the editing distance between the action character string indicating the above and the action character string indicating the behavior of a user belonging to each cluster other than the cluster is long. The conversion unit of the information processing system converts the information indicating the user's action defined as a series of actions into an action character string in which the characters given for each action are arranged in the order in which the actions are performed. The first step and
The derivation unit of the information processing system determines the editing distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user. The second step to derive and
A third step in which the cluster forming unit of the information processing system groups the actions of two or more users based on the editing distance derived by the derivation unit to form a plurality of clusters.
The behavior modeling unit of the information processing system derives the maximum common substring for each of the two combinations of the behavior character strings of the user's behavior belonging to the cluster for each of the clusters formed by the cluster formation unit. The fourth step of deriving a new character string by combining each of the maximum common substrings derived for each combination of the action character strings while maintaining the order of the characters in each maximum common character string is included. An information processing method characterized by the fact that. The conversion unit of the information processing system converts the information indicating the user's action defined as a series of actions into an action character string in which the characters given for each action are arranged in the order in which the actions are performed. The first step and
The derivation unit of the information processing system determines the editing distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user. Including the second step to derive
Actions include browsing a particular area of a web page or web page contained within a given website.
The predetermined website is intended to induce the user to perform a predetermined reaction.
In the first step, when the conversion unit converts the user's action into the action character string, the user makes the predetermined reaction or makes the predetermined reaction for each of the actions related to browsing the web page. An information processing method characterized in that the assigned characters are made continuous so that the number of consecutive actions increases as the action has a higher influence on not doing. A conversion unit that converts information indicating user actions defined as a series of actions into an action character string, which is a character string in which the characters assigned to each action are arranged in the order in which the actions are performed.
It is provided with a derivation unit for deriving an edit distance between the action character string converted by the conversion unit for the behavior of one user and the action character string converted by the conversion unit for the behavior of another user.
Actions include browsing a particular area of a web page or web page contained within a given website.
The predetermined website is intended to induce the user to perform a predetermined reaction.
It is an information processing method by an information processing system further including a cluster forming unit that forms a plurality of clusters by grouping the actions of two or more users based on the editing distance derived by the derivation unit.
A step in which the conversion unit converts each of the user's actions into the action character string,
A step in which the derivation unit executes a first derivation process for deriving an edit distance between the action character strings for each of two combinations of user actions.
The cluster forming unit groups the actions of two or more users based on the editing distance derived by the first derivation process of the derivation unit, and executes the first cluster forming process of forming a plurality of clusters. Steps to do and
The strength of the tendency that the conversion unit was viewed when the predetermined reaction was performed for each of the web pages for each cluster formed by the first cluster formation process of the cluster formation unit, or the predetermined The strength of the tendency that was viewed when the reaction was not made is derived, and each of the action character strings indicating the user's action is browsed when the predetermined reaction is made in the cluster to which the user's action belongs. Allocate so that the action related to browsing a web page with a strong tendency to be viewed, or the action related to browsing a web page with a strong tendency to be browsed when the predetermined reaction is not made, the number of consecutive times increases. A step to execute a correction process to make the characters continuous, and
After the correction process is executed by the conversion unit, the derivation unit executes a second derivation process for deriving the edit distance between the action character strings for each of the two combinations of the actions of the two users. ,
An information processing method, wherein the cluster forming unit includes a step of executing a second cluster forming process of forming a plurality of clusters based on an edit distance derived by the second derivation process of the derivation unit. ..