JP2011070396A - Method of analyzing trend of using commodity, method of recommending commodity, system for analyzing trend of using commodity, and system for recommending commodity - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method of analyzing a trend of using a commodity, a method of recommending commodity, a system for analyzing the trend of using the commodity, and a system for recommending commodity, for properly obtaining attributes of each consumer for effective analysis. <P>SOLUTION: The method of analyzing the trend of using the commodity includes the steps of: accessing a commodity purchase history database 214 and acquiring the purchase history information of the plurality of consumers of the plurality of commodities (S100); creating a vector for each predetermined vectorized interval for each consumer based on the acquired purchase history information (S200); clustering the created vector and classifying it into a plurality of data groups (S20); creating a representative vector which represents each of the plurality of classified data groups (S30); and classifying each of the plurality of consumers in association with any of the plurality of created representative vectors (S40). <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a product usage trend analysis method for analyzing a consumer's product usage trend, a product recommendation method using the product usage trend analysis system, and a product recommendation system.

  Generally, there are individual differences in consumer personality and consumption trends. For example, there are consumers with attributes that jump to fashion and new products that tend to be consumed in a relatively short time after a product is put on the market. In addition, there are consumers who have an attribute that thoroughly examines the state of a new product, which tends to be consumed after a relatively long time since the product was put on the market. When putting a product into the market, it is preferable to identify the type of each consumer and predict the sales of the product as much as possible.

As a conventional technique related to the above-mentioned sales prediction of a product, for example, there is a product sales prediction device described in Patent Document 1. This sales prediction device specifies consumer-oriented attributes and purchase timing patterns from past sales performance data. At a sales transition stage of a product, the trend of consumers who purchase the product, the purchase timing pattern, and the relationship between the sales of the product are modeled and analyzed for each product category. Based on the consumption model, the final sales of the product are predicted.

Japanese Patent Laid-Open No. 10-307808

  However, in the above prior art, each consumer is forcibly applied to a consumption model created in advance based on past purchase history of the consumer, and the sales of the product are predicted. Therefore, if the consumer's consumption behavior almost corresponds to the consumption model, a relatively effective analysis can be performed, but if the consumer's consumption behavior is far from the consumption model, the effective analysis can be performed. Impossible to do. As described above, it has been difficult to accurately grasp the attributes of each consumer with the above-described conventional technology.

  An object of the present invention is to provide a product usage trend analysis method, a product recommendation method, a product usage trend analysis system, and a product recommendation system that can accurately grasp and effectively analyze the attributes of each consumer.

  In order to achieve the above object, the product usage trend analysis method of the first invention is based on operation input information in the first information terminal on the analyst's side, and uses the identification information of the product and the use of a plurality of consumers for the product. Accessing a product usage history database storing usage history information associated with a history, and acquiring an acquisition history information of a plurality of consumers for a plurality of products, acquired in the information acquisition procedure, Based on the usage history information accumulated in the product usage history database, a vectorization procedure for creating a product usage frequency vector for each predetermined vectorization interval for each consumer, and the product usage created by the vectorization procedure A data group classification procedure for classifying the frequency vector into a plurality of data groups by clustering and the plurality of data classified by the data group classification procedure A representative vector creation procedure for creating a representative vector that individually represents each of the plurality of consumers, and classifying each of the plurality of consumers in association with one of the plurality of representative vectors created in the representative vector creation procedure. And a consumer classification procedure.

  In the first invention of this application, based on the usage history information acquired from the product usage history database in the information acquisition procedure, a product usage frequency vector is created for each consumer in the vectorization procedure. The product use frequency vector is classified into a plurality of data groups by performing a clustering process in the data group classification procedure. Then, after representative vectors for each data group are created in the representative vector creation procedure, each consumer is classified in association with one of the representative vectors in the consumer classification procedure.

  As described above, in the first invention of this application, the product usage frequency vectors created based on the usage history information of each of the plurality of consumers are classified into a plurality of data groups by clustering, and each consumer is classified into any data. The classification is performed in association with the representative vector of the group. In this way, unlike conventional methods in which each consumer is forcibly applied to a given consumption model, an analysis that flexibly reflects the usage trends of multiple consumers is performed, and a representative use period is determined for each data group. Can be calculated. As a result, the usage attributes of each consumer can be accurately grasped. Further, unlike the conventional method, it is not necessary for the analyst to manually set the time interval for analysis, and the labor burden on the analyst can be reduced.

  According to a second aspect, in the first aspect, the vectorization step includes creating an interval setting step for setting the vectorization interval and a product use frequency vector for each set vectorization interval for each consumer. A vector generation procedure to be performed, a zero vector determination procedure for determining whether or not the number of elements in a vector that is zero among the vectors generated in the vector generation procedure is equal to or less than a predetermined value, and zero in the zero vector determination procedure An interval resetting procedure for setting the vectorization interval to be longer than before when the number of elements in the vector becomes greater than a predetermined value, and the number of elements in the vector that becomes zero in the zero vector determination procedure is The interval resetting procedure, the vector creation procedure, and the zero vector determination procedure are repeated until the predetermined value or less is reached. To.

  By reducing the number of elements in the vector that becomes zero as much as possible, the curve representing the vector can be made as smooth as possible, and a highly accurate analysis can be performed.

  According to a third invention, in the second invention, the interval setting procedure and the interval resetting procedure set n as an integer equal to or greater than 1, and set the vectorization interval in units of n weeks.

  By making the number of Saturdays, Sundays, and weekdays included in each vectorization interval the same, the effect on usage history due to the difference in human life patterns between Saturdays, Sundays, and weekdays is evenly distributed at each vectorization interval. Can be included. Thereby, analysis can be performed with higher accuracy.

  In a fourth aspect based on the third aspect, the vector creation procedure includes a history acquisition procedure for acquiring the usage history information accumulated in the product usage history database, and the usage history information acquired in the history acquisition procedure. The elapsed time calculation procedure for calculating the elapsed time until the consumer's use, using the date and time of use by the consumer included in the product and the market launch time of the product included in the usage history information, and the elapsed time calculation Dividing the elapsed time calculated in the procedure by the vectorization interval set in the interval setting procedure or the interval resetting procedure to calculate the elapsed period number, and the elapsed period number Using the calculation result in the calculation procedure, an elapsed period number counting procedure for counting one consumer's product usage history for each of a plurality of products separately for each elapsed period number. Characterized in that it.

  In the elapsed time calculation procedure, the elapsed time until use is calculated using the consumer's use date and time and the product launch time, and in the elapsed period number calculation procedure, the elapsed time is divided by the vectorization interval, and the elapsed period number is calculated. calculate. Then, in the elapsed period number counting procedure, the product usage history of one consumer for each of the plurality of products is counted up separately for each elapsed period number. Thereby, the product use frequency vector of each consumer can be calculated.

  According to a fifth invention, in any one of the first to fourth inventions, the information acquisition procedure includes a category-specific acquisition procedure for acquiring the usage history information of a plurality of consumers for a plurality of products for each category of the products. It is characterized by providing.

  As a consumer trend, the number of uses and the use cycle may differ depending on the product category. In the fifth invention of the present application, the analysis performance can be further improved by dividing the product into categories and acquiring the usage history information.

  In a sixth aspect based on the fifth aspect, the information acquisition procedure is such that, among the usage history information of the plurality of consumers acquired in the category-specific acquisition procedure, the number of applicable consumers is equal to or greater than a predetermined value. It further comprises a history extraction procedure for extracting the usage history information within a specific period.

  As a result, the number of consumers that are subject to history extraction can always be ensured to be equal to or greater than a predetermined value, so that the analysis accuracy can be ensured with high accuracy.

  A seventh invention is the market of the product according to the sixth invention, wherein the information acquisition procedure is determined for each product out of the usage history information within the specific period extracted by the history extraction procedure. It is further characterized by further comprising a data extraction procedure for extracting the usage history information as a data sample from the time when the product is used until the consumer's usage of the product falls below a predetermined value.

  By excluding data that is substantially unnecessary for the analysis from the analysis target, it is possible to perform a quick and smooth analysis.

  In order to achieve the above object, the product recommendation method of the eighth invention of the present application is based on the operation input information in the first information terminal on the analyst side, and the identification information of the product and the usage histories of a plurality of consumers for the product, The information acquisition procedure for accessing the product usage history database storing the usage history information in association with each other, acquiring the usage history information of a plurality of consumers for a plurality of products, and the product usage acquired in the information acquisition procedure Based on the usage history information accumulated in the history database, a vectorization procedure for creating a product usage frequency vector for each predetermined vectorization interval for each consumer, and the product usage frequency vector created by the vectorization procedure Are classified into a plurality of data groups by clustering, and the plurality of data groups classified by the data group classification procedure A representative vector creation procedure for creating a representative vector that individually represents each of the plurality of consumers is associated with one of the plurality of representative vectors created in the representative vector creation procedure. A predetermined product at a predetermined recommended time for a consumer classification procedure to be classified and a consumer-side second information terminal classified in association with one of the plurality of representative vectors in the consumer classification procedure And a recommended information output procedure for outputting the recommended information.

  In the eighth invention of the present application, similarly to the first invention of the present application, the usage attribute of each consumer can be accurately grasped. In addition, if a product tends to be consumed in a relatively short time after being put on the market, the product is recommended early after the product is put on the market, and a relatively long time has passed since the product was put on the market. For those who tend to consume later, product recommendations can be made late after product launch. As a result, effective product recommendations can be made according to the individuality and consumption trends of each consumer.

  In order to achieve the above object, the product usage trend analysis system according to the ninth aspect of the present invention is a product usage history database storing usage history information in which product identification information and usage history of a plurality of consumers for the product are associated with each other. Based on the usage history information accumulated in the product usage history database, a vectorization unit that creates a product usage frequency vector for each predetermined vectorization interval for each consumer, and the vectorization unit Data group classification means for clustering the product usage frequency vector into a plurality of data groups, and a representative for creating a representative vector that individually represents each of the plurality of data groups classified by the data group classification means A vector creating means and each of the plurality of consumers are connected to the plurality of representative vectors created in the representative vector creating procedure. To any of Torr, and having a consumer classifying means for classifying association.

  In the ninth invention of this application, based on the usage history information in the product usage history database, the vectorization means creates a product usage frequency vector for each consumer. The product use frequency vector is classified into a plurality of data groups by being clustered by the data group classification means. Then, after representative vectors for each data group are created by the representative vector creating means, each consumer is classified by being associated with one of the representative vectors by the consumer classification means.

  As described above, in the ninth invention of the present application, the product usage frequency vectors created based on the usage history information of each of the plurality of consumers are classified into a plurality of data groups by clustering, and each consumer is classified into any data. Classification is performed in association with the representative vector of the group. In this way, unlike conventional methods in which each consumer is forcibly applied to a given consumption model, an analysis that flexibly reflects the usage trends of multiple consumers is performed, and a representative use period is determined for each data group. Can be calculated. As a result, the usage attributes of each consumer can be accurately grasped. Further, unlike the conventional method, it is not necessary for the analyst to manually set the time interval for analysis, and the labor burden on the analyst can be reduced.

  In order to achieve the above object, the product recommendation system according to the tenth aspect of the present invention includes a product use history database storing use history information in which product identification information is associated with use histories of a plurality of consumers for the product, Based on the usage history information accumulated in the product usage history database, vectorization means for creating a product usage frequency vector for each predetermined vectorization interval for each consumer, and the product created by the vectorization means Data vector classification means for classifying usage frequency vectors into a plurality of data groups by clustering processing, and representative vector creation for creating representative vectors that individually represent the plurality of data groups classified by the data group classification means And a plurality of representative vectors created by the representative vector creation procedure for each of the plurality of consumers. A consumer classifying unit that classifies in association with any of the two, and a second information terminal on the consumer side that is classified in association with any of the representative vectors by the consumer classifying unit. And a recommended information output means for outputting recommended information of a predetermined product at the recommended time.

  In the tenth invention of the present application, the usage attribute of each consumer can be accurately grasped in the same manner as the ninth invention of the present application. In addition, if a product tends to be consumed in a relatively short time after being put on the market, the product is recommended early after the product is put on the market, and a relatively long time has passed since the product was put on the market. For those who tend to consume later, product recommendations can be made late after product launch. As a result, effective product recommendations can be made according to the individuality and consumption trends of each consumer.

  According to the present invention, it is possible to accurately grasp the attributes of each consumer and perform effective analysis.

1 is a system configuration diagram conceptually showing an overall configuration of a product recommendation system according to an embodiment of the present invention. It is a figure showing an example of the memory content of a goods purchase history database. It is explanatory drawing explaining the method to acquire the purchase history information accumulate | stored in the product purchase history database for every category of goods. It is explanatory drawing explaining the method to extract the purchase history information in a specific period from the acquired purchase history information. It is explanatory drawing explaining the method to extract specific purchase history information as a data sample from the purchase history information in the extracted specific period. It is explanatory drawing explaining the method of producing the vector for every vectorization space | interval. It is a table | surface which shows an example of the vector table showing a vector. It is a table | surface which shows an example of the normalization vector table showing the normalized vector. It is explanatory drawing explaining an example of the method of clustering. It is a figure which shows an example of the produced representative vector. It is a table | surface which shows an example of the matching result of several consumers and several representative vectors. It is a flowchart showing the control content which the control part of a server performs. It is a flowchart showing the control content of step S100. It is a flowchart showing the control content of step S300. It is a flowchart showing the control content of step S400. It is a flowchart showing the control content of step S200. It is a flowchart showing the control content of step S500. It is a system block diagram which represents notionally the whole structure of a goods recommendation system in the modification which analyzes a goods use tendency based on the singing trend in Karaoke.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

  The overall configuration of the product recommendation system of this embodiment will be described with reference to FIG.

  In FIG. 1, a product recommendation system 1 includes a credit company terminal 100, a server 200 as a first information terminal on an analyst side, a base station 300, and a mobile terminal 400 as a consumer second information terminal. And have.

  The server 200 is connected to the credit company terminal 100 via a network NW such as the Internet. A display 220, a keyboard 221, and a mouse 222 are connected to the server 200. In addition, the server 200 includes a control unit 201 that controls the entire server 200, a storage unit 210 that includes a hard disk drive (HDD), a communication control unit 202, an output control unit 203, and an input control unit 204. have.

  The communication control unit 202 controls information communication performed between the credit company terminal 100 and the base station 300 via the network NW. The output control unit 203 performs control related to the output of the video signal to the display 220. The input control unit 204 performs control related to input of information via the keyboard 221 and the mouse 222.

  The storage unit 210 includes a product database 212, a program storage area 213, and a product purchase history database 214.

  The product database 212 stores recommendation information, which is information about products such as product names, product contents, prices, appearance images, etc., for recommending various products.

  The product purchase history database 214 stores purchase history information of a plurality of consumers for a plurality of products (refer to FIG. 3 described later for details). The product purchase history database 214 corresponds to a product use history database, and the purchase history information corresponds to use history information.

  The program storage area 213 includes a product usage trend analysis processing program and a product recommendation processing program for causing the control unit 201 to execute a product usage trend analysis method and a product recommendation method (see FIGS. 3 to 17 described later). Various programs are stored.

  The control unit 201 is a so-called microcomputer. In other words, although not shown, the control unit 201 includes a central processing unit such as a CPU, a ROM, and a RAM, and stores them in the program storage area 213 of the storage unit 210 while using the temporary storage function of the RAM. Signal processing is performed in accordance with the various programs.

  The server 200 having the above configuration is connected to the base station 300 via the network NW. Thereby, the recommended information of the predetermined product stored in the product database 212 of the server 200 can be transmitted by wireless communication to the mobile terminal 400 carried by each consumer via the base station 300. As the mobile terminal 400, for example, a mobile phone or the like is used. In addition, for example, an information terminal such as a Personal Digital Assistant (PDA) may be used.

  The portable terminal 400 includes a display unit 401 such as a liquid crystal display that performs various displays, an operation unit 402, and an antenna 410. The operation unit 402 includes numeric keys and operation buttons for the consumer to perform various operation inputs. The antenna 410 transmits / receives information to / from the base station 300 via wireless communication.

  In addition to the display unit 401, the operation unit 402, and the antenna 410, the mobile terminal 400 includes a wireless communication control unit 403 and a control unit 404 that controls the entire mobile terminal 400. The wireless communication control unit 403 controls information communication performed with the base station 300 via wireless communication. When receiving the recommended product information from the server 200 via the network NW and the base station 300, the control unit 404 causes the display unit 401 to display the recommended information.

  The server 200 functions as a product usage trend analysis system and a product recommendation system described in the claims.

  An example of the contents stored in the product purchase history database 214 will be described with reference to FIG.

  In FIG. 2, the product purchase history database 214 stores purchase history information in which information related to a product is associated with purchase histories of a plurality of consumers for the product. Note that the consumer purchase history corresponds to the consumer use history. The purchase history information includes a category ID, a product code, a sales start date / time, a consumer ID, and a purchase date / time.

  The category ID is identification information assigned to each of a plurality of categories to which a plurality of products belong. The product category is classified by, for example, the genre of products such as clothing, food, and books, and the price range of the products. The product code is identification information assigned to each of a plurality of products. These category IDs and product codes correspond to product identification information.

  The sales start date / time is date / time information when the product was first sold in the market by the product sales company. In this example, “year / month / day” is stored as the sales start date and time, but “time” may also be stored. This sales start date and time corresponds to the time when the product is put on the market.

  The consumer ID is identification information of the consumer who purchased the product. The purchase date / time is date / time information when the consumer purchases the product. In this example, “year / month / day” is stored as the purchase date and time, but “time” may also be stored. The purchase date and time corresponds to the use date and time by the consumer, and the consumer ID and purchase date and time correspond to the purchase history of the consumer, that is, the use history of the consumer.

  The feature of the product recommendation system 1 having the above configuration is that the server 200 accesses the product purchase history database 214 to acquire purchase history information, creates a product purchase frequency vector (described later) for each consumer, There are a procedure for clustering a purchase vector to create a representative vector (described later), and a procedure for classifying each of a plurality of consumers in association with one of the plurality of representative vectors. Hereinafter, details of these contents will be sequentially described.

  Here, as the purchase tendency of products by consumers, the number of purchases and the purchase cycle may differ depending on the category to which the product belongs. Therefore, in the present embodiment, the control unit 201 acquires the purchase history information accumulated in the product purchase history database 214 for each category ID corresponding to the product category. Then, subsequent processing, that is, extraction of purchase history information for a specific period described later, extraction as a data sample, creation of a product purchase frequency vector, normalization of a product purchase frequency vector, clustering, creation of a representative vector, and consumption A process such as classifying a person (see FIGS. 4 to 11 described later) is executed separately for each category ID.

  First, a method for acquiring the purchase history information accumulated in the product purchase history database 214 for each product category will be described with reference to FIG.

  In FIG. 3, the control unit 201 accesses the product purchase history database 214 according to operation input information on the keyboard 221 and the mouse 222 connected to the server 200 on the analyst side. Then, the purchase history information accumulated in the product purchase history database 214 is acquired for each category of products, that is, for each category ID. Note that only purchase history information related to a specific category ID, for example, only purchase history information related to the category ID “4” may be acquired.

  A method of extracting purchase history information within a specific period from the purchase history information acquired as described above will be described with reference to FIG. 4 (a) and 4 (b) show graphs with the horizontal axis representing time and the vertical axis representing the number of cumulative consumer IDs.

  When the purchase history information is acquired separately for each category ID from the product purchase history database 214 as described above, the purchase history information including purchase histories of the plurality of consumers is greater than a predetermined value, for example, 1000 or more. The purchase history information within a specific period including the consumer purchase history, that is, the consumer ID is extracted.

  As the method, for example, based on the purchase history information acquired from the product purchase history database 214, a corresponding purchase date and time among a plurality of consumer IDs included in the purchase history information is within a preset predetermined period. For example, the number of corresponding consumer IDs within the period from the current date and time to six months ago is counted. Then, the counted number of consumer IDs is compared with the predetermined value.

  At this time, as shown in FIG. 4 (a), the number of consumer IDs corresponding to the purchase date and time within a preset time period T0 to time T1 is C1, for example 1300, and the consumer The number of IDs C1 may be a predetermined value Cm or more, for example, 1000 or more. In this case, among the purchase history information acquired from the product purchase history database 214, purchase history information in which the purchase date and time included in the purchase history information falls within the period from the time T0 to the time T1 is extracted. In this case, the period from time T0 to time T1 corresponds to a specific period.

  On the other hand, as shown in FIG. 4B, the number of consumer IDs in which the corresponding purchase date / time falls within a predetermined period from time T0 to time T2 is C2, for example, 660, and the consumer ID May be less than a predetermined value Cm, for example, less than 1000. In this case, the period for extracting the purchase history information is extended to the date and time Tm at which the number of consumer IDs from the time T0 becomes the predetermined value Cm. That is, the date and time Tm at which the number of consumer IDs from the time T0 becomes the predetermined value Cm is detected. Then, purchase history information in which the purchase date and time included in the purchase history information falls within the period from the time T0 to the time Tm is extracted from the purchase history information acquired from the product purchase history database 214. In this case, the period from time T0 to time Tm corresponds to a specific period.

  A method of extracting specific purchase history information as a data sample from purchase history information within a specific period extracted as described above will be described with reference to FIG. FIG. 5 shows a time transition curve representing the elapsed time when the sales start date and time of the product is 0 on the horizontal axis and the temporal change as the number of purchases of the product on the vertical axis.

  When the purchase history information within a specific period is extracted as described above, the specific purchase history information is then extracted from the purchase history information within the specific period, as a data sample of consumer product usage trend analysis. Further extract as. Specifically, the number of purchases of the product is equal to or less than a predetermined value after the trend peak position (described later), for example, 0.01 times the number of purchases corresponding to the trend peak position, from the purchase history information extracted in the specific period. The purchase history information until it becomes etc. is extracted as a data sample. Note that the number of purchases of the product corresponds to the degree of consumer use of the product.

  As the method, for example, as shown in FIG. 5, the trend peak position P (Tp, Pp) of the time transition curve is detected based on the purchase history information within the extracted specific period. In this example, the vertical axis of the time transition curve is the number of purchases, but it may be the number of purchasers, that is, the number of consumer IDs. Then, after a time Tp corresponding to the detected trendy peak position P, a time Te at which the number of purchases is equal to or less than a predetermined value Pe is detected. The period from the sales start date and time of the product, that is, the time 0 to the purchase date and time corresponding to the time Te when the detected number of purchases of the product is equal to or less than the predetermined value Pe It becomes. That is, among the purchase history information extracted within the specific period, the purchase date / time included in the purchase history information is within the period from time 0, that is, from the sales start date / time to the date / time corresponding to the time Te. The purchase history information corresponding to is extracted as a data sample.

  In addition, since the purchase tendency by consumers varies depending on the product, the process of extracting the purchase history information as a data sample is performed for each product for all the products included in the extracted purchase history information of one category. That is, it is performed for each product code.

  When the data sample is extracted as described above, next, a product representing the purchase frequency of one consumer for a plurality of products for each consumer based on the purchase history information that is the extracted data sample. Create a purchase frequency vector. This product purchase frequency vector corresponds to a product use frequency vector. Hereinafter, the above-mentioned product purchase frequency vector will be abbreviated and simply referred to as “vector”.

  In the present embodiment, in creating the vector, first, a predetermined vectorization interval, which is an interval for creating the vector, is set in units of n weeks, i.e., 1 week, 2 weeks, 3 Set to be weekly. Thereafter, a vector for each set vectorization interval is created for each consumer.

  A method of creating a vector for each vectorization interval will be described with reference to FIGS.

  In the present embodiment, when the vectorization interval is set, the purchase time and the sales start date and time included in the purchase history information extracted as the data sample are used, and the elapsed time until the purchase of the consumer product, That is, the elapsed time from the sales start date and time of the product to the purchase date and time of the product is calculated. The elapsed time until the consumer purchases the product corresponds to the elapsed time until the consumer uses the product.

  That is, as shown in FIG. 6, one purchase history information is acquired from the purchase history information extracted as the data sample, and the sales start date and time and purchase date and time included in the acquired purchase history information are used. Based on this, the elapsed time is calculated. For example, in the case of purchase history information related to category ID “4”, product code “15498”, sales start date and time “2008/7/10”, consumer ID “972514”, and purchase date and time “2008/7/28”. The elapsed time can be calculated as 18 [days] from the time difference between the sales start date “2008/7/10” and the purchase date “2008/7/28”. This represents that the consumer purchased the product on the 18th day after the sale of the product started.

  When the elapsed time until the purchase of the product is calculated as described above, the elapsed time is then calculated by dividing the calculated elapsed time by the set vectorization interval. Specifically, the elapsed time is divided by the vectorization interval. And the fraction below the decimal point of the calculated numerical value is rounded up, and it calculates as the said elapsed period number. Specifically, if the calculated numerical value is an integer obtained by dividing the elapsed time by the vectorization interval, the numerical value is calculated as the elapsed period number, and the elapsed time is divided by the vectorization interval. If the obtained numerical value is not an integer, the decimal point of the value is rounded down, and a value obtained by adding 1 to the remaining integer value is calculated as the number of elapsed periods.

For example, when the set vectorization interval is 1 week, that is, 7 days, and the calculated elapsed time is 18 [days],
Elapsed time / vectorization interval = 18 / 7≈2.6
It becomes. Therefore, the number of elapsed periods is calculated as 3. This indicates that the consumer has purchased the product for 3 weeks after the start of sales of the product, that is, between the 15th and 21st days.

  After calculating the number of elapsed periods until the purchase of the product by the consumer in this way, next, based on the calculation result of the number of elapsed periods, one consumer's product purchase history for a plurality of products is Count separately for each period. Specifically, the product purchase history related to a plurality of consumer IDs included in the purchase history information extracted as the data sample is counted separately for each number of elapsed periods for each consumer ID. The consumer's product purchase history corresponds to the consumer's product use history.

  In the present embodiment, the count is performed using a vector table (refer to FIG. 7 described later for details) stored in an appropriate area of the storage unit 210 of the server 200. In this vector table, the number of product purchase histories related to each consumer ID can be recorded separately for each number of elapsed periods.

  In the above example, the product purchase history related to the consumer ID “972514” is calculated to have 3 corresponding elapsed periods. Accordingly, the numerical value in the column corresponding to the elapsed period number “3” is incremented among the rows in the illustrated horizontal row corresponding to the consumer ID “972514” of the vector table, that is, 1 is added to the numerical value. In this example, the numerical value in the column is incremented from 0 to 1.

  Since the initial value of the numerical value in each column of the vector table is “0”, acquisition of purchase history information from the purchase history information extracted as a data sample as described above, calculation of elapsed time, calculation of the number of elapsed periods, In addition, by repeatedly recording in the vector table, the product purchase history related to a plurality of consumer IDs included in the purchase history information extracted as the data sample is divided for each number of elapsed periods for each consumer ID. Can be counted up.

  FIG. 7 shows an example of a vector table in which product purchase histories related to a plurality of consumer IDs included in the purchase history information extracted as data samples are counted separately for each elapsed time period for each consumer ID.

  As shown in FIG. 7, in this example, for each of a plurality of consumer IDs included in the purchase history information extracted as a data sample, the number of corresponding product purchase histories is changed from the number of elapsed periods “1” to the number of elapsed periods “ It is divided and counted up to 52 ”. The range of the number of elapsed periods for counting the number of product purchase histories can be appropriately set via the server 200 on the analyst side, for example, according to the so-called product life.

  For example, for the consumer ID “7” shown at the top of the vector table shown in FIG. 7, the number of product purchase histories included in the purchase history information extracted as the data sample is 0 for the elapsed period number “1”. The elapsed period number “2” is 0, the elapsed period number “3” is 2, the elapsed period number “4” is 7, the elapsed period number “5” is 5, the elapsed period number “6” is 1, and the elapsed period number “7”. , 6 for the elapsed period number “4”, 4 for the elapsed period number “9”,..., 0 for the elapsed period number “51”, and 1 for the elapsed period number “52”.

  In this way, counting the purchase history of one consumer for each number of elapsed periods means how many items in one category each consumer purchases, that is, This represents the frequency of product purchase. In other words, the attributes are the number of purchases of a plurality of products in a category by the consumer and the number of elapsed periods until the purchase of the plurality of products, that is, corresponding to the consumer's product purchase frequency. This is equivalent to creating a vector. Therefore, the horizontal row of data relating to the consumer ID in the vector table corresponds to the vector corresponding to the consumer's product purchase frequency represented by the consumer ID. Further, each piece of data in a horizontal row related to the consumer ID in the vector table, that is, each count number of the product purchase history, corresponds to an element in the vector of the consumer represented by the consumer ID.

  In the example shown in FIG. 7, the number of the product purchase histories is the number of elapsed periods “52”, that is, a case where it is counted up to 364 [days] when converted to elapsed time. The number of periods may be counted up to “104” or may be counted up to a shorter period, for example, the number of elapsed periods “26”.

Then, when a vector is created for each consumer as described above, a known normalization process is performed in which each of the created vectors is multiplied by the inverse of the vector norm and the norm of each vector is 1. That is, if the created vector is V and the normalized vector is V2,

(Formula 1)
The relationship holds. However,
It is.

  FIG. 8 shows an example of a normalized vector table representing normalized vectors. In the example shown in FIG. 8, each of the plurality of consumer vectors shown in FIG. 7 represents a vector normalized using the above (Equation 1). That is, the norms of the vectors of the plurality of consumers shown in FIG.

  For example, the normalized vector data corresponding to the consumer ID “7” shown at the top of the normalized vector table shown in FIG. 8 has an elapsed period number “1” of 0.00 and an elapsed period number “ 0.00 for 2), 0.11 for the number of elapsed periods "3", 0.39 for the number of elapsed periods "4", 0.28 for the number of elapsed periods "5", 0.06 for the number of elapsed periods "6" Elapsed period number "7" 0.34, Elapsed period number "8" 0.22, Elapsed period number "9" 0.22, ... Elapsed period number "51" 0.00, elapsed The number of periods “52” is 0.06. The horizontal row of data related to the consumer ID in the normalized vector table corresponds to a normalized vector corresponding to the consumer represented by the consumer ID.

  When the normalization process is performed for each vector as described above, each of the normalized vectors is classified into a plurality of data groups by a known clustering process. Here, the clustering is a so-called “unsupervised data classification method” such as a known K-means method or a self-organizing map, and is a method of automatically classifying data without an external reference.

  An example of a clustering method will be described with reference to FIGS. 9A and 9B.

  9A and 9B, as an example in this example, a plurality of data D randomly existing in the XY coordinate system is obtained using the K-means method which is one of the clustering methods. Classify into the following data groups.

  That is, first, the number of clusters for classifying each data D, that is, the number of data groups to be classified is set. In this example, the number of clusters is set to 2, and the center position is determined randomly. Then, each data D is allocated to one of the two clusters set as described above. This allocation is performed so that the Euclidean distances are close to each other at the two center positions. Then, based on the position of each data D, the center position of each cluster is calculated. Then, the Euclidean distance between the position of each data D and the calculated center position of each cluster is calculated. Thereafter, each data D is reassigned to a cluster having a short Euclidean distance among the clusters. At this time, if the cluster is reassigned for any of the data D, the center position of the newly allocated cluster is calculated, and the same processing as described above is repeated. Then, by repeating the above processing, if the reallocation of the cluster is not performed for all the data D, each data D is finally classified into two clusters, that is, two data groups. In this example, each data D is classified into a data group DG1 for a plurality of data D represented by white circles and a data group DG2 for a plurality of data D represented by crosses. In this example, the number of clusters for classifying each data D is set to 2 to classify the data D into two data groups. However, more data can be obtained by increasing the number of clusters. It can also be classified into groups.

  Then, by performing the clustering process as described above, each of the normalized vectors is classified into a plurality of data groups based on the number of data groups to be classified in advance. Create a representative vector that represents

  FIG. 10 shows an example of the created representative vector. In FIG. 10, on the graph in which the horizontal axis represents the number of elapsed periods and the vertical axis represents the product purchase frequency, a plurality of normalized vectors represented by the normalized vector table of FIG. A representative vector is illustrated.

  In FIG. 10, when the normalized vectors are classified into a plurality of data groups by the clustering process, next, normalized vectors of a plurality of consumers included in the data group for each of the plurality of data groups. Are averaged for each data group, and a representative vector that individually represents each data group is created. A vector represented by data obtained by averaging vector data for each data group corresponds to a representative vector that individually represents each data group. Therefore, the representative vector is an averaged product purchase tendency of a plurality of consumers included in the data group, that is, a representative vector. When the K-means method is used as the clustering process, the representative vector creation method and the above-described procedure for updating the center position of each cluster are the same. In this way, when the calculated value of the clustering process can be directly regarded as a representative vector, this can be used as a representative vector.

  In this example, a plurality of normalized vectors represented by the normalized vector table of FIG. 8 are classified into four data groups by clustering processing, thereby generating four representative vectors RV1, RV2, RV3, and RV4. Has been.

  For example, the representative vector RV1 is a vector corresponding to a consumer group that tends to purchase a product after a relatively long time has elapsed since the product was sold. On the other hand, the representative vector RV4 corresponds to a consumer group that has a peak product purchase frequency in the number of elapsed periods “5” and tends to purchase the product after a relatively short time since the product was sold. It is a vector.

  Note that the number of data groups classified into the plurality of data groups may be increased, and the number of representative vectors to be created may be increased. Alternatively, the number of data groups classified into the plurality of data groups may be further reduced, and the number of representative vectors to be created may be further reduced.

  In the present embodiment, when a plurality of representative vectors are created based on the plurality of data groups classified as described above, each of the plurality of consumers is then assigned to one of the created representative vectors. On the other hand, it is classified and associated. Specifically, the distance between the normalized vector related to the consumer and each representative vector is calculated. Then, the consumer is associated with the representative vector whose calculated distance is closest. By performing this process for all consumers included in the purchase history information extracted as the data sample, each of the plurality of consumers related to each of the normalized vectors is changed to any one of the representative vectors. Can be classified in association with each other.

  FIG. 11 shows an example of a correspondence result between a plurality of consumers and a plurality of representative vectors.

  In the example shown in FIG. 11, each of the consumer IDs related to a plurality of normalized vectors represented by the normalized vector table of FIG. 8 is represented by four representative vectors RV1, RV2, RV3, RV4 shown in FIG. Based on this distance, any one of the four representative vectors RV1, RV2, RV3, and RV4 is classified and associated.

  For example, the consumer ID “7” is the representative vector RV2, the consumer ID “43” is the representative vector RV3, the consumer ID “461” is the representative vector RV1, the consumer ID “589” is the representative vector RV2, and the consumer ID “V”. 629 ”is the representative vector RV2, the consumer ID“ 769 ”is the representative vector RV4, the consumer ID“ 860 ”is the representative vector RV2, the consumer ID“ 907 ”is the representative vector RV3, and the consumer ID“ 1220 ”is the representative vector RV1. The consumer ID “1325” is classified and associated with the representative vector RV3.

  In the present embodiment, the consumer mobile terminal 400 represented by the consumer ID that is classified and associated with one of the plurality of representative vectors created in this manner has a predetermined product at a predetermined recommended time. Output recommended information. Specifically, a recommended information plan including information such as recommended products and recommended times is created based on the peak time of the product purchase frequency corresponding to the created representative vectors. Then, every time when the recommended time related to the representative vector comes according to the created recommended information plan, the recommended information of the predetermined product is output to the mobile terminals 400 of a plurality of consumers associated with the representative vector. To do.

  Note that the content of the recommended information plan is different for each of the plurality of representative vectors. In the recommended information plan related to the representative vector RV1 shown in FIG. 10, for example, for a plurality of consumers associated and classified in the representative vector RV1, recommendation is started later after the sale of the product, and the entire recommendation is made. It includes contents such as increasing the number of times, extending the recommended period and the recommended interval once, and ending the entire recommendation later. On the other hand, in the recommended information plan related to the representative vector RV4 shown in FIG. 10, for example, for a plurality of consumers associated and classified in the representative vector RV4, recommendation is started early after the sale of the product, It includes contents such as reducing the recommended number of times, shortening the recommended period and recommended interval, and ending the entire recommendation earlier.

  As described above, when the recommended information is output to the consumer's mobile terminal 400, the received recommended information is displayed on the display unit 401 of the consumer's mobile terminal 400.

  Control contents executed by the control unit 201 of the server 200 will be described with reference to FIG.

  In FIG. 12, for example, operation input information on the keyboard 221 and mouse 222 connected to the server 200 is input to the control unit 201 via the input control unit 204, so that it is represented at a “START” position in the figure. This flow is started.

  First, in step S100, the control unit 201 accesses the product purchase history database 214 and executes information acquisition processing for acquiring the purchase history information of a plurality of consumers for a plurality of products. This step corresponds to an information acquisition procedure. Although the detailed contents will be described later with reference to FIG. 13, in this step, the control unit 201 acquires purchase history information separately for each category ID.

  Thereafter, in step S200, the control unit 201 acquires the product purchase history acquired in step S100, specifically extracted as a data sample in step S430 of FIG. 15 described later (the same applies to FIG. 12 below). Based on the purchase history information stored in the database 214, a vectorization process is executed for creating a vector for each predetermined vectorization interval for each consumer. This step corresponds to a vectorization procedure and functions as vectorization means. This detailed content will be described later with reference to FIG.

  In step S10, the control unit 201 normalizes the vectors created for each consumer in step S200 using a known method, and normalizes the norm of each vector to 1. Note that the vector before normalization (see the horizontal row in the vector table shown in FIG. 7) and the vector after normalization (see the horizontal row in the normalized vector table shown in FIG. 8) are described above. The relationship is expressed by (Formula 1).

  Thereafter, in step S20, the control unit 201 uses a method such as a known K-means method (see FIG. 9) or a self-organizing map for the vector created in step S200 and normalized in step S10. And clustering processing to classify the data into a plurality of data groups. This step corresponds to a data group classification procedure and functions as a data group classification means. Note that the normalization processing in step S10 may be omitted, and the vectors created in step S200 may be classified into a plurality of data groups by clustering processing.

  Then, the process proceeds to step S30, and the control unit 201 creates a representative vector (see FIG. 10) that individually represents each of the plurality of data groups classified in step S20. Specifically, for each of the data groups, the normalized vector data of a plurality of consumers included in the data group is averaged for each data group to create a representative vector for each data group. To do. This step corresponds to a representative vector creation procedure and also corresponds to a representative vector creation means.

  Thereafter, in step S40, the control unit 201 classifies each of the plurality of consumer IDs in association with any of the plurality of representative vectors created in step S30. Specifically, for each of the plurality of vectors normalized in step S10 corresponding to each of the plurality of consumer IDs included in the purchase history information acquired in step S100, each of the plurality of representative vectors Is calculated for each vector. Then, based on the calculation result of the distance, each of the plurality of consumer IDs corresponding to the plurality of vectors is associated with the representative vector having the closest calculated distance. This step corresponds to the consumer classification procedure and also corresponds to the consumer classification means.

  Then, the process proceeds to step S50, where the control unit 201 selects one of the representative vectors based on the result of creating the representative vector in step S30 and the result of associating each consumer ID with the representative vector in step S40. Create a recommended information plan for a plurality of consumers classified in association with. This recommended information plan includes information such as recommended time and recommended products.

  After that, in step S60, the control unit 201 recommends the recommended product information for the consumer mobile terminal 400 classified in association with any of the plurality of representative vectors in step S40 at a predetermined recommended time. Is output. Specifically, according to the recommended information plan created in step S50, recommended information for a predetermined product is acquired from the above-described product database 212 every time the recommended time related to the representative vector comes. Then, the recommended information is output to the mobile terminals 400 of a plurality of consumers associated with the representative vector. Thereby, the received recommendation information is displayed on the display unit 401 of the mobile terminal 400 of the consumer. Thereafter, this flow is terminated. This step corresponds to a recommended information output procedure and functions as a recommended information output unit.

  In addition, the control part 201 performs the process of said step S200 and said step S10-step S60 separately for every category ID about all the category IDs contained in the purchase history information acquired separately for every category ID in said step S100. To do. For example, when purchase history information relating to category IDs “1” to “4” is acquired in step S100, step S200 and step S10 are separately performed for the category IDs “1” to “4”. -The process of step S60 is performed.

  The detailed contents of step S100 will be described with reference to FIG.

  In FIG. 13, first, in step S <b> 110, the control unit 201 accesses the product purchase history database 214 and acquires the purchase history information separately for each category ID, that is, for each product category. This step corresponds to the category-specific acquisition procedure.

  After that, in step S300, the control unit 201 purchases the purchase history within a specific period such that the number of corresponding consumers among the purchase history information of the plurality of consumers acquired in step S110 is a predetermined value or more. A history extraction process for extracting information is executed. This step corresponds to a history extraction procedure. This detailed content will be described later with reference to FIG.

  Then, the process proceeds to step S400, and the control unit 201 purchases the product from the sales start time of the product determined for each product in the purchase history information within the specific period extracted in step S300. A data extraction process for extracting purchase history information as a data sample until the value becomes equal to or less than a predetermined value is executed. This step corresponds to a data extraction procedure. This detailed content will be described later with reference to FIG. Thereafter, this routine is terminated.

  In addition, the control part 201 performs the process of said step S300 and step S400 separately for every category ID about all the category IDs contained in the purchase history information acquired for every category ID in said step S110.

  Details of step S300 will be described with reference to FIG.

  In FIG. 14, first, in step S310, the control unit 201 determines the corresponding purchase date and time among a plurality of consumer IDs included in the purchase history information based on the purchase history information acquired in step S110 of FIG. The number of consumer IDs corresponding to a predetermined period, for example, a period from the current date and time to six months ago, is counted.

  Thereafter, in step S320, the control unit 201 determines whether the number of consumer IDs counted in step S310, in other words, the number of consumers has reached a predetermined value or more, for example, 1000 or more. If the number of consumer IDs counted in step S310 is greater than or equal to the predetermined value, the determination is satisfied and the process moves to step S330.

  In step S330, the control unit 201 extracts purchase history information in which the purchase date and time included in the purchase history information falls within the predetermined period from the purchase history information acquired in step S110 of FIG. . In this case, the predetermined period corresponds to a specific period. Thereafter, this routine is terminated.

  On the other hand, in step S320, if the number of consumer IDs counted in step S310 is less than the predetermined value, the determination in step S320 is not satisfied, and the process proceeds to step S340.

  In step S340, the control unit 201 extends the predetermined period as appropriate so that the number of consumer IDs included in the purchase history information is based on the purchase history information acquired in step S110 of FIG. Detects the period of the predetermined value.

  Then, the process proceeds to step S350, and the control unit 201 matches the purchase date / time included in the purchase history information among the purchase history information acquired in step S110 of FIG. 13 within the period detected in step S340. The purchase history information to be extracted is extracted. In this case, the period detected in step S340 corresponds to a specific period. Thereafter, this routine is terminated.

  Details of step S400 will be described with reference to FIG.

  In FIG. 15, first in step S410, the control unit 201 determines the purchase history information based on the purchase history information within the specific period extracted in step S300, specifically, step S330 or step S350 in FIG. For all product codes included in the above, in other words, for all products, the trend peak position P (see FIG. 5) appearing in the time transition curve of the number of purchases from the sales start date and time is detected for each product code.

  Thereafter, in step S420, the control unit 201 performs the above-described time transition for all product codes included in the purchase history information based on the purchase history information within the specific period extracted in step S330 or step S350 of FIG. The date and time when the number of purchases from the sales start date and time after the trendy peak position P detected in step S410 appearing on the curve is a predetermined value, for example, 0.01 times the number of purchases corresponding to the trendy peak position P, Detect for each product code.

  Then, the process proceeds to step S430, and the control unit 201 purchases all the product codes included in the purchase history information among the purchase history information within the specific period extracted in step S330 or step S350 in FIG. The purchase history information corresponding to the purchase date / time included in the history information from the sales start date / time to the date / time detected in step S420 is extracted as a data sample. Thereafter, this routine is terminated.

  Details of step S200 will be described with reference to FIG.

  In FIG. 16, first in step S210, the control unit 201 sets the vectorization interval to n weeks, for example, one week. Note that n is an integer of 1 or more. This step corresponds to the interval setting procedure.

  Thereafter, in step S500, the control unit 201 executes a vector creation process for creating a vector for each set vectorization interval for each consumer. This step corresponds to a vector creation procedure. This detailed content will be described later with reference to FIG.

  Then, the process proceeds to step S220, where the control unit 201 determines that all consumer IDs included in the purchase history information of the data sample created in step S500, that is, a vector that is zero among vectors corresponding to all consumers. The number of elements, that is, the number where the numerical value in each column of the vector table (see FIG. 7) is “0” is counted.

  Thereafter, in step S230, the control unit 201 determines whether or not the number of elements in the vector counted as zero in step S220 is equal to or less than a predetermined value, for example, 10% or less of all the elements in the vector. judge. If the number of elements in the vector that becomes zero is greater than the predetermined value, the determination is not satisfied and the routine goes to Step S240. This step 230 corresponds to a zero vector determination procedure.

  In step S240, the control unit 201 sets the value of the variable n to twice the previous value, and resets the vectorization interval to n weeks. That is, the vectorization interval is set longer than before. This step corresponds to the interval resetting procedure. Note that the value of the variable n is not limited to 2 times, but may be an integer multiple such as 3 times, 4 times, 5 times, etc.

  And the control part 201 repeats step S240, step S500, step S220, and step S230 until the number of elements in the vector used as said zero becomes below the said predetermined value, and the determination of said step S230 is satisfy | filled. When the number of elements in the vector that becomes zero becomes equal to or smaller than the predetermined value, the determination in step S230 is satisfied, and this routine is terminated.

  Details of step S500 will be described with reference to FIG.

  In FIG. 17, first, in step S510, the control unit 201 initializes the vector table, and sets all the numerical values in each column of the vector table to “0”.

  Thereafter, in step S520, the control unit 201 selects and acquires purchase history information related to one product purchase history from the purchase history information including a plurality of product purchase histories extracted in step S430 of FIG. . This step corresponds to a history acquisition procedure. Note that purchase history information regarding all purchase histories may be acquired in a lump.

  Then, the process proceeds to step S530, and the control unit 201 calculates the elapsed time until the purchase of the product by the consumer based on the purchase date and time and the sales start date and time included in the purchase history information acquired in step S520 (details). (See FIG. 6). This step corresponds to an elapsed time calculation procedure.

  Thereafter, in step S540, the control unit 201 divides the elapsed time until the purchase of the product by the consumer calculated in step S530 by the vectorization interval set in step S210 or step S240 of FIG. And the fraction after the decimal point of the calculated numerical value is rounded up and calculated as the number of elapsed periods (refer to FIG. 6 for details). This step corresponds to an elapsed period number calculation procedure.

  Then, the process proceeds to step S550, and the control unit 201 uses the calculation result of the number of elapsed periods in step S540 to calculate the number of consumer product purchase histories as a numerical value in the corresponding column of the vector table, that is, By incrementing the elements in the vector, counting is performed separately for each number of elapsed periods (see FIG. 6 for details). This step corresponds to an elapsed period number counting procedure.

  Thereafter, in step S560, the control unit 201 determines whether or not all the purchase history information extracted in step S430 in FIG. 15 has been acquired in step S520. If not all have been acquired yet, the determination is not satisfied, and the process returns to step S520 to acquire purchase history information that has not yet been acquired, and the same procedure is repeated. When all the purchase history information extracted in step S430 in FIG. 15 is acquired, the determination in step S560 is satisfied, and this routine ends. It should be noted that when the vector creation process in step S500 ends, the horizontal row of the vector table indicates the number of purchases of a plurality of products in a category by one consumer and the purchase of the plurality of products. This is a vector corresponding to the consumer's product purchase frequency.

  As described above, in the present embodiment, a vector is created for each consumer based on the purchase history information acquired from the product purchase history database 214 (see FIG. 2). Then, the created vectors are classified into a plurality of data groups by clustering processing. Then, after a representative vector for each data group is created, each consumer is classified in association with one of the representative vectors.

  As described above, in the present embodiment, vectors created based on purchase history information of each of a plurality of consumers are classified into a plurality of data groups by clustering processing, and each consumer is represented by one of the data groups. Classify in association with the vector. In this way, unlike the conventional method that forcibly applies each consumer to a given consumption model, an analysis that flexibly reflects the usage trend of multiple consumers, the purchase timing in the above example, and each data group For each, a typical use time, the purchase time in the above example, can be calculated. As a result, the usage attributes of each consumer can be accurately grasped. Further, unlike the conventional method, it is not necessary for the analyst to manually set the time interval for analysis, and the labor burden on the analyst can be reduced.

  Here, as a consumer's tendency, the number of purchases and the purchase cycle may differ depending on the category to which the product belongs. Therefore, in the present embodiment, purchase history information of a plurality of consumers for a plurality of products is acquired for each product category (see step S110 in FIG. 13). Thereby, the analysis performance can be further improved.

  In the present embodiment, in particular, purchase history information within a specific period in which the number of corresponding consumers is equal to or greater than a predetermined value among purchase history information of a plurality of consumers acquired from the product purchase history database 214. (See step S310 to step S350 in FIG. 14). From the purchase history information within the specific period extracted in this way until the purchase date and time by the consumer of the product from the sales start date and time determined for each product is equal to or less than a predetermined value. Are further extracted as data samples (see steps S410 to S430 in FIG. 15). As a result, the number of consumers that are subject to history extraction can always be ensured to be equal to or greater than a predetermined value, so that the analysis accuracy can be ensured with high accuracy. In addition, data that is substantially unnecessary for analysis can be quickly and smoothly analyzed by excluding the data from the analysis target.

  In this embodiment, in particular, a vectorization interval that is an interval for generating the vector is set, and a vector for each set vectorization interval is generated for each consumer. Thereafter, it is determined whether or not the number of elements in the vector that is zero among the created vectors is equal to or less than a predetermined value. At this time, if the number of elements in the vector that becomes zero is larger than a predetermined value, the vectorization interval is set longer than before. Then, the above processing is repeated until the number of elements in the vector that is zero among the generated vectors is equal to or less than a predetermined value. As described above, by reducing the number of elements in the vector that becomes zero as much as possible, the curve representing the vector can be made as smooth as possible, and a highly accurate analysis can be performed.

  In step S210 and step S240 in FIG. 16, the vectorization interval is set in units of n weeks, where n is an integer of 1 or more. Thus, by making the number of Saturdays, Sundays, and weekdays included in each vectorization interval the same, the influence on the consumer's purchase history due to different human life patterns on Saturdays, Sundays, and weekdays, It can be evenly included in each vectorization interval. As a result, the analysis can be performed with higher accuracy.

  In the present embodiment, the purchase history information accumulated in the product purchase history database 214 is acquired. In the above example, the purchase history information extracted as a data sample in step S300 in FIG. 14 is acquired (see FIG. 17). (See step S520). Then, using the purchase date and time of the consumer included in the acquired purchase history information and the sales start date and time of the product included in the purchase history information, an elapsed time until the purchase of the consumer is calculated ( (See step S530). Thereafter, the calculated elapsed time is divided by the set vectorization interval to calculate the number of elapsed periods (see step S540). Then, using the calculation result, the product use history of one consumer for each of the plurality of products is counted separately for each number of elapsed periods (see step S550). Thereby, the vector corresponding to each consumer's product purchase frequency can be calculated.

  In the present embodiment, in particular, recommended information on a predetermined product is output at a predetermined recommended time to the mobile terminals 400 of consumers that are classified in association with any of a plurality of representative vectors. As a result, for those who tend to purchase a product in a relatively short time after being put on the market, the product is recommended early after the product is put on the market, and a relatively long time has passed since the product was put on the market. For those who tend to purchase after the lapse of time, the product can be recommended later after the product is introduced. As a result, effective product recommendations can be made according to the individuality of each consumer and purchase trends.

  The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit and technical idea of the present invention. Hereinafter, such modifications will be described.

(1) Case of analyzing product usage trend based on singing trend in karaoke In the above embodiment, the content of analyzing the product usage trend based on purchase history information of the product has been described. Also included is music data sung by the consumer. That is, when a consumer who is a karaoke user sings in a karaoke room, the distribution of music data by a music distribution company to a karaoke device is equivalent to the sale of the music data, and for consumers to purchase the music data. Equivalent to. Therefore, the consumer's product use tendency may be analyzed based on the purchase history information of the music data corresponding to the singing history in karaoke.

  The overall configuration of the product recommendation system 1 ′ that realizes such a system will be described with reference to FIG.

  In FIG. 17, a product recommendation system 1 ′ includes a karaoke device 100 installed in a karaoke room KR, a server 200, a base station 300, a mobile terminal 400 carried by each consumer, and a karaoke device via a network NW. 100 has a music distribution company side terminal 500 that distributes music data. The configurations of the server 200, the base station 300, and the mobile terminal 400 are the same as those in the above-described embodiment.

  The karaoke device 100 includes a karaoke device main body 110, a display device 120, and a microphone 115. The karaoke apparatus main body 110 reproduces music data distributed from the terminal 500 on the music distribution company side via the network NW. The display device 120 displays a video related to singing when the music data is reproduced. The microphone 115 is connected to the karaoke apparatus main body 110 and is used for singing by consumers. The karaoke apparatus main body 110 and the server 200 are connected so as to be able to communicate with each other via a network NW. When a consumer reproduces music data in the karaoke room KR, the purchase history information of the music data is displayed for each music data. Are stored in the product purchase history database 214 of the server 200.

  The stored contents of the product purchase history database 214 in this case are the same as those shown in FIG. That is, in this modification, the category ID and the product code shown in FIG. 2 correspond to the music data identification information. The sales start date and time corresponds to the distribution start time when the music data is first distributed to the market from the terminal 500 on the music distribution company side. The consumer ID is identification information of the consumer who purchased the music data, that is, the karaoke user. The purchase date and time corresponds to the singing action time when the consumer M first reproduced the music data in the karaoke room KR.

  Therefore, the merchandise usage tendency analysis method in the present modification is the same as in the above-described embodiment. That is, after acquiring purchase history information corresponding to the singing history at karaoke from the product purchase history database 214, the control unit 201 creates a vector for each predetermined vectorization interval for each consumer. Then, the created vectors are classified into a plurality of data groups by clustering processing, and representative vectors for the plurality of data groups are created. When a plurality of representative vectors are created in this way, each of the plurality of consumers is classified in association with one of the plurality of representative vectors.

  Effects similar to those of the above-described embodiment can also be obtained by the modification described above.

  In addition, in the above, the arrow shown in each figure of FIG.1, FIG.18 etc. shows an example of the flow of a signal, and does not limit the flow direction of a signal.

  Further, the flowcharts shown in FIGS. 12 to 17 do not limit the present invention to the procedure shown in the above-described flow. May be.

  In addition to those already described above, the methods according to the above-described embodiments and modifications may be used in appropriate combination.

  In addition, although not illustrated one by one, the present invention is implemented with various modifications within a range not departing from the gist thereof.

1 Product recommendation system 1 'Product recommendation system 200 server (first information terminal, product usage trend analysis system, product recommendation system)
214 Product Purchase History Database (Product Usage History Database)
400 Mobile terminal (second information terminal)

Claims (10)

In accordance with operation input information on the first information terminal on the analyst side, access to a product use history database storing use history information in which product identification information and use histories of a plurality of consumers for the product are associated, An information acquisition procedure for acquiring usage history information of multiple consumers for multiple products,
A vectorization procedure for creating a product usage frequency vector for each predetermined vectorization interval for each consumer based on the usage history information accumulated in the product usage history database acquired in the information acquisition procedure;
A data group classification procedure for classifying the product use frequency vector created in the vectorization procedure into a plurality of data groups by clustering, and a plurality of data groups classified in the data group classification procedure are individually represented. A representative vector creation procedure for creating a representative vector to be performed;
A product usage trend analysis comprising: a consumer classification procedure for classifying each of the plurality of consumers in association with any of the plurality of representative vectors created in the representative vector creation procedure Method. The vectorization procedure is:
An interval setting procedure for setting the vectorization interval;
A vector creation procedure for creating a product usage frequency vector for each set vectorization interval for each consumer;
Of the vectors created in the vector creation procedure, a zero vector determination procedure for determining whether or not the number of elements in the vector that becomes zero is equal to or less than a predetermined value;
An interval resetting procedure for setting a vectorization interval longer than before when the number of elements in a vector that becomes zero in the zero vector determination procedure is greater than a predetermined value;
2. The interval resetting procedure, the vector creating procedure, and the zero vector judging procedure are repeated until the number of elements in a vector that becomes zero in the zero vector judging procedure becomes a predetermined value or less. Product usage trend analysis method. The interval setting procedure and the interval resetting procedure are:
3. The product usage trend analysis method according to claim 2, wherein n is an integer of 1 or more, and the vectorization interval is set in units of n weeks. The vector creation procedure includes:
A history acquisition procedure for acquiring the usage history information accumulated in the product usage history database;
Using the date and time of use by the consumer included in the usage history information acquired in the history acquisition procedure and the time to market for the product included in the usage history information, the elapsed time until use of the consumer is calculated. The elapsed time calculation procedure to calculate,
Dividing the elapsed time calculated in the elapsed time calculation procedure by the vectorization interval set in the interval setting procedure or the interval resetting procedure, and calculating an elapsed period number calculating procedure;
Using a result of the calculation of the number of elapsed periods, and a procedure for counting the number of elapsed periods for counting the consumption history of one consumer for each of a plurality of products for each number of elapsed periods. The product usage trend analysis method according to claim 3. The information acquisition procedure includes:
The product usage tendency according to any one of claims 1 to 4, further comprising: a category-based acquisition procedure for acquiring the usage history information of a plurality of consumers for a plurality of products for each category of the product. Analysis method. The information acquisition procedure includes:
History extraction procedure for extracting the usage history information within a specific period such that the number of applicable consumers is equal to or greater than a predetermined value among the usage history information of the plurality of consumers acquired by the category-specific acquisition procedure The product usage trend analysis method according to claim 5, further comprising: The information acquisition procedure includes:
Of the usage history information within the specific period extracted in the history extraction procedure, the consumer's usage for the product is determined to be less than or equal to a predetermined value from the market launch time of the product determined for each product. The product usage trend analysis method according to claim 6, further comprising a data extraction procedure for extracting the usage history information up to the date as a data sample. In accordance with operation input information on the first information terminal on the analyst side, access to a product use history database storing use history information in which product identification information and use histories of a plurality of consumers for the product are associated, An information acquisition procedure for acquiring usage history information of multiple consumers for multiple products,
A vectorization procedure for creating a product usage frequency vector for each predetermined vectorization interval for each consumer based on the usage history information accumulated in the product usage history database acquired in the information acquisition procedure;
A data group classification procedure for classifying the product use frequency vector created in the vectorization procedure into a plurality of data groups by clustering processing;
A representative vector creation procedure for creating a representative vector that individually represents each of the plurality of data groups classified in the data group classification procedure;
A consumer classification procedure for classifying each of the plurality of consumers in association with any of the plurality of representative vectors created in the representative vector creation procedure;
Recommended information for outputting recommended information of a predetermined product at a predetermined recommended time to the second information terminal on the consumer side classified in association with any of the plurality of representative vectors in the consumer classification procedure A product recommendation method comprising: an output procedure. A product use history database storing use history information in which product identification information and use history of a plurality of consumers for the product are associated;
Vectorizing means for creating a product usage frequency vector for each predetermined vectorization interval for each consumer based on the usage history information stored in the product usage history database;
Data group classification means for classifying the product use frequency vector created by the vectorization means into a plurality of data groups by clustering processing;
Representative vector creation means for creating a representative vector that individually represents each of the plurality of data groups classified by the data group classification means;
Commodity usage trend analysis, comprising: consumer classification means for classifying each of the plurality of consumers in association with any of the plurality of representative vectors created in the representative vector creation procedure system. A product use history database storing use history information in which product identification information and use history of a plurality of consumers for the product are associated;
Vectorizing means for creating a product usage frequency vector for each predetermined vectorization interval for each consumer based on the usage history information stored in the product usage history database;
Data group classification means for classifying the product use frequency vector created by the vectorization means into a plurality of data groups by clustering processing;
Representative vector creation means for creating a representative vector that individually represents each of the plurality of data groups classified by the data group classification means;
Consumer classification means for classifying each of the plurality of consumers in association with any of the plurality of representative vectors created in the representative vector creation procedure;
Recommended information for outputting recommended information of a predetermined product at a predetermined recommended time to the second information terminal on the consumer side classified in association with any of the plurality of representative vectors by the consumer classification means A product recommendation system comprising an output means.