JP2001297115A - Hypertext access pattern analysis device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a hypertext access pattern analysis device capable of providing an access pattern effective to an analyzer. SOLUTION: A sorting means 2 sorts plural texts (hypertexts) provided from a system 1 into plural clusters on the basis of the contents of the texts, an access history detection means 2 detects the access history of each user for texts, an access pattern generation means 5 generates an access pattern obtained by adding texts linked to texts included in a set defined as a species in which the rate of users simultaneously accessing all texts included in a text set obtained by combining plural texts in each cluster is more than a prescribed value to the set, and a providing means 7 provides the access pattern.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a hypertext system configured on a network, for example,
A hypertext access (browsing) pattern analysis device for efficiently finding a user's access pattern on a hypertext system, and a hypertext access pattern analysis for realizing the analysis device on a computer It relates to a medium on which a program is recorded.

[0002]

2. Description of the Related Art In recent years, the demand for providing information on the World Wide Web (hereinafter abbreviated as Web) has been increasing.
Web site for providing information on the Web in companies
(Equivalent to an example of a hypertext system), and websites are actively promoting the sale of goods and services by means of advertising activities of their products or electronic commerce. Under these circumstances, technology that discovers, as an access pattern, browsing sequences / sets of hypertexts common to many users from the access history of users (customers) who have accessed the Web effectively operates Web sites It is indispensable above.

[0003] Information provided on a Web site is called content, and is represented by hypertext (for example, HTML). Individual hypertext information is linked by links. The user who accesses the Web site can access the hypertext in the Web site by following the link to obtain various information. Web
The server can record the behavior of the user who has accessed the server (the recorded content is hereinafter referred to as access history).

The access history includes the IP address of the accessing computer (the computer used by the user to access the Web server or the computer relaying the user's usage information), the access time, and the information. Instructions sent to the Web server for access are recorded. The instruction contains the identifier of the hypertext accessed by the user on the server (for example, UR
L) is included.

[0005] A conventional technique for finding a user's trend by finding a user's access pattern using an access history will be described. For example, “Efficient Data Mining fo
r Path Traversal Patterns ”, IEEE Trans. on Knowled
ge and Data Engineering, 1998 ”(hereinafter referred to as Reference 1) describes a technique for extracting a sequence called MF (Maximal Forward references). In the example shown in FIG. 12, when a certain user browses the nodes in the order of A → B → C → D → E → F → G, the MF [ABCD] [ABE] [AFG] is obtained. The MF is found for all users who have accessed, and the number of appearances of the MF subsequence is counted.
Is obtained as a user access pattern.

However, in this technique, when an MF sequence is generated, a portion close to the entrance of the site is counted a plurality of times, and a partial sequence of the MF sequence from the entrance of the site to a page group having a small number of link transitions is generated. Tend to increase the number of appearances. For this reason, in this technology, only the access pattern in the part where the number of link transitions from the top page is small is emphasized, and it is necessary to effectively find the access pattern for the hypertext group that has to transit many links. There was a problem that it was difficult.

[0007] For example, "Mining Association Rul
e in Hypertext Databases “In Proc of KDD-98” (hereinafter referred to as reference 2) extracts the number of user transitions between nodes linked by a link from a log and associates the node sequence accessed by the user with a correlation rule. Is described as a user access pattern.

However, in this technique, since only the number of transitions of a user between nodes is observed, there is a problem that the same user does not always follow the sequence. Referring to FIG. 13 as an example, in this technique, a correlation is obtained in which a user following the link of B-> A tends to follow the link of A-> F. It does not indicate that you have visited all of F, and does not accurately reflect the user's access trends.

[0009] Further, in both the techniques described in the above-mentioned Documents 1 and 2, since the contents of the contents arranged in the server are not taken into consideration, meaningless access patterns that do not help the user to know the trend of the user are considered. Are also found in large quantities. For this reason, there is a problem that it takes a lot of time to scrutinize the found access pattern.

[0010]

As described in the above-mentioned prior art, the conventional technique has an insufficient point for finding a user's access pattern, so that the above-mentioned problems can be solved. Technology development was desired. The present invention has been made in view of such a conventional situation. For example, by examining the relationship between the intent of the site configuration of the analyst and the user's interest in consideration of the contents of the content, the access pattern is closely examined. It is an object of the present invention to provide a hypertext access pattern analysis device and a storage medium that can facilitate the above.

The present invention is also applicable to a hypertext system in which the number of accesses in a hypertext system is short and the link distance is relatively short, as well as a hypertext in which the number of accesses is relatively short and the link distance is deep. It is an object of the present invention to provide a hypertext access pattern analysis device capable of discovering a characteristic access pattern among users who visit. Another object of the present invention is to provide a hypertext access pattern analysis device capable of improving the efficiency of access pattern generation.

[0012]

In order to achieve the above object, in the hypertext access pattern analyzing apparatus according to the present invention, the hypertext classifying means converts a plurality of hypertexts provided by the hypertext system into contents of each hypertext. The access history detecting means detects the access history of each user with respect to the hypertext, and the access pattern generating means classifies each cluster based on the detected access history of each user. In a hypertext set formed by combining a plurality of hypertexts included in a hypertext set in which the percentage of users who accessed all hypertexts included in the hypertext set at the same access is equal to or more than a predetermined value as a seed, Seed high Hypertext which is connected by links to hypertext contained in over text set generates an access pattern formed by adding to the hypertext set presents an access pattern by the access pattern presenting means is generated.

According to a further specific embodiment of the present invention, the hypertext classification means classifies each hypertext in the hypertext system into a plurality of clusters based on the contents described in each hypertext. ,
The access history detection means pre-processes the access history for each hypertext collected by the server to extract a hypertext access sequence for each user, and the access pattern generation means based on the access sequence for each user, A hypertext set included in a seed hypertext set is directly linked to a hypertext included in the seed hypertext set by using the hypertext set which is common to users having a ratio equal to or greater than a predetermined value in each cluster and accessed by the respective users at the same access. An access pattern is generated by adding a hypertext that is stored, and the generated access pattern is stored in an access pattern storing unit. The access pattern presenting unit stores all of the stored access patterns in an analyst (user). And ordered according to a measure of interest Presented to the. In this case, for example, the ordering unit performs the ordering based on the number of users who have accessed an arbitrary page (hypertext) of each of the classified clusters (each cluster is formed of, for example, a set of hypertext).

Accordingly, an access pattern is generated for each cluster classified based on the content of each hypertext, and the access pattern is analyzed by an analyst (user).
For example, by considering the contents of the hypertext content and showing the relationship between the analyst's intention of the site configuration and the user's interest, it is possible to easily examine the access pattern.

As the hypertext system, for example, various systems may be used as long as a plurality of hypertexts are connected and held by links. Also, accessing all the hypertexts included in the hypertext set at the same access means that, for example, the user follows links between the hypertexts and stops at all the hypertexts without stopping the access to the hypertexts. Indicates that access was made consecutively.

As the ratio of users who have made such access, for example, a minimum support, a support, or both of them, which will be described later, can be used. Also, various values may be used as the above-mentioned predetermined value, for example, according to the usage status of the device. Further, as a method of presenting the access pattern, for example, a method of displaying and displaying the access pattern on a display screen or the like, a method of presenting by printing and outputting with a printer, or the like can be used.

Further, in the hypertext access pattern analysis apparatus according to the present invention, all the hypertexts included in the target hypertext set with respect to the number of users who have accessed the hypertext provided by the hypertext system are accessed at the same time. The ratio of the number of users who accessed the site is the minimum support,
The ratio of the number of users accessing all hypertexts included in the target hypertext set at the same access to the number of users accessing the hypertexts included in the target cluster, as the support level.

Here, the number of users who have accessed the hypertext provided by the hypertext system (hereinafter referred to as the number of users A) refers to, for example, at least one hypertext provided by the hypertext system. When the same user accesses two or more hypertexts, the same user is counted as one user. The number of users who have accessed all hypertexts included in the target hypertext set at the same access (hereinafter, referred to as the number of users B) is as follows.
For example, the order of access may be arbitrary, but refers to the number of users who have accessed all the hypertexts at the same access.

The number of users accessing the hypertext included in the target cluster (hereinafter referred to as the number of users C) is, for example, the number of users accessing at least one hypertext included in the cluster. If the same user accesses two or more hypertexts, the same user is counted as one user. The number of users A described above
And the number of users B, the minimum support is represented by, for example, (number of users B / number of users A). When the number of users C and the number of users B described above are used, the support level is represented by, for example, (the number of users B / the number of users C).

Then, the access pattern generating means determines that the minimum support is equal to or more than a predetermined minimum support threshold and the support is within a hypertext set formed by combining a plurality of hypertexts included in the target cluster. A hypertext set that is equal to or greater than a predetermined support threshold and the number of included hypertexts is equal to or greater than a predetermined number is used as a seed, and with respect to various hypertext sets, the hypertext included in the seed hypertext set is directly used. A hypertext set obtained by adding a hypertext linked by a link to the hypertext set, and the hypertext set in the same manner as above.
Among the hypertext sets obtained by adding the above hypertexts, a hypertext set whose support level is the same as the support level of the hypertext set as the kind is regarded as an access pattern candidate, and the access pattern candidate is determined for each access pattern candidate. A hypertext set formed by adding hypertext directly linked to the hypertext included in the pattern candidate to the access pattern candidate, and similarly adding one or more hypertexts to the hypertext set. In the hypertext set, a hypertext set whose minimum support is equal to or more than a predetermined minimum support threshold and whose support is equal to or more than a predetermined support threshold is set as an access pattern.

As an example of a more specific embodiment of the present invention, the access pattern generation means includes a cluster whose access number is determined to be "the number of users" by the ordering means and an access history. Is input, and among the combinations of hypertexts included in the cluster, the ratio of the number of users who browsed all the hypertexts in the combination to the number of users in the entire access history is set as the minimum support, and the “user The ratio with respect to Expression 1 "is defined as the support. Then, a combination in which both the minimum support and the support exceed the predetermined minimum support threshold and the support threshold, respectively, and the number of hypertexts included in the combination is equal to or more than a predetermined number is seeded. When the hypertext linked by a direct link to the hypertext included in the species is added for each generated species, the hypertext included in the species and the added hypertext are accessed at the same access The number of active users is calculated as “the number of users 2”.

At this time, "number of users 1" of "number of users 2"
The addition of a hypertext that does not change the support rate, which is the ratio of the access pattern candidate, is used as an access pattern candidate, and the same operation is performed for the obtained new access pattern candidate.There is no additional hypertext that does not change the support rate. Repeat until With regard to the access pattern candidates obtained in this way, when a hypertext directly linked by a link to the hypertext included in the access pattern candidate is added, the hypertext included in the access pattern candidate and the added hypertext are added. The number of users accessing at the same access is calculated as “the number of users 3”.

At this time, a support which is a ratio of "the number of users 3" to "the number of users 1" and a minimum support which is a ratio of "the number of users 3" to the number of users of the entire access history are given in advance. A hypertext exceeding the support threshold and the minimum support threshold is added as a new access pattern candidate, and this operation is repeated until there is no more hypertext that can be added. Then, an access pattern candidate obtained by the above process is generated as an access pattern.

Accordingly, since the access pattern is generated in consideration of not only the minimum support as described above but also the support as described above, for example, the hypertext system in which the number of accesses is large and the link distance is short is short. Similarly to a set, for a hypertext with a relatively small number of accesses and a deep link distance, a characteristic access pattern can be found among users who are visiting the hypertext in the same area. Note that various values may be used as the above-described predetermined minimum support threshold, the predetermined support threshold, and the predetermined number in accordance with, for example, the usage status of the apparatus.

Further, in the hypertext access pattern analysis device according to the present invention, the access pattern generation means may execute an access pattern candidate generation process and an access pattern generation process relating to various hypertext sets as a seed hypertext set. Are executed in ascending order of the number of users who have accessed the hypertext included in the access pattern candidate generation process and the access pattern generation process. If one of them is less than the predetermined threshold value (that is, the minimum support threshold value or the support threshold value), information on the under-threshold information specifying the combination pattern of the hypertexts constituting the hypertext set is stored. Pattern candidate In the process of the processing and the access pattern generation processing, the hypertext set is added in response to the additional hypertext set including all the hypertexts of the combination pattern specified by the stored information less than the threshold. In the course of the access pattern candidate generation processing, it is excluded from the access pattern candidates, and in the course of the access pattern generation processing, it is excluded from the access patterns.

As an example of a more specific aspect of the present invention, the access pattern generating means generates an access pattern in the order of the number of accessed users in descending order of the number of users. In this case, in the process of generating an access pattern from a seed, a hypertext set whose minimum support is less than the minimum support threshold or whose support is less than the support threshold is recorded by the generation process holding means. I do. Then, when generating an access pattern candidate or an access pattern from each species, a minimum support or support level is set for a hypertext set having, as a partial pattern, the hypertext set held by the generation process holding means in the previous process. Is less than the threshold value and is excluded from the access pattern candidates and access patterns.

Therefore, access pattern candidate generation processing and access pattern generation processing relating to various hypertext sets in the descending order of the number of accessed users are executed, and it is determined that the minimum support and the support are already less than the threshold. For the hypertext set including the pattern, the pattern is excluded from the access pattern candidates and the access pattern without performing the determination.
The efficiency of the access pattern generation processing can be improved.

Further, in the hypertext access pattern analyzing apparatus according to the present invention, all the hypertexts included in the target hypertext set with respect to the number of users accessing the hypertext provided by the hypertext system are accessed at the same time. The ratio of the number of users who accessed the site is the minimum support,
The ratio of the number of users who accessed all hypertexts included in the target hypertext set at the same access to the number of users accessing hypertexts included in the target cluster as the support level is provided by the hypertext system. The ratio of the number of users who accessed in the same order to all hypertexts included in the target hypertext set with respect to the number of users who accessed the hypertext to be performed in the same order as the minimum order of the same order, The ratio of the number of users accessing the hypertext included in the target hypertext set to the number of users accessing the hypertext included in the target cluster in the same order at the same access to the number of users accessing the hypertext included in the target cluster is defined as the same order support.

Here, the number of users who access all the hypertexts included in the target hypertext set in the same order at the same access (hereinafter referred to as the number of users D) means, for example, the access order. This is the number of users who are the same and have accessed all the hypertexts at the same time. When the number of users A and the number of users D described above are used, the same order minimum support is expressed by, for example, (number of users D / number of users A).
When the number of users C and the number of users D are used, the same order support is represented by, for example, (number of users D / number of users C).

Then, the access pattern generating means firstly sets, in a hypertext set formed by combining a plurality of hypertexts included in the target cluster,
A hypertext set whose minimum support is equal to or greater than a predetermined minimum support threshold and whose support is equal to or greater than a predetermined support threshold is selected. A hypertext set whose order support is equal to or more than the minimum order support threshold, the same order support is equal to or larger than the predetermined same order support threshold, and the number of included hypertexts is equal to or larger than a predetermined number is used as a seed.

Next, the hypertext generating means adds, to the hypertext set, hypertexts that are directly linked to hypertexts included in the hypertext set to be various types. And the same order support of the hypertext set having the same order support in the hypertext set obtained by adding one or more hypertexts to the hypertext set in the same manner. The hypertext set which is the same as that described above is set as an access pattern candidate.

Then, the access pattern generating means adds, to each access pattern candidate, a hypertext set obtained by adding a hypertext directly linked to the hypertext included in the access pattern candidate to the access pattern candidate. And similarly, in a hypertext set obtained by further adding one or more hypertexts to the hypertext set, the minimum support is equal to or greater than a predetermined minimum support threshold and the support is equal to or greater than a predetermined support threshold. Is selected, and in the selected hypertext set, the same order minimum indicating degree is equal to or greater than a predetermined same order minimum supporting threshold and the same order supporting degree is equal to or greater than a predetermined same order supporting degree threshold. Let a certain hypertext set be an access pattern.

Therefore, for example, an access pattern that matches the access order of each user is generated and presented instead of an access pattern based merely on whether or not each user has accessed, so that the access trend of the user is accurately reflected. The access pattern can be presented to the analyst. The predetermined minimum support threshold, the predetermined support threshold, the predetermined same-order minimum support threshold, the predetermined same-order support threshold, and the predetermined number are, for example, depending on the usage status of the device. Various values may be used accordingly.

In the hypertext access pattern analysis device according to the present invention, the hypertext classifying means classifies the plurality of hypertexts into a plurality of clusters having a hierarchical relationship, and the access pattern presenting means sets the hypertext as a seed of the access pattern. An access pattern that includes a species other than the species to which the user belongs, and an access pattern that includes species other than the species to which the user belongs. Presents the access patterns generated from the seeds.

Here, the number of users who have accessed the hypertext included in the access pattern is, for example, the number of users who have accessed at least one hypertext included in the access pattern. When the above hypertext is accessed, the same user is counted as one user.

As an example of a more specific embodiment of the present invention, the hypertext classifying means classifies the hypertext into clusters having a hierarchical relationship, and the access pattern holding means stores the generated access pattern and its source. Keep the pair with the seed that became. Then, a search is made for access patterns that include other species as a part, in order from a cluster having a deeper cluster to which the held paired species belongs and a large number of users sharing the access pattern. A user who shares the discovered access pattern and the access pattern generated from the other species included in the access pattern with the analyst, sharing the hierarchical relationship of the cluster, the number of hypertexts included in the access pattern, and the access pattern. It is ranked and presented based on the number of.

Accordingly, an access pattern including a species other than the species to which the user belongs and the access pattern and the corresponding Since an access pattern generated from a species other than the species to which the subject belongs is presented, an access pattern considered to be interesting for the analyst can be presented.

Further, in the hypertext access pattern analysis device according to the present invention, the accepting means accepts the designation of the access pattern from the user, and the hypertext presenting means acquires the hypertext included in the designated access pattern and the hypertext included in the designated access pattern. Among the hypertexts included in the cluster to which the seed hypertext set belongs, the hypertext directly connected to the hypertext included in the seed hypertext set by a link, and the hypertext as the seed of the access pattern. A hypertext that can be reached from the hypertext included in the hypertext set as the seed among the hypertexts included in the cluster to which the text set belongs via any one hypertext included in the access pattern. Text and the hypertext included in the specified access pattern and another hypertext (that is, the hypertext not included in the access pattern) are provided in a different manner, so that the number of accesses of the user is reduced. The hypertext to be presented is presented in a tree structure with the largest hypertext as the root.

As an example of a more specific embodiment of the present invention, when the user designates an arbitrary access pattern, the accepting means accepts the designation and displays the hypertext in the designated access pattern for displaying the hypertext. It is stored as a set, and among the hypertexts in the cluster to which the species belongs, the hypertext directly linked by a link from the hypertext included in the species that is the source of the access pattern is stored as a display hypertext set, and the Of the hypertext in the cluster to which it belongs,
The hypertext that can be reached from the hypertext included in the seed that is the source of the access pattern via any one hypertext in the access pattern is stored as a display hypertext set. Then, the access pattern presenting means displays the hypertext held as the display hypertext set in a tree structure with the hypertext having the largest number of accesses by the user as a root, and at this time, the hypertext included in the designated access pattern is displayed. Text and hypertext not included are presented to the user in an identifiable manner.

Therefore, according to the designation from the user, the presentation mode of the hypertext included in the designated access pattern is made different from that of another hypertext to present the hypertext to be presented. Since the hypertext to be presented is presented in a tree structure with the largest hypertext as the root, it is possible for the analyst to easily see and grasp the presented content.

Here, various methods may be used as a method for making the presentation mode of the hypertext included in the designated access pattern different from that of the other hypertexts. What is necessary is to be presented so that the hypertext included in the.

The various processes according to the present invention as described above can also be realized by, for example, reading and executing a program stored in a storage medium by a computer. As an example, the storage medium according to the present invention stores a program to be executed by a computer in a manner readable by input means of the computer, and the program stores a plurality of hypertexts provided by a hypertext system in each hypertext system. A process of classifying into a plurality of clusters based on the content of the hypertext, a process of detecting an access history of each user with respect to the hypertext, and a process of detecting a plurality of clusters included in the cluster based on the detected access history of each user. Of the hypertext set that combines all the hypertexts included in the hypertext set as a seed, and a hypertext set in which the percentage of users accessing the hypertext at the same access is equal to or more than a predetermined value is used as a seed. Included in text set And generating an access pattern comprising a hypertext connected by links to Lee hypertext in addition to the hypertext set, it causes the process of presenting the generated access pattern executed on the computer.

[0043]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment according to the present invention will be described with reference to the drawings. FIG. 1 shows an example of a system including a hypertext access pattern analysis device according to the present embodiment. The system includes a Web server 1 having a content providing unit 11 and an access history recording unit 12.
An access history preprocessing unit 2 forming an access history detecting unit; a hypertext classifying unit 3 forming a hypertext classifying unit; an ordering unit 4 forming an ordering unit; and an access pattern forming an access pattern generating unit. The generating unit 5, the access pattern holding unit 6 that forms the access pattern holding unit, the access pattern presenting unit 7 that forms the access pattern presenting unit, and the receiving unit that receives instructions and the like from the user by using a keyboard, a mouse, or the like. A reception unit (not shown) is provided.

The Web server 1 is equivalent to an example of a hypertext system, and has a function of transmitting information on a network. The web server 1 stores information (contents) to be provided to the user as hypertext. The Web server 1 provides the content by the content providing unit 11 according to the access from the user, and the identifier (IP address) for identifying the user every time the user accesses by the access history recording unit 12.・ Time ・ Record the hypertext address (URL) accessed by the user.

Here, the IP address recorded on the server side may be the IP address of a computer called a proxy server which performs access from a client actually used by a user. In this case, a plurality of users use the same IP address. Therefore, in this embodiment, the user is identified based on the IP address of the computer that has accessed the server and the access interval. That is, even if access is performed using the same IP address, if the access time interval is longer than a predetermined time interval set in advance, the access is identified as access from a different user.

The access history used in the present invention is not necessarily limited to the one described above.
For example, as disclosed in Japanese Patent Application Laid-Open No. H10-224349, a system in which a user is identified by using an access history recorded by a proxy together can be used. It is also possible to use an access history recorded by a mechanism having an access notification mechanism on the client side, such as a Java applet, as in the invention described in (1). In short, for example, the location of information accessed by the user and the Anything that can be understood may be used.

The access history preprocessing unit 2 extracts the identifiers and the hypertexts accessed by the users having the identifiers in chronological order, and thereafter retrieves a set of hypertexts visited by each user for a new access for each user. It has a function to use as a history.

Next, the hypertext classification unit 3
It has a function of classifying the content stored in the server 1 into clusters. As an example of a method of realizing the hypertext classification unit 3, a thesaurus as shown in FIG. 2 is given in advance, and each word is used as a classifier, and the classification is preliminarily assigned to each hypertext using an RDF (Resource Description Frame).
work) or other framework can be used. In this case, a method of using hypertexts belonging to the same classification as one cluster, a method of belonging to the classification having the same depth from the root of the thesaurus as one cluster, or the like can be used.

Here, as a specific example, it is assumed that there is a site configuration shown in FIG. 3 (O represents one hypertext) and the thesaurus shown in FIG. In this case, for example, the user classifies the hypertexts B, G, and K in FIG. 3 into “products”, classifies the hypertexts F, I, and L into “price”, and hypertexts C, D, E, H, and Assuming that J is classified into “functions”, when constructing a cluster for each classifier, {B, G, K}, {F, I, L}, {C, D, E, H, J} Are generated. Also, when clusters are formed not by classifiers but by the thesaurus hierarchy, 階層 B, G,
Two initial sets are formed, K｝ and {C, D, E, H, J, F, I, L}. Note that the cluster {B, G, K} represents a cluster composed of hypertexts B, G, and K, and the same applies to other clusters.

Hereinafter, the description will be given by taking as an example a case where a cluster is constituted by being divided by the thesaurus hierarchy (shown as the latter). The method of classifying the hypertext by the hypertext classification unit 3 is not necessarily limited to the method described in the present embodiment. For example, based on the content / content described in the hypertext, It should just be something which reflects. For example, the frequency of occurrence of words is extracted from the text described in the hypertext, the similarity of the frequency of occurrence of words between the hypertexts is evaluated using a known Vector Space Model, and a known hierarchical clustering method based on the word similarity is used. It is also possible to use a cluster obtained by applying. The above-mentioned Vector Space Model is described in, for example, “G. Sallton,“ Developments in automatic text
retrieval ”, Science, 253, 1991”
The above-described hierarchical clustering method is described in, for example, `` Ellen M.
Voohres, “IMPLEMENTING AGGLOMERATIVEHERARCHIC CL
USTERING ALGORITHMS FOR USE IN DOCUMENT RETRIEVA
L ”, Information Processing & Management, Vol. 22, N
o. 6, pp. 465-476, 1986 ".

Next, the ordering section 4 has a function of ordering the clusters according to the number of users who have visited the hypertext belonging to the cluster. When counting the number of users, a user visiting a plurality of hypertexts is counted only once. In this embodiment, regarding the number of users, the number of users who visited the clusters {B, G, K} is larger than the number of users who visited the clusters {C, D, E, H, J, F, I, L} (User (｛B, G, K｝)> User (｛C, D, E, H, J, F, I,
Suppose that the relationship L｝) holds.

Next, the access pattern generation unit 5 has a function of generating an access pattern in order from a cluster having a small number of users. In the present embodiment, the efficiency of the processing in the access pattern generation process is improved as described in detail later by generating the data in the order of the small number of users.

FIG. 4 shows an example of the configuration of the access pattern generation unit 5. The access pattern generation unit 5 of this embodiment includes a simultaneous access pattern discovery unit 2 in a cluster.
1, an access pattern finding unit 22 that does not change the support, which is a component for growing a seed into an access pattern, an access pattern finding unit 23 that satisfies the support condition, and a generation process that constitutes a generation process holding unit Holder 2
4 are provided.

First, the simultaneous access pattern finding unit 21 in a cluster has a function of finding a combination (simultaneous access pattern) of hypertexts that the user tends to visit at the same time for the hypertexts in the target cluster. are doing. A hypertext in a cluster is a collection of hypertexts that a website creator considers to be similar or related information, and discovering the relationship between those hypertexts is It will be interesting for the analyst. Therefore, the access pattern generation unit 5 obtains a simultaneous access pattern in the cluster as a seed, analyzes how users who visited the seed have visited the Web site, and obtains an access common to many users. Get the pattern.

An interesting access pattern for users making various accesses is an access pattern common to many users. Therefore, as a measure of the depth of interest of the pattern, the ratio of users who share the predetermined pattern to all users who access (hereinafter referred to as support) is used. In the present invention, since an access pattern is generated based on individual clusters, the ratio of users who share the access pattern to all users who have accessed the hypertext belonging to the cluster (initial set) is used as support.

However, if the number of users who have accessed the hypertext belonging to the cluster is extremely small using only this support level, a pattern that is shared by only a small number of users when viewed from the entire site is discovered. Will be lost. Therefore, the ratio of the number of users sharing the pattern to the number of users belonging to the entire site (that is, the total number of users accessing the hypertext provided by the Web server 1) is also used as the minimum support. By using two degrees of support in this way, it is possible to appropriately find a pattern not only in a region with a large number of accesses but also in a region with a small number of accesses.

The above-described simultaneous access pattern finding unit 21 in the cluster performs various combinations of two or more hypertexts included in the cluster (that is, a combination of some hypertexts included in the cluster or a combination of the hypertexts included in the cluster). In all the included hypertext combinations), a search is made for a combination in which the minimum support and the support are higher than a predetermined minimum support threshold and a support threshold, respectively. Among the searched combinations, for example, a combination that is not included in other combinations and has the number of hypertexts included in the searched combination that is equal to or greater than a given number is used as a seed, and two support levels are set. Combinations that do not satisfy the threshold are held in the generation process holding unit 24. At this time, the minimum support and the support are calculated by the following equations.

[0058]

(Equation 1)

[0059]

(Equation 2)

For example, of the above initial set (cluster), for {C, D, E, H, J, F, I, L} with a small number of users,
When the pattern Pt (F, I, L) and the pattern Pt (C, D, F) are generated as access pattern candidate seeds, the generation process holding unit 24 stores another pattern set {Pt (D, E), Pt (C, H),…}
Will be held. Note that the pattern Pt (F, I, L) represents an access pattern composed of hypertexts F, I, and L, and the same applies to other patterns.

As a means for realizing a process of finding a simultaneous access pattern in an initial set (cluster) at a high speed, a known method described in, for example, JP-A-8-287106 can be used. For reference, in this known method, a minimum support value (minimum number of times) defined by a user in an item set of a consumer transaction stored in a database for the purpose of rapidly mining a large-scale database. Identifies the set of items that appear in the database as a large item set, and compares the number of times that large item set appears in the database with the number of times that a particular subset of the item set appears in the database. It has been made to discover.

Here, the process of finding a simultaneous access pattern will be described by taking as an example a case where the access order is considered.
When considering the access order, counting the number of users in a pattern that considers the order from the beginning would increase the computational cost. The calculation cost is reduced by taking the order into consideration.

More specifically, first, at the stage of generating each combination, the minimum support and the support are calculated without considering the access order. Next, only when the minimum support and the support are equal to or more than the respective thresholds, the minimum support (the same order minimum support) and the support (the same order support) for the number of users who match up to the access order are calculated. . Then, if both the same-order minimum support and the same-order support are equal to or larger than the respective thresholds, a hypertext is further added to increase the number of elements in the combination, and a seed access pattern is generated. Hereinafter, when the access order is considered in all the processing steps, the same method is adopted.

Next, an access pattern to which a hypertext is further added is generated based on the generated seed.
The process of this process is performed by two processing units: an access pattern finding unit 22 that does not change the support level and an access pattern finding unit 23 that satisfies the support level condition.
In this embodiment, an access pattern is generated by adding a hypertext to a seed access pattern according to the scale of support and minimum support. For this reason,
The hypertext viewed by all the users sharing the seed access pattern is always included in the finally generated access pattern. Therefore, by first generating an access pattern including such a hypertext, it becomes possible to narrow the search range of access pattern candidates as a whole.

Here, the two processing units 22 and 23 described above
Will be described using a specific example. For example, L ₀ = {P ₀₀ , P ₀₁ ,..., P _0k } is given as a seed of an access pattern related to a certain cluster L ₀ , and a generation process holding unit is generated in a process where an access pattern was generated before the cluster L _0. A set of access patterns held in 24 is LC. Here, each P represents a seed obtained from the cluster. The first (first) suffix of the seed P and the suffix of the cluster L indicate the number of hypertexts added in addition to the seed included in the access pattern. For example L ₂ represents a set of the obtained access pattern candidate by adding two hypertext access pattern as a seed. The last (second) subscript of the seed P indicates the seed number generated for a certain cluster.

FIG. 5 is a flowchart of a process performed by the access pattern finding unit 22 that does not change the support. That is, in the finding unit 22, in order from i = 0 (step S1), first, each access pattern in the given access pattern candidate set L _i can be reached from the hypertext set in the access pattern. Then, a hypertext not included in the original cluster is added, and the access pattern set L _{i + 1}
Is generated (step S2).

Here, the reachable hypertext means a hypertext directly connected to the hypertext in the access pattern by a link, and does not consider the direction of the link. For example, in this embodiment, the initial set
Pt (F, one of the species generated from {C, D, E, H, J, F, I, L}
From (I, L), the elements of L ₁ include pattern Pt (F, I, L, G), pattern Pt (F, I, L, B), and pattern Pt (F, I, L, K). Generated.

Next, the generation process holding unit 24 holds a set LC of access patterns that did not satisfy the minimum support or the threshold of the support in the past access pattern generation process. In the present embodiment, branch hunting in the search of the access pattern generation process is realized by using the access pattern stored in the generation process storage unit 24.

That is, in the generation of the access pattern, since the access pattern is generated in order from the cluster having the smallest number of users, when the number of users of the cluster used for generating the access pattern for the j-th cluster is D _j , D _k <D _l (where k <l). Further, when the newly generated access pattern includes a certain access pattern in the subsequence, the number of users sharing the original k-th access pattern candidate A is User (A), and the newly generated access pattern is newly generated. Assuming that the number of users sharing the l-th access pattern candidate B is User (B), User (A) ≧ User (B) holds. Therefore, the support degree Sup of the access pattern candidate A
Support (B) for port (A) and access pattern candidate B
Has the relationship shown in the following equation.

[0070]

(Equation 3)

That is, among the access pattern candidates,
An access pattern having an access pattern included in the set LC held in the generation process holding unit 24 as a partial pattern can be removed from Li _{+ 1} as an access pattern candidate that does not satisfy the support threshold (step S3).

Next, the support is calculated for each access pattern candidate. If the support is changed, it is removed from Li _{+ 1.} If the support or the initial support is less than the threshold, the generation process holding unit 24 Hold with. Further, except for the support of access pattern candidate is the source unless changed from L _i (step S4). And the processing of the above process is L
_{The process} is repeatedly performed until _{i + 1} becomes empty (steps S5 and S7), and finally remaining L ₀ , L ₁ ,..., L _n (n is an integer of 0 or more) are set as the generated access pattern candidates. Output (Step S6).

Next, FIG. 6 is a flowchart of a process for finding a final access pattern by the access pattern finding unit 23 satisfying the support condition based on the access pattern candidates generated as described above. . That is, in the discovery unit 23, first, among the access pattern candidate set generated as described above, selects a set L _i number of hypertext included in the access pattern is the smallest (step S11), and for each of the access pattern candidate in the L _i, add hypertext reachable from hypertext set in access patterns, generating an access pattern set L _{i + 1} (step S12). It should be noted that, in this case, and hypertext included in the initial set (cluster) that is the source of L _i,
The hypertext included in L _k (k> i) is not added.

Next, among the access patterns included in L _{i + 1} , those that include, as a partial pattern, an access pattern held in the generation process holding unit 24 and that did not satisfy the threshold value for the support degree in the past are considered as L patterns. _{It is} removed from _{i + 1} (step S13). Next, for each access pattern candidate of _{Li + 1} , the support level and the minimum support level are calculated, and those that do not exceed both thresholds are excluded from the _{Li + 1} and stored in the generation process storage unit 24. , for which exceeds both thresholds, except access pattern candidate became its original from L _i (step S14). And L
_{i + 1} until empty and and not empty L _k (k> i + 1) does not exist (step S15, step S16, step S18, step S19), performed by repeating the above processing processes, the final An access pattern set that satisfies the conditions of support and minimum support is output as a generated access pattern (step S17).

Here, in the case of the present embodiment, an example of the relationship between access pattern candidates searched in the process of generating an access pattern from Pt (C, D, F) which is one of the species, for example, is shown. It is shown in FIG. In the figure, each alphabet corresponds to each hypertext shown in FIG. Further, an access pattern surrounded by a double square means an access pattern whose support degree is not changed, an access pattern surrounded by a dashed square means an access pattern whose support degree is changed, The used access pattern means an access pattern that does not satisfy the support condition (the minimum support threshold or the support threshold).

An access pattern surrounded by a dashed-line square and further marked with an X mark means a pattern which is found to not satisfy the conditions of minimum support or support when the number of users is counted. The access pattern of only means a pattern determined not to satisfy the condition of the support based on an access pattern set that did not satisfy the condition of the past support without calculating the support.

In the example shown in the figure, Pt (B, C, D, F) is generated by the access pattern finding unit 22 which does not change the support. If Pt (B, C, D, F) does not change the support, then Pt (B, C, D, F) and Pt (A, B, C, D, F) and Pt (B, G, C, D, F
) Is generated. For Pt (A, B, C, D, F) and Pt (B, G, C, D, F), Pt (B, G, C, D, F) is the minimum support threshold or support threshold condition Is not satisfied, the Pt (B, G, C, D, F) is transferred to the generation process holding unit 24, and Pt (A, B, C, D, F) is output as an access pattern candidate.

Next, an access pattern candidate that satisfies the support condition is further searched for access pattern candidates. That is, from Pt (A, B, C, D, F), Pt (A, B, G, C,
D, F) and Pt (A, B, K, C, D, F) are generated as access pattern candidates. Pt (A, B, G, C, D, F) is generated by the generation process holding unit 24
Since the Pt (B, G, C, D, F) held in is included as a partial pattern, it is removed without calculating the support. Pt (A, B,
K, C, D, F), the minimum support and support are calculated,
It is excluded if the support condition is not met. In the example shown in the figure, Pt (C, D, F, E) and Pt (B, C, D, E, F) are not generated because the hypertext ( Here, E) is not added in a later generation process. In addition, the species Pt (F, I, L) and Pt (C, D, F)
After the access pattern is generated from, the access pattern is generated based on the cluster {B, G, K} with the next smallest number of users.

The access pattern holding unit 6 has a function of holding the access pattern generated by the access pattern generation unit 5 as a set of the seed from which the access pattern is generated and the access pattern. The access pattern presenting unit 7 has a function of presenting the access pattern held in the access pattern holding unit 6 to an analyst (user). In the present embodiment, the following two modes of presentation are provided. Here is an example.

First, an example of a presentation process performed by the access pattern presentation unit 7 will be described. FIG. 8 is a flowchart of the presentation process in this example of the mode. In this example, the access pattern presenting unit 7 determines that the set C of the set of the access pattern and the seed becomes empty (step S2).
1) The following processing is performed. That is, first, a set Pa of the access pattern and the seed that is the access pattern based on the seed generated from the lowest cluster in the hierarchical relationship of the cluster and that shares the largest number of users is selected (step S22). .

Next, it is determined whether or not the access pattern of the selected set Pa is a pattern including the seed of another higher-level hierarchical cluster in the set C (step S23), and it is determined that the pattern is the pattern. In this case, the access pattern (discovered access pattern) and the number of access users pertaining to the set Pb of the pattern and the seed, and the selected set
The access pattern of Pa (original access pattern) and the number of accessing users are presented to the analyst (step S2).
4). It should be noted that after such a presentation is made or the access pattern of the set Pa selected in the above determination is set C
If it is determined that the pattern does not include the seed of the other higher-level hierarchical cluster, the selected set Pa is excluded from the set C (step S25), and the same processing is repeated.

In the method of presenting such an access pattern, a hypertext in which more detailed information is described
When the user browses (the hypertext included in the lower cluster), it can be considered that the access pattern has changed. For example, in the configuration shown in FIG. 3, the access pattern Pt (A, B, G) is found from Pt (B, G), and the access pattern Pt (F, I, L) and Pt (C, D, F) are Access pattern Pt
(A, B, G, K, F, I, L) and Pt (A, B, C, D, F) are found. In this case, the hypertexts B, G, and K are pages on which information of different products is described, respectively.
C, D, and F are pages on which detailed information of the product B is described, respectively, and hypertexts F, I, and L are pages on which information on the specification and price of the product is described.

Then, although there is a high correlation between the hypertexts B and G for the product itself, it can be said that a user who has watched the product in more detail has no correlation between the products and sees only the product B. On the other hand, it can be seen that there is a high correlation between the user's browsing behavior for the specifications and prices of all three products. This means that if the user is interested in both products B and G, but sees the details of interest in product B, something that makes the user feel that they do not need to see the details of product G is It is believed that there is. In addition, only the hypertext B, G, and K pages have a small correlation,
The fact that the correlation is increased by limiting the users to those who saw the specifications and prices suggests that, as one possibility, some of the user groups may be comparing products incorrectly. In any case, the access patterns of these users are likely to give suggestions about the problems involved in the site, and are considered to assist in site analysis.

As another specific example using such an embodiment, for example, in a site that sells a product, the hierarchy of the hypertext displayed when the user tries to purchase the product in the thesaurus hierarchy, By separately providing a hierarchy that leads to the purchase of a product and classifying the hypertexts, it is possible to discover the relationship between the user's willingness to purchase and the actual purchasing activity and present it to the user.

Next, another example of the presentation process performed by the access pattern presentation unit 7 will be described. FIG. 9 is a flowchart of the presentation process in this example of the mode. In this embodiment, the access pattern presentation unit 7 performs a presentation process based on the specification of the access pattern received from the analyst (user) via the reception unit. That is, first, for the access pattern selected by the user, the hypertext in the access pattern is added to the display hypertext set D (step S31).

Next, a set L of hypertexts in the cluster including the selected access pattern type is extracted (step S32), and the hypertexts in the set L, which are hypertexts constituting the type, are extracted. Is added to the display hypertext set D (step S33). Then
Any one of the hypertexts in the set L from the hypertexts constituting the seed
The hypertext reachable via one hypertext is added to the display hypertext set D (step S34).

In this way, for example, the hypertext in the access pattern, which is a hypertext linked by a link from the hypertext included in the cluster that is the source of the seed, is defined using the page with the largest number of accessing users as the root. It is expanded and displayed in a tree structure (step S35).
In this display processing, for example, by marking hypertext included in the access pattern specified by the analyst on the tree structure, the hypertext can be distinguished from the hypertext not included in the access pattern. (Step S3)
6).

Here, FIG. 10 shows an example in which the access pattern Pt (A, B, C, D, F) is designated and displayed in the site configuration shown in FIG. 3, and the access pattern Pt (A, B, G,
FIG. 11 shows an example in which display is performed by designating (K, F, I, L). FIG.
In the example of the tree structure shown in FIG. 0, for Pt (A, B, C, D, F),
Since the hypertext E in the cluster from which the seed Pt (C, D, F) is based is directly linked to the hypertext C by a link, it is added to the hypertext to be displayed. In the example of the tree structure shown in FIG. 11, Pt (A, B, G, K, F, I,
L), the three hypertexts C in the cluster,
H, J are hypertexts B, G, K in the access pattern
Have been added to the hypertext to be displayed because they are directly linked to each other.

In the examples shown in FIGS. 10 and 11, hypertexts included in the access pattern specified by the analyst are indicated by colored circles. Is indicated by a white circle. Note that the display form of each hypertext is not necessarily limited to the one shown in the present embodiment, and for example, a method of displaying a part of information in the hypertext may be used.

As a method of presenting a hypertext in a distinguished manner, for example, a link included in an access pattern specified by an analyst is distinguished from others (for example, a line representing a link included in a specified access pattern is thickened). (Or the like). With such a presentation process, it is possible to find a difference between the classification considered by the user and the classification of the site builder, which helps in considering the link structure of the site.

As described above, in the hypertext access pattern analysis apparatus according to the present embodiment, the relationship between the intention of the site configuration of the analyst and the interest of the user is shown by considering the contents of the content, for example. Pattern scrutiny can be facilitated, and for example, hypertext systems with a relatively small number of accesses, such as hypertexts with a relatively small number of accesses, as well as a set of hypertexts with a large number of accesses in a hypertext system.
Characteristic access patterns can be found among users visiting the hypertext in the same area, and, for example, access pattern generation can be made more efficient.

Although the preferred embodiment of the present invention has been described in the present embodiment, the configuration of the hypertext access pattern analyzing apparatus according to the present invention is not necessarily limited to that shown in the present embodiment. Any configuration may be used. For example, in the present embodiment, as a preferable mode, various processes performed by the hypertext access pattern analysis apparatus according to the present invention include, for example, a control in which a processor is stored in a ROM in a hardware resource including a processor and a memory. Although the configuration is controlled by executing the program, for example, each functional unit for executing the processing may be configured as an independent hardware circuit.

The present invention also relates to a floppy (registered trademark) disk or CD-ROM storing the above-described control program.
It can be understood as a computer-readable storage medium such as a ROM, and the processing according to the present invention can be performed by inputting the control program from the storage medium to the computer and causing the processor to execute the control program.

[0094]

As described above, according to the hypertext access pattern analysis apparatus and storage medium of the present invention, an access pattern is generated for each cluster classified based on the contents of each hypertext, and the access pattern is generated. By presenting patterns to analysts,
For example, by comparing the access pattern with the classification, it is possible for the analyst to discover the relationship between the intention of the site configuration and the interest of the user.

In the hypertext access pattern analysis device according to the present invention, not only the minimum support (or the same order minimum support) as described above, but also the support (or the same order support) as described above. Since the access pattern is generated taking into account, for example, an access pattern can be generated according to the number of users accessing the hypertext in the classified cluster, and the access of the user in the hypertext system can be performed. Even in the case of a large deviation, it is possible to find a characteristic pattern in each region.

The hypertext access pattern analysis apparatus according to the present invention executes access pattern candidate generation processing and access pattern generation processing relating to various hypertext sets in ascending order of the number of users who have accessed, and has already performed minimum support. For a hypertext set including a pattern whose degree or support is determined to be less than the threshold, without performing the determination, because it is excluded from access pattern candidates and access patterns,
The efficiency of the access pattern generation process can be improved.

Further, in the hypertext access pattern analysis apparatus according to the present invention, for example, the order of the number of users who have accessed the hypertext included in the hierarchy of the cluster to which the seed hypertext set belongs is deeper, and The access pattern including the species other than the species to which the user belongs and the access pattern generated from the species other than the species to which the subject belongs are presented to the analyst. For example, the hypertext included in the designated access pattern and another hypertext are included. It is effective for the analyst because the presentation mode with the text is made different and the predetermined hypertext to be presented is presented to the analyst in a tree structure with the hypertext with the largest number of accesses of the user as the root. Information can be presented in an easy-to-view form.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of a system including a hypertext access pattern analysis device according to an embodiment of the present invention.

FIG. 2 is a diagram showing an example of data of a thesaurus.

FIG. 3 is a diagram showing an example of a site configuration.

FIG. 4 is a diagram illustrating a configuration example of an access pattern generation unit.

FIG. 5 is a diagram illustrating an example of a procedure of a process of finding an access pattern candidate whose support level does not change.

FIG. 6 is a diagram illustrating an example of a procedure of a process of finding an access pattern satisfying a support condition.

FIG. 7 is a diagram illustrating an example of generating an access pattern.

FIG. 8 is a diagram illustrating an example of a procedure of an access pattern presentation process.

FIG. 9 is a diagram illustrating an example of a procedure of an access pattern presentation process.

FIG. 10 is a diagram showing a presentation example of an access pattern.

FIG. 11 is a diagram showing a presentation example of an access pattern.

FIG. 12 is a diagram illustrating an example of an access pattern.

FIG. 13 is a diagram illustrating an example of an access pattern.

[Explanation of symbols]

1. Web server 2. Access history preprocessing unit
3 ・ ・ Hypertext classification section 、 4 ・ ・ Order section 、
5 ··· Access pattern generation unit, 6 ··· Access pattern holding unit, 7 ··· Access pattern presentation unit, 11 ·
・ Content providing unit, 12 ・ ・ Access history recording unit, 2
1. Discovery part of simultaneous access pattern in cluster, 2
2 ······································································································· 23

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Yohei Yamane 430 Sakai-Nakai-cho, Ashigara-gun, Kanagawa Green Tech Naka Fuji Xerox Co., Ltd. F-term (reference) 5B075 ND02 ND36 NR12 PQ02 PR03 5B082 AA11 BA09 CA13

Claims (7)

[Claims] 1. Hypertext classification means for classifying a plurality of hypertexts provided by a hypertext system into a plurality of clusters based on the content of each hypertext, and an access history detecting an access history of each user to the hypertext. Detecting means, based on the detected access history for each user, for each cluster, in a hypertext set formed by combining a plurality of hypertexts included in the cluster,
A hypertext set in which the percentage of users who accessed all hypertexts included in the hypertext set at the same access is equal to or greater than a predetermined value as a seed, and a hypertext linked by a link to the hypertext included in the seed hypertext set A hypertext access pattern analysis device, comprising: an access pattern generation unit configured to generate an access pattern formed by adding text to the hypertext set; and an access pattern presentation unit configured to present the generated access pattern. . 2. The hypertext access pattern analysis apparatus according to claim 1, wherein the number of users accessing the hypertext provided by the hypertext system is the same as all the hypertexts included in the target hypertext set. The ratio of the number of users who accessed at the time of access is set as the minimum support, and all the hypertexts included in the target hypertext set with respect to the number of users accessing the hypertext included in the target cluster at the same access Using the ratio of the number of users who have accessed the support as the support, the access pattern generating means sets the minimum support to a predetermined minimum support threshold in a hypertext set formed by combining a plurality of hypertexts included in the target cluster. Above A hypertext set whose support level is equal to or higher than a predetermined support level threshold and the number of included hypertexts is equal to or higher than a predetermined number is used as a seed, and various hypertext sets are included in the seed hypertext set. Hypertext set obtained by adding hypertext directly linked to the hypertext to the hypertext set, and a hypertext set obtained by similarly adding one or more hypertexts to the hypertext set Among the hypertext sets whose support is the same as the support of the hypertext set that is the seed, as access pattern candidates, and for each access pattern candidate, a direct link to the hypertext included in the access pattern candidate is made. Hypertext connected by In a hypertext set added to the access pattern candidate and a hypertext set in which one or more hypertexts are similarly added to the hypertext set, the minimum support is equal to or greater than a predetermined minimum support threshold. A hypertext access pattern analysis apparatus, wherein a hypertext set whose support level is equal to or higher than a predetermined support level threshold is set as an access pattern. 3. The hypertext access pattern analysis device according to claim 2, wherein the access pattern generation means performs an access pattern candidate generation process and an access pattern generation process regarding various hypertext sets as a seed hypertext set. Are executed in ascending order of the number of users who have accessed the hypertext included in the access pattern candidate generation process and the access pattern generation process. In the process of the access pattern candidate generation process and the access pattern generation process, at least the minimum support or the support of the added hypertext set If one of them is less than the predetermined threshold, information on the under-threshold specifying the combination pattern of the hypertexts constituting the hypertext set is stored, and the access pattern candidate generation processing and the access pattern generation processing are performed. The hypertext set is excluded from the access pattern candidates or access patterns in response to the additional hypertext set including all the hypertexts of the combination pattern specified by the stored information less than the threshold value. A hypertext access pattern analysis apparatus. 4. The hypertext access pattern analysis apparatus according to claim 1, wherein the number of users accessing the hypertext provided by the hypertext system is the same as all the hypertexts included in the target hypertext set. The ratio of the number of users who accessed at the time of access is set as the minimum support, and all the hypertexts included in the target hypertext set with respect to the number of users accessing the hypertext included in the target cluster at the same access Using the percentage of the number of users who accessed the document as support, the same access to all the hypertexts included in the target hypertext set relative to the number of users who accessed the hypertext provided by the hypertext system The ratio of the number of users who accessed in the same order is the same order minimum support, and all the hypertexts included in the target hypertext set for the number of users who accessed the hypertexts included in the target cluster The access pattern generation means sets the ratio of the number of users who accessed in the same order at the same access to the same order as the same order support level, and sets the minimum among the hypertext sets formed by combining a plurality of hypertexts included in the target cluster. A hypertext set whose support is equal to or greater than a predetermined minimum support threshold and whose support is equal to or greater than a predetermined support threshold is selected. The same-order support is equal to or more than the minimum support threshold and the same-order support is a predetermined same-order support The hypertext set that is the above and the number of included hypertexts is a predetermined number or more is used as a seed, and various hypertext sets are directly linked to the hypertext included in the seed hypertext set by a link. In the hypertext set obtained by adding the obtained hypertext to the hypertext set and the hypertext set obtained by further adding one or more hypertexts to the hypertext set in the same manner, the same order support is the same. A hypertext set having the same order of support as the seed hypertext set is regarded as an access pattern candidate, and for each access pattern candidate, a hypertext directly connected to the hypertext included in the access pattern candidate by a link. The access In the hypertext set added to the hypertext candidate and the hypertext set obtained by further adding one or more hypertexts to the hypertext set, when the minimum support is equal to or more than a predetermined minimum support threshold, A hypertext set whose support level is equal to or higher than a predetermined support level threshold is selected, and in the selected hypertext set, the same order minimum instruction is equal to or higher than a predetermined same order minimum support level threshold and the same order support is set. A hypertext access pattern analysis device, wherein a hypertext set whose degree is equal to or greater than a predetermined same order support degree threshold is used as an access pattern. 5. The hypertext access pattern analysis device according to claim 1, wherein the hypertext classification unit classifies the plurality of hypertexts into a plurality of clusters having a hierarchical relationship, The access pattern presenting means is arranged so that the number of users who have accessed the hypertext included in the access pattern in the descending order of the hierarchy of the cluster to which the hypertext set serving as the seed of the access pattern belongs and the number of the users other than the species to which the own belongs belongs. A hypertext access pattern analysis apparatus for presenting an access pattern including a species and an access pattern generated from a species other than the species to which the subject belongs. 6. The hypertext access pattern analysis device according to claim 1, further comprising: a receiving unit that receives designation of an access pattern from a user, wherein the hypertext presenting unit receives the designated access pattern. The hypertext included in the access pattern is directly linked to the hypertext included in the hypertext set as the seed among the hypertexts included in the cluster to which the hypertext set as the seed of the access pattern belongs. And any one hypertext included in the access pattern from the hypertext included in the hypertext set as the seed among the hypertexts included in the cluster to which the hypertext set as the seed of the access pattern belongs Texture The target of presentation is hypertext that can be reached via the designated access pattern, and the presentation mode of the hypertext included in the specified access pattern is different from that of other hypertexts, so that the number of user accesses is maximized. A hypertext access pattern analysis device characterized by presenting a hypertext to be presented in a tree structure with the hypertext as a root. 7. A storage medium in which a program to be executed by a computer is stored readable by input means of the computer, the program stores a plurality of hypertexts provided by a hypertext system based on the contents of each hypertext. A process of classifying into a plurality of clusters, a process of detecting an access history of each hypertext user, and a plurality of hypertexts included in the cluster for each cluster based on the detected access history of each user Of the hypertext set consisting of all the hypertexts included in the hypertext set as a seed, and a hypertext set in which the ratio of users accessing the hypertext at the same access is equal to or greater than a predetermined value is included in the seed hypertext set. Hyperte A storage medium for causing a computer to execute a process of generating an access pattern formed by adding hypertext linked to a strike to a hypertext set to the hypertext set, and a process of presenting the generated access pattern. .