JP2008176398A - Information processing apparatus and method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately recommend items for each user. <P>SOLUTION: An item I-01 is purchased from the purchase histories of a plurality of users, and information showing that there are many users who purchase the item I-03 is generated, and a time-series pattern relating to the purchase of the item is generated. When the item I-01 is purchased by a user U001, the item I-03 which is highly likely to be purchased after the item I-01 is purchased is selected by referring to the time-series pattern, and recommended to the user U001. This invention can be applied to a server which is managed by a store which sells predetermined items on a network. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an information processing apparatus and method, and a program. In particular, when recommending other items related to the item to a user who purchased a predetermined item, the recommended item is purchased by the other user. The present invention relates to an information processing apparatus and method, and a program, which are items selected in consideration of the order pattern of items.

  2. Description of the Related Art Conventionally, a so-called content personalization invention has been proposed in which content such as television programs and music and items such as products are recommended based on customer preferences (see, for example, Patent Document 1). ). For content personalization, a technique called collaborative filtering (CF) and a technique called content-based filtering (CBF) are widely used.

JP 2004-194107 A

  As described above, an item or content different from the purchased item or the browsed content is recommended to the item purchaser or the content viewer.

  For example, when a book is purchased via the Internet, other users who have purchased the book recommend the purchased book. Such recommendations include, for example, collaborative filtering (P. ResnIck, N. Iacovou, M. Suchak, P. Bergstrom, and J. RIedl. “GroupLens: Open ArchItecture for CollaboratIve FIlterIng of Netnews.” Conference on Computer Supported CooperatIve Work , pp. 175-186, 1994.).

  According to the recommendation by collaborative filtering, the recommendation can be realized by the similarity of the history even if there is no metadata, but since the processing is performed based on the similarity of the history, items that are not related to each other There was a possibility of being recommended.

  Even if there is a relationship between items, if the user cannot find the relationship, the user cannot understand why the item was recommended, and it may be difficult to accept the recommendation.

  For example, even if a beginner (starter) purchases a book for beginners, if a book purchased by another user is a book for advanced users, a book for advanced users should be recommended to beginners was there. Such recommendations are not considered appropriate.

  The present invention has been made in view of such a situation, and makes it possible to perform more appropriate recommendation.

  An information processing apparatus according to an aspect of the present invention generates a time-series history when items are arranged in order of purchase from an access history associated with an item accessed by a user and an access time to the item. Means, a frequency extracting means for counting the number of transitions from the first item to the second item using a plurality of time-series histories, and extracting a frequent transition, and the frequency extracting means Based on the pattern generation means for generating a transition pattern between a plurality of items from the transition, and the item accessed by the user who recommends the information based on the rule for selecting the recommended information, the item Determining means for determining an item positioned before or after the position in the pattern as an item of the recommended information; Obtain.

  The frequency extraction further comprises grouping means for grouping the users into a plurality of groups from the access history, and specifying means for specifying which of the plurality of groups the user recommending the information belongs to The means may extract the transition using a time-series history of users belonging to the group.

  The rule may be a rule for recommending an item extracted as having a high frequency transition by the extraction means.

  The rule may be a rule that recommends an item whose ratio of the frequency of the item extracted as the frequency of the frequency extracted by the extraction unit to the number of users is equal to or higher than a preset ratio. .

  The rule is a user in which the number of items is recorded in the access history of a user who is a recommendation target and the number of items is less than a certain number, and the earlier item of the pattern is included in the history. If it is determined that the rule is a rule that recommends an item relatively earlier in the pattern.

  In the rule, the number of items is recorded in the access history of the user who is the target of recommendation, and the earlier item of the pattern is included in the history, and with respect to other genres If it is determined that the user can see the same history, the access history of similar users is extracted from a group different from the group determined to belong to the user, and the extracted history is included in the extracted history. It can be a rule that recommends an item.

  The rule is a user whose number of items is recorded in a certain number or more in the access history of the user who is a recommendation target and the later item of the pattern is included in the history. If it is determined that the item is a rule that recommends an item relatively later in the pattern.

  According to the rule, a user whose number of items is recorded in the access history of a user who is a recommendation target, and the later item of the pattern is included in the history. If it is determined that the item is a rule, it is possible to make a rule that recommends an item relatively later in the pattern than a predetermined number.

  The rule is a user whose number of items is recorded in a certain number or more in the access history of a user who is a recommendation target, and the earlier item of the pattern is included in the history. If it is determined that the number of items in the pattern is relatively earlier than the predetermined number, the rule may be recommended.

  The rule is a ranking of items arranged in the order in which the number of items recorded in a plurality of the access histories is larger than a certain number in the access history of the user to be recommended. If it is determined that the user contains a lot of items that do not fall in the top of the ranking, a rule that recommends items of lower rank among similar patterns be able to.

  An information processing method according to an aspect of the present invention generates a time-series history when items are arranged in order of purchase from an access history in which an item accessed by a user and an access time to the item are associated with each other Using a step and a plurality of time-series histories, the number of transitions from the first item to the second item is counted, a frequency extraction step for extracting a high frequency transition, and extraction by the processing of the frequency extraction step Based on the pattern generation step for generating a transition pattern between a plurality of items and the rule for selecting information to be recommended based on the item accessed by the user who recommends the information, An item positioned before or after the position of the item in the pattern is determined as the item of recommended information. And a determining step of.

  A program according to one aspect of the present invention generates a time-series history when items are arranged in order of purchase from an access history in which an item accessed by a user and an access time to the item are associated. Using a plurality of time-series histories, the number of times of transition from the first item to the second item is counted, and a frequency extraction step for extracting a high frequency transition and the frequency extraction step are extracted. From the transition, a pattern generation step for generating a transition pattern between a plurality of items and a rule for selecting recommended information, based on an item accessed by a user who recommends the information, The item located before or after the position in the pattern is determined as the item of recommended information To execute processing including the determination step that the computer.

  In the information processing apparatus, method, and program according to one aspect of the present invention, a time series when items are arranged in the order of purchase from an access history in which an item accessed by a user and an access time to the item are associated with each other History is generated, a plurality of time-series histories are used, the number of times of transition from the first item to the second item is counted, and transitions whose count number is equal to or greater than a predetermined value are extracted and extracted A transition pattern in which a plurality of items exist is generated from the transition. When information is recommended to the user, an item accessed by the user is used as a reference, and an item positioned before or after the position in the pattern of the item is recommended.

  According to one aspect of the present invention, it is possible to recommend a more appropriate item.

  Embodiments of the present invention will be described below. Correspondences between constituent elements of the present invention and the embodiments described in the specification or the drawings are exemplified as follows. This description is intended to confirm that the embodiments supporting the present invention are described in the specification or the drawings. Therefore, even if there is an embodiment which is described in the specification or the drawings but is not described here as an embodiment corresponding to the constituent elements of the present invention, that is not the case. It does not mean that the form does not correspond to the constituent requirements. Conversely, even if an embodiment is described here as corresponding to a configuration requirement, that means that the embodiment does not correspond to a configuration requirement other than the configuration requirement. It's not something to do.

  When the information processing apparatus according to one aspect of the present invention (for example, the server 2 in FIG. 3) arranges items in order of purchase from an access history in which an item accessed by a user and an access time to the item are associated with each other. The number of times of transition from the first item to the second item is counted using a generating means (for example, the item extraction unit 56 in FIG. 3) that generates the time-series history and a plurality of time-series histories, A pattern of transition between a plurality of items from the frequency extraction means for extracting a high-frequency transition (Etto Enami, transition count section 57, transition extraction section 58 in FIG. 3) and the transition extracted by the frequency extraction section The user who recommends the information accesses based on the pattern generation means (for example, the time-series generation unit 59 in FIG. 3) and the rule for selecting the recommended information. Determination means (for example, a recommendation rule selection unit 60 and a recommended item selection unit 61 in FIG. 3) that determines an item positioned before or after the position of the item in the pattern as a reference item based on the item that has been received. ).

  From the access history, a grouping unit (for example, the grouping unit 54 in FIG. 3) for grouping the users into a plurality of groups and a group to which the user who recommends the information belongs belong to the group An identification unit (for example, a similar group determination unit 55) may be further provided.

  Embodiments of the present invention will be described below with reference to the drawings.

[System configuration]
FIG. 1 is a diagram showing a configuration of an embodiment of a system to which the present invention is applied. A server 2 and terminals 3-1 to 3 -N are connected to the network 1. The system shown in FIG. 1 is a system that provides predetermined information to a user. Also, analysis for generating information to be provided is performed. The network 1 includes the Internet, a LAN (Local Area Network), and the like.

  The server 2 accumulates information to be provided and generates information for providing. Terminals 3-1 to 3-N are user-side terminals. For example, the terminal is operated when the user purchases and views a predetermined item by the terminal 3-1.

  In the following description, when it is not necessary to individually distinguish the terminals 3-1 to 3-N, they are simply described as the terminal 3. In FIG. 1, only one server 2 is shown, but it is of course possible to provide a plurality of servers 2.

  In this embodiment, an item is a product that is bought and sold on the Internet. For example, a CD (Compact DISc), a book, furniture, etc., a purchase procedure is performed at a store that is open on the Internet, and the item is delivered. It is assumed that the product is a product that is purchased and delivered over the Internet, such as an incoming product, music data, or a program. In addition, the item includes products that are not subject to trading, and includes, for example, information such as squeeze, program guide information for television programs, and the like.

  In the system shown in FIG. 1, the predetermined information provided to the user is information for recommending an item. For example, for the user who purchased the CD, other information related to the CD is available. Information for recommending the CD. In the following description, for example, the description “recommend an item” means that information on a predetermined product is provided to the user.

[About server configuration and operation]
FIG. 2 is a diagram illustrating an internal configuration example of the server 2. A CPU (Central Process Ing UnIt) 21 executes various processes according to a program stored in a ROM (Read Only Memory) 22 or a storage unit 28. A RAM (Random Access Memory) 23 appropriately stores programs executed by the CPU 21 and data. The CPU 21, ROM 22, and RAM 23 are connected to each other by a bus 24.

  An input / output interface 25 is also connected to the CPU 21 via the bus 24. The input / output interface 25 is connected to an input unit 26 including a keyboard, a mouse, and a microphone, and an output unit 27 including a display and a speaker. The CPU 21 executes various processes in response to commands input from the input unit 26. Then, the CPU 21 outputs the processing result to the output unit 27.

  The storage unit 28 connected to the input / output interface 25 includes, for example, a hard disk, and stores programs executed by the CPU 21 and various data. The communication unit 29 communicates with an external device (for example, the terminal 3) via the network 1. A program may be acquired via the communication unit 29 and stored in the storage unit 28.

  The drive 30 connected to the input / output interface 25 drives a removable medium 31 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, and drives the program or data recorded therein. Get etc. The acquired program and data are transferred to and stored in the storage unit 28 as necessary.

  Since the terminal 3 can be basically configured with the same configuration as the server 2, the description thereof is omitted here.

[Functions of the server in the first embodiment]
FIG. 3 is a diagram for explaining functions of the server 2 according to the first embodiment. The server 2 has a function for recommending items. The server 2 has a function of executing a grouping process as a process for recommending an item, generating time-series information, and selecting an item for recommendation using the information. Specifically, it has a function as shown in FIG.

  The server 2 includes a history table management unit 51, an item information table management unit 52, a grouping information storage unit 53, a grouping unit 54, a similar group determination unit 55, an item extraction unit 56, a transition count unit 57, a transition extraction unit 58, a time series The configuration includes a generation unit 59, a recommendation rule selection unit 60, a recommended item selection unit 61, and a recommendation information providing unit 62.

  The history table management unit 51 manages a history table 81 as shown in FIG. 4 in which an item purchased by a predetermined user is associated with the date and time when the item was purchased. Referring to FIG. 4, in the history table 81, for example, information that the user “U001” purchased the item “I-01” at the date “2005/08/22/9: 25”. Are described in association with each other.

  Information described in the history table 81 managed by the history table management unit 51 is supplied from a server (not shown) managed by a store that sells items. The supply of information may be performed via the network 1 (FIG. 1), or may be performed by being written on a paper medium or the like and sent.

  Note that, as described above, the description will be made assuming that the history of the purchased item is managed here, but it is also possible to record that the item or its information has been accessed as a history, and the record may be used as the history table 81. In other words, the history table 81 may be managed as an access history table in which a user who has accessed predetermined information is associated with the time of access.

  The server 2 itself may be a server managed by a store that sells items. In such a case, the description of the history table 81 may be updated when the item is sold on the server 2 itself.

  Here, the description will be continued assuming that the information is described in the history table 81 when the item is purchased. In other words, the history table management unit 51 has the history table 81 related to the item purchased by the user. Continue to explain as managing. However, for example, not only the purchased item but also the information regarding the item for which the user has browsed the information may be described in the history table 81 without purchasing.

  The item information table management unit 52 manages an item information table 82 in which a predetermined item and information related to the item are associated as shown in FIG. Referring to FIG. 5, for example, when the item I-01 is a book, information such as the author and the issue date is described as information of the item I-01.

  The grouping information management unit 53 manages the results grouped by the processing by the grouping unit 54. The grouping unit 54 refers to the history table 81 managed by the history table management unit 51, groups users or items, and generates a group having some common points.

  The similar group determination unit 55 determines to which group the user who recommends the item belongs, by referring to the group information managed by the grouping information management unit 53.

  The item extraction unit 56 refers to the history table 81 managed by the history table management unit 51 and extracts items in which a history with a certain frequency or more remains. “Items with a history of a certain frequency or more” means that, as described above, when the history table 81 managed by the history table management unit 51 is a history of purchased items, attention is paid to one item. When the item is purchased, the number of times the item is purchased is a predetermined number or more. In other words, since the number of times an item has been purchased is often the same as the number of users who purchased the item, depending on the item, when focusing on one item, it may be the number of users who purchased the item. .

The transition counting unit 57 counts transitions between items extracted by the item extracting unit 56. For example, when the item I-02 is purchased after the item I-01 and the item I-02 are extracted by the item extraction unit 56 and the item I-01 is purchased,
Item I-01 → Item I-02
The number of transitions from item I-01 to item I-02 is counted as one. The transition counting unit 57 counts the number of such transitions for each item.

  The transition extracting unit 58 extracts a transition in which the number of transitions counted by the transition counting unit 57 is a predetermined value or more.

The time series generation unit 59 rearranges the transitions extracted by the transition extraction unit 58 in time series, and generates time series data. For example, if the transition from item I-01 to item I-02 is extracted and the transition from item I-02 to item I-03 is extracted,
Item I-01 → Item I-02 → Item I-03
The time series data (time series pattern) is generated.

  When recommending an item to the user, the recommendation rule selection unit 60 selects a rule related to selection of the recommended item. As will be described later, a plurality of rules are set according to the characteristics of the user who recommends an item, and one rule is selected from the rules, and an item recommended according to the rule is selected. Make sure that appropriate items are recommended.

  The recommended item selection unit 61 refers to the item information table 82 managed by the item information table management unit 52 based on the rule selected by the recommendation rule selection unit 60, and selects an item to be provided to the user.

  The recommended information providing unit 62 provides information on the item selected by the recommended item selecting unit 61 to the user side. For example, when providing information to the terminal 3-1 connected via the network 1, the recommended information providing unit 62 provides information on items to be recommended via the network 1.

  The operation of the server 2 having such a function will be described with reference to the flowchart of FIG. Assuming that the processing of the flowchart of FIG. 6 is started, the history table management unit 51 and the item information table management unit 52 each include a history table 81 and item information in which a predetermined number of information is described. It is assumed that the table 82 is being managed.

  In step S21, the grouping unit 54 performs a grouping process. The grouping process is performed at a time prior to recommending the item to the user, and the result is managed by the grouping information management unit 53.

  Here, the grouping process in step S21 will be described. As a first grouping method, a method may be applied in which a history is dimensionally compressed based on principal component analysis or PLSA, and a plurality of groups are generated by clustering the results. The processing of the grouping unit 54 when performing grouping by clustering will be described with reference to the flowchart of FIG.

  In step S51, the grouping unit 54 acquires information. The information acquired in step S51 is the history table 81 (FIG. 4) managed by the history table management unit 51. For the subsequent processing, from the history table 81, a history table 81 ′ as shown in FIG. 8 (denoted with a dash (′) to distinguish it from the history table 81 shown in FIG. 4). Is created.

  The history table 81 ′ illustrated in FIG. 8 is a table in which an item and a user who has purchased the item are associated with each other. For example, in the history table 81 ′ illustrated in FIG. 8, the item I-01 is purchased by the user U001, the user U002, and the user U003, the item I-02 is purchased by the user U003, and the item I-03 is the user. It states that U001 has purchased.

  In step S51, the history table 81 'as shown in FIG. 8 is acquired. In step S52, multivariate analysis is performed. For example, the quantification management table 83 as shown in FIG. 9 is created by performing multivariate analysis on the information described in the history table 81 ′ as shown in FIG. 8. That is, from the purchase pattern of the history table 81 ′ as shown in FIG. 7, the features for each item that the user prefers to purchase are quantified into several-dimensional data by multivariate analysis such as correspondence analysis and principal component analysis. Is done.

  In step S53, information (numerical values) described in the quantification management table 83 is mapped. For example, consider mapping the numerical values described in the columns of dimension 1 and dimension 2 in the quantification management table 83 shown in FIG. By associating dimension 1 with, for example, the X axis and dimension 2 with, for example, the Y axis, the features (information) of each item described in the quantification management table 83 in two dimensions consisting of the X axis and the Y axis. Can be mapped. When such mapping is performed, for example, a graph as shown in FIG. 10 can be acquired.

  The graph shown in FIG. 10 is an example in which the items described in the quantification management table 83 as shown in FIG. 9 are mapped using the numerical values of the dimension 1 and the dimension 2 for each item. Mapping is performed for the number of items described in the quantification management table 83 (for example, for 100 items). Since such processing is performed, the dimension for performing multivariate conversion may be determined depending on how many dimensions are mapped and the processing described later is executed. For example, when mapping is performed two-dimensionally and processing to be described later is executed, the multivariate transformation may be quantified into two-dimensional data.

  When the graph as shown in FIG. 10 is created, a cluster is determined in step S54 (FIG. 7). For example, the cluster analyzes the graph shown in FIG. 10 and divides the data into clustered parts (parts where points are concentrated), and sequentially assigns the divided areas to cluster 1, cluster 2, and the like. Thus, a cluster is determined.

  Clustering is performed by a technique such as k-means or hierarchy.

  By performing the clustering, for example, the result shown in FIG. 11 is acquired from the graph shown in FIG. The result shown in FIG. 11 shows a state when the cluster is classified into five clusters 1 to 5. Each cluster is set with no overlap. In this way, by setting each cluster so as not to overlap, one cluster can be assigned to one item.

  It is also possible to set so that each cluster may overlap. When clustering is performed using a clustering method called a soft clustering method, the clusters may overlap. When there is an overlap in clusters, for example, a proportion of a plurality of clusters is assigned to one item such that cluster 1 is 90% and cluster 2 is 10%. Here, as shown in FIG. 11, the description will be continued assuming that there is no overlap between the clusters.

  Thus, when a cluster is derived from predetermined information, the derived cluster is stored in step S55 (FIG. 7). When the cluster is determined, the grouping process is completed. Therefore, this cluster information is managed by the grouping information management unit 53 as a grouping result.

  As a second grouping method, a method may be used in which the history is vectorized, several representative patterns are set, the distance when the set pattern is centroided is obtained, and the closest pattern is grouped.

  A grouping method based on vectorization will be described with reference to FIG. In FIG. 12, a black triangle indicates a representative vector, and a black circle indicates a vector of each user. A representative vector is set, and a vector space centered on the representative vector is set. In this way, after creating some representative patterns (groups), it is determined in which group the vector of each user is grouped according to which vector space, and the determination result is The grouping information is managed by the grouping information management unit 53 as grouping information.

  When grouping is performed by vectorization, the history items may be weighted and the grouping may be performed using the weighted result.

  As a third grouping method, a grouping method by selecting items with high frequency from the history of all users may be applied. For example, before performing grouping, a time series pattern described later is generated, and grouping is performed based on the time series pattern. The time series pattern is data in which items are arranged in the order of purchase. A plurality of time series patterns may be generated, and grouping may be performed by making each data into one group.

  As a fourth grouping method, grouping may be performed by genre. When an appropriate genre is assigned to an item, for example, the same genre may be grouped into the same group in the assigned genre.

  Further, in the first to third methods, grouping may be performed on each grouped group so that further subgroups are formed in the group. When subgroups are formed, grouping may be performed by applying the above-described fourth grouping method to the subgroup grouping method.

  The grouping process executed in step S21 (FIG. 6) may be performed based on any of the first to fourth methods.

Documents describing details of methods applicable to the above grouping will be described.
Reference 1: MJZakI, EffIcIently mInIng frequent trees In a forest, Proc. Of the 8th InternatIonal Conference on Kknowledge DIscovery and Data MInIng, pp71-80, 2002.
Reference 2: A.TermIer, MC Rousset, MSebag., TreeFInder A fIrst Step towards XMLDDataMInIng, IEEE ICDM 02, pp450-457, 2002.

  Returning to the description of the flowchart of FIG. 6, the grouping process is executed in step S <b> 21, and when the result is accumulated in the grouping information management unit 53, the process proceeds to step S <b> 22.

  In step S <b> 22, the similar group determination unit 55 specifies a user group similar to the history of a user who is a target for recommending an item. To specify a similar group, a plurality of groups are generated by performing grouping in step S21. Among the groups, the user to be processed determines which group the group belongs to. It is.

  For example, the history of items purchased by the user to be processed is read with reference to the history table 81 managed by the history table management unit 51, and the read history is matched with the history in each group. The group having the highest matching result may be determined to be the group to which the user to be processed belongs.

  Also, for example, when an item is recommended when a user purchases an item, it is determined that the group including the item purchased by the user to be processed is a group to which the user belongs. You may be made to do.

  When the process in step S22 is executed by the similar group determination unit 55, it is determined that the history of the user group to which the users U001 to U005 belong is similar to the history of the user to be processed For example, a time series history table 91 as shown in FIG. 13 is created as the process of step S23.

  The time series history table 91 shown in FIG. 13 is a table in which items purchased by a predetermined user are described in time series. In the time series history table 91, for example, it is described that the user U001 has purchased in the order of item I-01, item I-03, item I-05, item I-09, and item I-10.

  The time series history table 91 shown in FIG. 13 can be created from the history table 81 as shown in FIG. 4, and the history table management unit 51 stores both the history table 81 and the time series history table 91. You may make it manage. Further, since the time series history table 91 is generated from the history of users included in the user groups determined to be similar, information on groups managed by the grouping information management unit 53 is also referred to as appropriate. .

  A time-series history table 91 shown in FIG. 13 is an example of a table created based on the item I-01. That is, this is a table in which users who have purchased the item I-01 are extracted from the user group, and the items purchased by the user are listed in time series from the item I-01. The time series history table 91 shown in FIG. 13 shows an example as a table starting from the item I-01 as described above, but there are cases where other items have been purchased before the item I-01. Before and after I-01, purchased items are listed in chronological order.

  In step S <b> 24, the transition counting unit 57 counts the number of transitions between items, and supplies the result to the transition extracting unit 58. In step S25, the transition extraction unit 58 extracts a transition having a high frequency. Processing in steps S24 and S25 will be described.

  Referring to the time-series history table 91 shown in FIG. 13 again, the number of times counted from the history of five users U001 to U005 is the transition from item I-01 to item I-03. The transition from item I-01 to item I-05 is twice (one person from user U002), and item I-01 to item I-10 is twice (user U001 and user U003). Transition to is one time (one user U004), and transition from item I-01 to item I-11 is one time (one user U005).

  Such a count regarding the transition between items is performed. When paying attention to a predetermined item, only an item purchased immediately after that item may be counted (the above example), or an item purchased during a predetermined period may be counted. (Example shown below).

  The time-series history table 91 shown in FIG. 13 also describes items other than items purchased immediately after the item I-01. The items described in the time-series history table 91 shown in FIG. 13 are items that are purchased in a predetermined period thereafter, based on the time when the item I-01 is purchased.

  When an item described in such a time series history table 91, that is, an item purchased within a predetermined period, is counted, for example, the transition from item I-01 to item I-03 is as follows: 3 times (user U001, user U003, and user U005).

  If only the items purchased immediately after are counted, there may be few users who purchase exactly the same products in the same order, and the result of counting increases as a result of a small number of times such as 0 times or 1 time. there is a possibility. When there is such a possibility, an item purchased within a predetermined period may be processed as a count target.

  Moreover, the period before and after may be set as a predetermined period so that the item purchased before the item made into the process target may be included.

  When an item whose transition count is 3 or more is described from the time-series history table 91 shown in FIG. 13, it is as shown in FIG. The table shown in FIG. 14 is a transition frequency table 92. The transition frequency table 92 is a table that describes the transition frequency for a predetermined item, in this case, the item I-01. Referring to the transition frequency table 92 shown in FIG. 14, the transition from the item I-01 to the item I-03 is three times, the transition from the item I-01 to the item I-05 is three times, and the item I-01 It can be seen that the transition from item I-09 to item I-09 is four times, and the transition from item I-01 to item I-10 is three times.

  In this way, the transition frequency table 92 shown in FIG. 14 is created from the time-series history table 91 shown in FIG. In this way, by narrowing down to items whose number of transitions is equal to or greater than a predetermined number, it is possible to reduce subsequent processing and to appropriately select recommended items.

  When the transition frequency table 92 shown in FIG. 14 is created by executing the processing of step S24 and step S25, the processing proceeds to step S26. In step S <b> 26, the time series generation unit 59 generates a time series pattern. The time series pattern is, for example, a time series table 93 or a pattern that can be read from the time series table 93 as shown in FIG.

  The time series table 93 shown in FIG. The time series table 93 is a table obtained by extracting only items described in the transition frequency table 92 from the time series history table 91. In the example shown in FIG. 15, the time series table 93 is an item purchased by the user who purchased the item I-01 after purchasing the item I-01, and the items purchased by a plurality of users are displayed in time series. It is a table arranged and described.

  For example, the user U001 purchases in the order of item I-01, item I-03, item I-05, item I-09, item I-10, and those items are also purchased by users U002 to U005. It can be seen from the time series table 93 shown in FIG.

  After such a time series table 93 is generated, the process proceeds to step S27. In step S27, the recommendation rule selection unit 60 (FIG. 3) selects a recommendation rule. Here, the rule which the recommendation rule selection part 60 has as an option is demonstrated.

  As rule 1, there is a rule for recommending an item extracted as a high-frequency transition in the process of step S25. When an item to be recommended is selected in accordance with this rule 1, items described in the transition frequency table 92 shown in FIG. 14 are recommended. When the number of items described in the transition frequency table 92 is large, the frequency threshold may be increased to reduce the number of items.

  For example, the transition frequency table 92 shown in FIG. 14 has been described as describing items with a frequency of 3 times or more, but when it is determined that the number of items is large when the frequency is 3 times or more, the frequency is set to 4 The number of items may be narrowed by raising the number of times or more.

  When rule 1 is applied and the transition frequency table 92 shown in FIG. 14 is created, for the user who purchased item I-01, item I-03, item I-05, item I-09, item I-10 is recommended.

  As rule 2, there is a rule that recommends only items with a high time-series frequency. The rule 2 is the same as the rule 1 in that an item recommended with reference to the transition frequency table 92 shown in FIG. 14 is selected. However, in the case of the rule 2, only the item with the highest frequency is selected. Is different from the recommendation target. If there are a plurality of items having the highest frequency, a plurality of items are recommended.

  Rule 2 is applied when the number of items described in the transition frequency table 92 is large, and can be used in combination with Rule 1. For example, rule 1 may be applied basically, and rule 2 may be applied (switched to rule 2) when it is determined that the number of items is large.

  When the rule 2 is applied and the transition frequency table 92 shown in FIG. 14 is created, the item I-09 is recommended to the user who purchased the item I-01.

  As rule 3, a threshold value such as an adjacent period, for example, within one week is set, and items existing in the period are determined to be items purchased at the same time and are recommended at a time. There is a rule. Users who purchased the same item at the same time are considered to have strong connections (something in common), and other items purchased by other users who purchased the same item at the same time are recommended. To.

  When the time series history table 91 shown in FIG. 13 is created in a state where the rule 3 is applied and items within a predetermined period are described, the user who has purchased the item I-01 On the other hand, for example, item I-03 described in the time series history table 91 shown in FIG. 13 is recommended.

  As rule 4, there is a rule that it is determined that the reverse order has a certain frequency or more, and that the order does not matter and is recommended at a time. In the above-described embodiment, it is described that the item purchased at a later time from the time when a predetermined item is purchased as a reference, but the item purchased at the later time is the target of recommendation. Items may also be recommended.

  For example, in the case of simultaneous purchase, if there is a rule that only recommends items purchased at a time later than the time of purchase, the items purchased at the same time may not be recommended. By applying the rule 4, it is possible to make an item for recommendation at the time of simultaneous purchase as well. In other words, rule 4 can be said to be a rule that includes simultaneous purchase or a past period in the adjacent period.

  Rule 4 is similar to rule 3, users who purchased the same item at the same time are considered to have strong connections (something in common), and other users who purchased the same item at the same time Other items purchased by are recommended.

  When the rule 4 is applied and the time series history table 91 shown in FIG. 13 is created, for example, the item I-01 is recommended to the user who purchased the item I-03.

  As rule 5, there is a rule that recommends only patterns that are seen by users of a certain ratio or higher. Rule 1 or the like is a rule that determines an item to be selected based on the number of times of frequency. However, Rule 5 further sets the number of times of frequency as a ratio, and an item with a ratio equal to or higher than a preset ratio is recommended. To.

  When rule 5 is applied and a transition frequency table 92 as shown in FIG. 14 is created, and the ratio threshold is set to 80%, the transition frequency table 92 is stored in the user U001. From the history of 5 to U005, the frequency of item I-09 is 4, and the ratio is 80% (= 4 times / 5 people). Item I-09 is recommended.

  Rule 6 is that the number of items is recorded in the history of the user who is the target of recommendation, and the initial time series (before a certain order) among the frequent time series patterns. If there is only a history of the item, there is a rule for such a user to recommend a relatively earlier item in the time series pattern in which the items are rearranged in time series .

  In the history of the user to be recommended in Rule 6, the number of items is only recorded below a certain number, and among the frequent time-series patterns, the initial time series (before a certain order) A user who has a history of only one item can be determined to be a beginner.

  The fact that the number of items is only a certain number or less in the history means that it is not long before the purchase of the item is started. In addition, only the earliest items in the time series of the most frequent time series patterns have a history, which means that other users who continue to purchase items will purchase items shortly after purchasing them. This means that only the item you have left remains in the history. Therefore, it can be determined that the user is a beginner.

  For users who are judged to be such beginners, items for beginners and items for intermediate players are recommended. Therefore, when items are rearranged in chronological order, items that are relatively earlier, that is, items that are determined to be purchased when other users are beginners, are recommended. To do.

  In addition, if a lot of items are recommended for beginners, it is conceivable that on the side receiving the recommendation, which item should be selected from the recommended items is lost. For beginners, the number of items to be recommended tends to increase. Therefore, a rule for limiting the number of recommended items to a certain number or less may also be set.

  As rule 7, more than a certain number of items are recorded in the history of the user to be recommended, and only the item in the early time series (before a certain order) of the frequent time series patterns. When there is no history and when a similar history can be seen for other genres, similar user histories are extracted in other genres (groups), and items described in the extracted history are displayed. There are rules for recommending users to be recommended.

  In the rule 7, the number of items is recorded in the history of the user who is the target of recommendation, and only the item in the early time series (before a certain order) among the frequent time series patterns. If there is no history, and a user who can see the same history for other genres, it can be determined that the user has purchased the item widely and shallowly.

  The fact that the number of items is recorded in the history more than a certain number means that the time has passed since the purchase of the item has started and that the user is used to purchasing the item. Moreover, it seems that beginners prefer that when there is a history only for the items in the early time series (before a certain order) among the frequent time series patterns, and the same history can be seen for other genres as well. This means that they are interested in a wide variety of genres rather than purchasing a single item and investigating one genre.

  For such a wide and shallow type of user who purchases items, for example, the best seller or the best version is recommended regardless of genre. In addition, such best sellers and best versions are suitable for beginner users, and for beginners, as described above, when items are rearranged in chronological order, It is assumed that such a recommendation method is also adopted in the rule 7 because the item in the item is recommended.

  As rule 8, the number of items is recorded in the history of the user who is to be recommended, and an item in the latter half of the time series (after a certain order) among the frequent time series patterns. If it is included in the history at a certain frequency or more, items in the later direction are recommended in the time series pattern in which items are rearranged in time series, and more than a certain number of items are recommended at a time. There is a rule to make it.

  Similar to rule 7, the fact that a certain number or more of items are recorded in the history means that time has passed since the purchase of the item has started, and that the user is used to purchasing the item. In addition, items that are in the latter half of the time series (after a certain order) in the frequent time series pattern are included in the history at a certain frequency or more, so other users who continue to purchase items This means that a history of purchased items remains when a considerable amount of time has passed since the purchase began. Therefore, it can be determined that the person is an expert.

  For users who can be judged as such experts, items for experts are recommended. For this purpose, items that are rearranged in the time series pattern after the items are rearranged in time series are recommended.

  In addition, even if a skilled person recommends a relatively large number of items at once, it is considered that the item that suits him / her can be selected accurately, so that a certain number of items or more are recommended at a time. .

  As rule 9, the number of items is only recorded below a certain number in the history of the user to be recommended, and the latter of the time series (after a certain order) among the frequent time series patterns When items are included in the history at a certain frequency or more, a rule that recommends items that are relatively later in the time series pattern in which the items are rearranged in a time series, a number less than the predetermined number. There is.

  Similar to rule 6, it can be determined that the user is a beginner because only a certain number of items are recorded in the history. However, since the items in the latter half of the time series (after a certain order) in the time series pattern are included in the history at a certain frequency or more, it can be determined that the person is an expert. Thus, it may not be possible to determine whether the user is a beginner or an expert. For example, it is considered that such a situation that it is difficult to make a determination occurs when an expert person purchases an item managed by the server 2 for the first time.

  In such a case, although it is determined that the person is an expert, there is a possibility of being a beginner, so an item for the expert is recommended, but the number of items smaller than the number of items recommended for the expert of the rule 8 Make the number recommended.

  As rule 10, the number of items is recorded in the history of the user who is to be recommended, and the initial item in the time series (before a certain order) of the frequent time series patterns is displayed. There is a rule that when items are included in the history at a certain frequency or more, items are rearranged in time series and a relatively earlier item is recommended in a time series pattern than a predetermined number.

  Similar to rule 8, since a certain number or more of items are recorded in the history, it can be determined that the user is an expert. However, since the items in the previous period (before a certain order) of the time series patterns are included in the history at a certain frequency or more, it can be determined that the user is a beginner. Thus, it may not be possible to determine whether the user is a beginner or an expert. For example, if you are an expert in a given field, but have purchased an item to start a new study in another field, it may be difficult to make such a determination.

  In such a case, although it is determined that the person is a skilled person, there is a possibility that the person is a beginner, so an item for beginners is recommended. Make it recommended.

  As the rule 11, the history of the user who is to be recommended has a certain number of items, and the items other than the items included in the user history of the certain number or more, in other words, included in the user history, If there is a lot of items that do not enter the top of the ranking when a ranking is made such that the items are arranged in descending order, the frequency is constant among similar time-series patterns. There is a rule that items below a few ranks are recommended.

  Like rule 7 and rule 8, having a certain number or more of items in the history means that time has passed since the purchase of the item has started, and that the user is used to purchasing the item. In addition, the fact that there are many items other than the items included in the history of a certain number of users or more means that items that are generally purchased, for example, best-seller items, are not so much purchased. A user having such a history can be said to be a user who cannot be defined such as a beginner or an expert.

  For users who cannot be defined, it is determined that they prefer items that a relatively small number of people prefer, and items whose frequency is a certain number or less are recommended.

  In the rules 6 to 11, it is determined that the user is a beginner using the user's history. However, there are cases where items managed and sold by the server 2 can be purchased at stores other than the server 2, for example, in a real-world store, and the user does not necessarily purchase items under the management of the server 2. Not exclusively. Therefore, the server 2 may not be able to manage all items purchased by the user. In such a case, the determination using the user history in the rules 6 to 11 may not be an essential requirement.

  Although rules other than the above-described rules 1 to 11 may be further provided, the description will be continued here assuming that the rules 1 to 11 are set in the recommended rule selection unit 60 (FIG. 3).

  Also, for example, when selecting items to be recommended by applying rules 6 to 11, that is, when determining the type of user and selecting items to be recommended based on the determination, a time series pattern is generated for each type. However, only the expert-type time-series pattern may be recommended to the beginner-type user.

  In the recommendation rule selection unit 60, rules 1 to 11 are preset, and a rule suitable for the user who recommends the item is selected as the process in step S27 (FIG. 6). The recommended item selection unit 61 is instructed to wash items recommended by the selected rule.

  In step S <b> 28, the recommended item selection unit 61 refers to the item information table 82 managed by the item information table management unit 52, and reads information on items that meet the rule. The read information is supplied to the recommendation information providing unit 62.

  In step S29, the recommended information providing unit 62 provides the supplied information to the user side.

  In this way, items are arranged in time series in the order of purchase, and information on the items arranged in time series is used to purchase a user after purchasing the predetermined item. It is possible to recommend items that are desirable to do.

  Therefore, for example, it is possible to recommend an item for beginners to a beginner and recommend an item for an expert to an expert.

  According to the above-described embodiment, as described with reference to FIG. 16, for example, when a user who has purchased a predetermined item falls under the above-described rule 6 and is determined to be a beginner type, Item U-01 at the front in the series is recommended. For example, for a user who purchased item I-05, item I-09, which is the next item after item I-05 in the time series pattern, is recommended.

  In addition to recommending items (providing item information), some advice may be provided to the user. As advice, for example, when a user who is determined to be a beginner type tries to purchase an item for an expert, it may be advised that there is an item for a beginner that should be purchased first. .

  This will be described with reference to FIG. When the time series pattern is item I-01 or item I-05, item I-03 or item I-05 next to item I-09 or item I-10 User U001 purchased item I-01. At this time, the item I-03 or the item I-05 is recommended to the user U001 by executing the above-described processing.

  However, user U001 selected item I-09 after purchasing item I-01. In other words, the user U001 tried to purchase the item I-09 without purchasing the item I-03 or the item I-05. It can be determined by referring to the history of the user U001 in the history table 81 that the item I-03 or the item I-05 has not been purchased.

  In this way, when the next item is selected in a user having a certain history, it is judged from the time-series pattern, and when the item is more than a certain order and there is no item in the history, It is also possible to advise the user to reselect an item and advise the user to select an item in the middle of the time series pattern, that is, in this case, item I-03 or item I-05.

  Now, consider a case where a television broadcast program is handled as an item. When the above-described item is a program, the history table 81 or the like becomes a history of the program viewed by the user. Therefore, the time series history table 91 created from such a history table 81, the transition frequency table 92, and the like also describe programs viewed by the user.

  When such a viewed program is recommended to a user who has viewed a predetermined program, the program that has already been broadcast is recommended. Even if the recommended user side is recommended, the user cannot view the program. Therefore, when the item is a program, a program broadcast in the future is recommended instead of a program broadcast in the past.

  FIG. 18 shows an example of a time series pattern created when a program is handled as an item. In the example shown in FIG. 18, a viewer who has viewed program 01 has a pattern of viewing program 03 or program 05, and a viewer who has viewed program 03 or program 05 has a pattern of viewing program 09 or program 10. Suppose that it is determined by performing the process of arranging items in time series as described above.

  Since the program 01 and the program 03 are already viewed by the viewer, for example, even if the viewer who viewed the program 01 recommends that the viewer watch the program 03, the program has already been broadcast. 03 cannot be viewed. Therefore, the time series pattern related to the program as shown in FIG. 18 is rewritten to a general time series pattern using the metadata of the program.

  That is, when it is necessary to recommend future information as information, such as a television broadcast program, metadata tends to be difficult to make recommendation according to maturity by the time-series pattern described above as rule 6 or the like. Using an indirect collaborative filter at the level, programs can be recommended in time series.

  FIG. 19 is a table summarizing the metadata of the program shown in FIG. For example, the program information table 111 shown in FIG. 19 is a table in which production country, performers, and keywords are associated as program information. For example, it is described that the production country of the program 01 is “Japan”, the performer is “actor 1”, and the keyword is “romance”. Similarly, the production country of program 03 is “Japan”, the performer is “actor 2”, the keyword is “romance”, the production country of program 05 is “Japan”, and the performer is “actor” 3 ”, the keyword is“ romance ”, the production country of program 09 is“ Italy ”, the performer is“ actor 4 ”, the keyword is“ romance ”, and the production country of program 10 is“ “America”, the performer is “Actor 5”, and the keyword is “Human”.

  When the time series pattern of the program shown in FIG. 18 is rewritten with the information described in the program information table 111, the time series pattern shown in FIG. 20 is obtained. The metadata 01 is metadata extracted from the metadata of the program 01, and describes that the production country is “Japan”, the performer is “actor 1”, and the keyword is “romance”. Yes. The metadata 02 is metadata extracted from the metadata of the program 03 and the program 05, the production country is “Japan”, the performer is “actor 2 or actor 3”, and the keyword is “romance”. It is described that there is. The metadata 03 is metadata extracted from the metadata of the program 09 and the program 10, the production country is “Italy or America”, the performer is “actor 4 or actor 5”, and the keyword is “romance”. Or “human”.

  A viewer who has watched a program having metadata 01 tends to view a program having metadata 02, and a viewer who has watched a program having metadata 02 tends to view a program having metadata 03. It can be seen from the time-series pattern in FIG.

  For a viewer who has watched a program having the metadata 01, a program having the metadata 02 (a program that does not need to be completely matched and has metadata similar to the metadata 02) is, for example, an EPG (ElectronIc Program). Search from GuIde) and recommend the search results. In this way, future programs can be recommended using past time-series patterns. Therefore, it is possible to recommend a program in accordance with the tendency of the viewer to view.

[Regarding Server Function in Second Embodiment]
As described with reference to the flowchart of FIG. 6, the server 2 in the first embodiment executes the process from the grouping process to the recommendation. By including the grouping in the process of recommending the grouping as in the server 2 in the first embodiment, for example, it is possible to make the process suitable when the number of items is large or the number of users is large.

  When the number of items and the number of users are small, even if grouping is performed, the number of items and the number of users belonging to the group may decrease, and it may not function well. Therefore, a case where the recommendation process is executed without performing the grouping process will be described below as a second embodiment.

  For example, when the store has just opened and the number of purchasers is small and the history is small, the recommendation processing is performed by applying the second embodiment in which no grouping is performed, and the grouping is performed when the history is accumulated. It is also possible to switch to perform recommendation processing by applying the embodiment. Further, the first embodiment and the second embodiment can be used in combination.

  FIG. 21 is a diagram for explaining the function of the server 2 in the second embodiment. The server 2 includes a history table management unit 201, an item information table management unit 202, an item extraction unit 203, a transition count unit 204, a transition extraction unit 205, a time series generation unit 206, a recommendation rule selection unit 207, a recommended item selection unit 208, The recommendation information providing unit 209 is included.

  History table management unit 201, item information table management unit 202, item extraction unit 203, transition count unit 204, transition extraction unit 205, time series generation unit 206, recommendation rule selection unit 207, recommended item selection unit 208, and recommendation information provision The unit 209 includes the history table management unit 51, the item information table management unit 52, the item extraction unit 56, the transition count unit 57, the transition extraction unit 58, the time series generation unit 59, the recommendation rule selection unit 60, and the recommendation shown in FIG. The item selection unit 61 and the recommendation information providing unit 62 have basically the same configuration and perform the same processing, and thus detailed description thereof is omitted. Hereinafter, a description will be given of processing that is different from when grouping processing is performed by not performing grouping processing.

  FIG. 22 is a flowchart for explaining the operation of the server 2 shown in FIG.

  In step S <b> 101, the item extraction unit 203 extracts an item having a history of a certain frequency or more with reference to a table stored in the history table management unit 201. The item extraction unit 203 refers to the table shown in FIG. 4 managed by the history table management unit 201, pays attention to a predetermined item (for example, item I-01), and the predetermined item is purchased. Count (number of purchased users). Such a count is performed for each item. And the result is referred and the item purchased more than predetermined times is extracted.

  By the process in step S101, the total number of items that can be candidates for items recommended to the user is narrowed down. When this narrowing is performed, the history table 81 as shown in FIG. 4 is referred to, but a time series history table 91 as shown in FIG. 13 may be referred to.

  In step S102, the transition counting unit 204 counts the number of transitions between items. This count is determined by one transition before and after the item when a predetermined item is noticed.

  As will be described with reference to FIG. 23, when attention is paid to the item I-01, if there is a history that the item I-05 has been purchased before the item I-01 is purchased, only one is counted. As a result of counting all items purchased before item I-01, if item I-05 has the highest count value, the transition from item I-05 to item I-01 is determined.

  Similarly, when attention is paid to the item I-01, if there is a history of purchasing the item I-03 or the item I-04 after purchasing the item I-01, the number of times of transition for each item is counted. As a result, if the count value transitioned from item I-01 to item I-03 is 10, and the count value transitioned from item I-01 to item I-04 is 20, item I-01 to item I Since the transition to -04 has a higher count value, the transition from item I-01 to item I-04 is determined.

By performing such processing, for item I-01, including the transition before and after,
Item I-05 → Item I-01 → Item I-04
A time-series pattern is generated.

  In step S103, a time series pattern is generated for each item. And in step S104, the time series pattern as a whole is produced | generated by analyzing the time series pattern for every produced | generated item. In other words, a time series pattern of the entire item extracted in the process of step S101 is generated. For example, when 100 items are extracted, a time series pattern related to the 100 items is generated.

  As a result, for example, a time series pattern as shown in FIG. 24 is generated. The time series pattern shown in FIG. 24 will be described. The item I-05 changes to the item I-01. The item I-01 transitions to the item I-04 or the item I-10 because an item with the same count value exists (or because it does not fall within a predetermined range).

  When transitioning to item I-04, transition from item I-04 to item I-20. When the item changes to the item I-10, the item I-10 changes to the item I-11. The item I-11 changes to the item I-20. The item I-20 has the same count value in the case of transition from the item I-04 and the case of transition from the item I-11.

  As described above, the time series pattern may have a branching pattern (path).

  Thus, when the time-series pattern is generated, the process proceeds to step S105 (FIG. 22). In step S105, a recommendation rule is determined. As recommendation rules, rules 1 to 11 described in the first embodiment can be applied. Further, the recommended rule selection unit 207 can be configured to select one rule from the rules 1 to 11 set in advance.

  However, in the case of the second embodiment, it is considered to be relatively effective when applied when the number of items and the number of users are small. And when the number of items and the number of users are small, it is conceivable that the number of items included in the history is small even if the user's history is referenced. If the number of items included in the history is small, as a result, it is determined that the user is a beginner. Therefore, it is considered that the number of users determined to be beginners increases. Therefore, when using the second embodiment, instead of using the rules 1 to 11 described above, a rule that the next item after the item purchased by the user is recommended in the time series pattern is set. You may make it apply.

  For example, in the case of the time-series pattern as shown in FIG. 24, if the item purchased by the user is item I-01, item I-04 or item I-10 is recommended (both as recommended) Is of course good). The item I-11 may also be recommended together with the item I-10. That is, not only items on one route but also items existing on a branch route may be recommended together.

  In step S106, the recommended item selection unit 208 selects an item to be recommended. For example, as described above, an item existing after the position of the time series pattern of the purchased item is selected as the recommended item. The information of the selected item is provided to the user side by the recommendation information providing unit 209 in step S107.

  As described above, in the second embodiment, a time series pattern is generated without performing grouping, and an item recommended to the user is selected with reference to the time series pattern.

  In addition to the method of generating a time series pattern as described above, for example, a tree structure mining technique using TreeMIner or TreeFInder, a technique using GA, or an item time series ( It may be generated by taking the cross-correlation of the item purchase order) and counting highly correlated items under a certain lag (delay).

  Even if the recommendation process is performed based on the second embodiment, the same effect as that of the first embodiment can be obtained. That is, it is possible to recommend an item suitable for the user type, the learning progress of the user, and the like.

  When the recommendation to which the present embodiment is applied is performed, it is possible to accurately recommend an item that learns a predetermined field (area) gradually and deeply on an appropriate path.

  In addition, in the case of the recommendation using the conventional collaborative filter, it is highly likely that the recommended item with a similar history is felt suddenly for the recommended user, but this embodiment is applied. Thus, it is possible to reduce the possibility of making such a sudden recommendation. Also, for beginners in a certain area, those who are interested in the area for the first time, the contents for beginners are easy to be familiar with, and if the contents for experts and advanced users are recommended from the beginning, the interest will continue. However, by applying this embodiment, items for beginners are recommended for beginners, and items for advanced users are recommended for advanced users. It becomes possible.

[About recording media]
The series of processes described above can be executed by hardware or can be executed by software. When a series of processing is executed by software, a program constituting the software executes various functions by installing a computer incorporated in dedicated hardware or various programs. For example, it is installed from a program storage medium in a general-purpose personal computer or the like.

  As shown in FIG. 2, a program storage medium for storing a program that is installed in a computer and can be executed by the computer includes a magnetic disk (including a flexible disk), an optical disk (CD-ROM (Compact DIsc-Read Only). Memory), DVD (including DIgItal VersatIle DIsc)), magneto-optical disk (including MD (MInI-DIsc)), or removable media 31 that is a package medium made of semiconductor memory, or the program is temporary or permanent The ROM 22 is stored in the hard disk, the hard disk constituting the storage unit 28, and the like. The program is stored in the program storage medium using a wired or wireless communication medium such as a local area network, the Internet, or digital satellite broadcasting via a communication unit 29 that is an interface such as a router or a modem as necessary. Done.

  In the present specification, the step of describing the program stored in the program storage medium is not limited to the processing performed in chronological order according to the described order, but is not necessarily performed in chronological order. Or the process performed separately is also included.

  Further, in this specification, the system represents the entire apparatus constituted by a plurality of apparatuses.

  The embodiment of the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the gist of the present invention.

It is a figure which shows the structure of one Embodiment of the system to which this invention is applied. It is a figure which shows the structural example of a server. It is a figure which shows the function of a server. It is a figure for demonstrating a log | history table. It is a figure for demonstrating an item information table. It is a flowchart for demonstrating a recommendation process. It is a flowchart for demonstrating grouping. It is a figure for demonstrating a log | history table. It is a figure for demonstrating a quantification management table. It is a figure for demonstrating a cluster. It is a figure for demonstrating a cluster. It is a figure for demonstrating the other method of grouping. It is a figure for demonstrating a time series history table. It is a figure for demonstrating a transition frequency table. It is a figure for demonstrating a time series table. It is a figure explaining selection of the item to recommend. It is a figure for demonstrating advice. It is a figure for demonstrating the time series pattern of a program. It is a figure for demonstrating metadata. It is a figure which shows the time-sequential pattern rewritten by metadata. It is a figure which shows the function of a server. It is a flowchart for demonstrating a recommendation process. It is a flowchart explaining the count of the transition between items. It is a figure for demonstrating the time series pattern with a branch.

Explanation of symbols

  1 network, 2 server, 3 terminal, 21 CPU, 22 ROM, 23 RAM, 28 storage unit, 31 removable media, 51 history table management unit, 52 item information table management unit, 53 grouping information management unit, 54 grouping unit, 55 Similar group judging unit, 56 item extracting unit, 57 transition counting unit, 58 transition extracting unit, 59 time series generating unit, 60 recommended rule selecting unit, 61 recommended item selecting unit, 62 recommended information providing unit

Claims (12)

Generating means for generating a time-series history when items are arranged in order of purchase, from an access history associated with an item accessed by a user and an access time to the item;
Using a plurality of time-series histories, counting the number of times of transition from the first item to the second item, a frequency extracting means for extracting a high-frequency transition;
From the transition extracted by the frequency extraction means, a pattern generation means for generating a transition pattern between a plurality of items,
Based on a rule for selecting information to be recommended, an item accessed by a user who recommends the information is used as a reference, and an item positioned before or after the position of the item in the pattern is used as the item of recommended information. An information processing apparatus comprising: determination means for determining. Grouping means for grouping the users into a plurality of groups from the access history;
A specifying means for specifying to which group of the plurality of groups the user recommending the information belongs, and
The information processing apparatus according to claim 1, wherein the frequency extraction unit extracts the transition using a time-series history of users belonging to the group. The information processing apparatus according to claim 1, wherein the rule is a rule that recommends an item extracted as a transition having a high frequency by the extraction unit. The rule according to claim 1, wherein a ratio of an item frequency extracted as having a high frequency by the extraction unit to a number of users recommends an item having a ratio equal to or higher than a preset ratio. Information processing device. The rule is a user in which the number of items is recorded in the access history of a user who is a recommendation target and the number of items is less than a certain number, and the earlier item of the pattern is included in the history. The information processing apparatus according to claim 1, wherein the information processing apparatus is a rule that recommends an item that is relatively earlier in the pattern when it is determined. In the rule, the number of items is recorded in the access history of the user who is the target of recommendation, and the earlier item of the pattern is included in the history, and with respect to other genres If it is determined that the user can see the same history, the access history of similar users is extracted from a group different from the group determined to belong to the user, and the extracted history is included in the extracted history. The information processing apparatus according to claim 2, wherein the rule recommends an item. The rule is a user whose number of items is recorded in a certain number or more in the access history of the user who is a recommendation target and the later item of the pattern is included in the history. The information processing apparatus according to claim 1, wherein the information processing apparatus is a rule that recommends an item that is relatively later in the pattern. According to the rule, a user whose number of items is recorded in the access history of a user who is a recommendation target, and the later item of the pattern is included in the history. 2. The information processing apparatus according to claim 1, wherein when it is determined that the number of items is relatively later than the predetermined number in the pattern, the information processing apparatus according to claim 1. The rule is a user whose number of items is recorded in a certain number or more in the access history of a user who is a recommendation target, and the earlier item of the pattern is included in the history. 2. The information processing apparatus according to claim 1, wherein when it is determined, the rule is to recommend a relatively larger number of items that are relatively ahead in the pattern. The rule is a ranking of items arranged in the order in which the number of items recorded in a plurality of the access histories is larger than a certain number in the access history of the user to be recommended. If it is determined that the user includes a large number of items that do not fall in the top of the ranking, a rule that recommends an item having a lower rank among similar patterns is created. The information processing apparatus described in 1. A generation step of generating a time-series history when items are arranged in the order of purchase from an access history associated with an item accessed by a user and an access time to the item,
Using a plurality of time-series histories, a frequency extraction step of counting the number of times of transition from the first item to the second item and extracting a frequent transition;
From the transition extracted in the process of the frequency extraction step, a pattern generation step for generating a transition pattern between a plurality of items,
Based on a rule for selecting information to be recommended, an item accessed by a user who recommends the information is used as a reference, and an item positioned before or after the position of the item in the pattern is used as the item of recommended information. An information processing method including a determination step for determining. A generation step of generating a time-series history when items are arranged in order of purchase from an access history associated with an item accessed by a user and an access time to the item;
Using a plurality of time-series histories, a frequency extraction step of counting the number of times of transition from the first item to the second item and extracting a frequent transition;
From the transition extracted in the process of the frequency extraction step, a pattern generation step for generating a transition pattern between a plurality of items,
Based on a rule for selecting information to be recommended, an item accessed by a user who recommends the information is used as a reference, and an item positioned before or after the position of the item in the pattern is used as the item of recommended information. A program that causes a computer to execute a process including a determining step for determining.

Images