JP2012027757A - Commodity recommendation device, commodity recommendation system, commodity recommendation method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a commodity recommendation device and the like capable of recommending an appropriate commodity at an appropriate timing without missing an opportunity to recommend.SOLUTION: A commodity recommendation server calculates a relationship between selected commodities selected from a commodity group and other commodities based on commodity purchase history information of a plurality of users stored in a purchase log DB (S12). The commodity recommendation server calculates deviation of release times for every commodity category based on release time information (release dates) of the respective commodities (S13). The commodity recommendation server corrects relevance ratios of the respective commodities for every appropriate commodity category with the deviation used as weighting by time (S14), and presents a commodity with high relevance ratio as a recommendation commodity to a user terminal (S15).

Description

  The present invention relates to a product recommendation device that performs information recommendation using collaborative filtering.

  In recent years, with the development of information technology and the spread of the Internet, information sources have been diversified and a large amount of information has been accumulated. For this reason, information filtering techniques for organizing a large amount of accumulated information and appropriately obtaining information desired by a user have been studied. For example, a service for presenting another product related to a product to a user who wants to purchase the product is provided on an EC site or the like. In this case, a technique called collaborative filtering is used to extract appropriate recommendation information.

In general, in collaborative filtering, the degree of relevance (similarity) between one product and another product is calculated using accumulated information such as past product purchase histories of multiple users, user attribute information, and product attribute information. Then, products with high relevance (similarity) are presented as recommended products. As an information recommendation device using such collaborative filtering technology, for example, technologies shown in Patent Literature 1 and Patent Literature 2 have been proposed.
In Patent Literature 1, the purchase history information of a plurality of users' products and the purchase cycle information are used to extract the history of changes of the user's purchased products, and whether or not it is time for a certain user to change the purchased products. When it is time to make a decision and change the purchased product, an information recommendation device or the like for determining and recommending a product to be recommended to a user based on the purchase change history of another user has been proposed.
Moreover, in patent document 2, while collecting a product usage pattern, the purchase history of goods is accumulate | stored, the past goods purchase history of the existing customer similar to the customer for replacement is extracted, and between the goods before and after replacement by purchase. From the relationship, there has been proposed a product recommendation system or the like that presents a product that is highly satisfactory to the existing customer or a similar offerable product as a product that should be recommended to the replacement target customer.

JP 2004-220152 A JP 2008-225629 A

  However, when related products (similar products, recommended products) are determined using the collaborative filtering technique as described above, the past product purchase history of the entire user is referred to, so there is no record of purchase to some extent. It will not be presented to the top as a related product. For example, “postcard creation software” that is extracted as a related product of the product “printer” is calculated to have low relevance unless a certain record of purchase is left. However, since this “postcard creation software” has a tendency to be released and purchased from autumn to winter, purchase of new products is recommended when the recommended products are determined by collaborative filtering based on past purchase history. When the number increased, there was a problem that the time to recommend was already missed.

  The present invention has been made in view of such problems, and the object of the present invention is to provide a product recommendation device that can recommend an appropriate product at an appropriate time without missing the time to recommend. Is to provide.

  In order to solve the above-described problem, the first invention includes a degree-of-association calculating unit that calculates a degree of association between a selected product selected from a group of products and another product based on purchase history information of a plurality of users. Based on the release time information of each product, the bias calculation means for calculating the bias of the release time for each product category, and the relevance calculation means based on the bias of the release time calculated by the bias calculation means A product recommendation device comprising: a correcting unit that corrects the degree of relatedness for each product category; and a presentation unit that determines and presents a recommended product based on the degree of association corrected by the correcting unit. It is.

According to the product recommendation device of the first invention, the degree of association between the selected product selected from the product group and other products is calculated based on the purchase history information of a plurality of users, and the release time of each product Based on the information, calculate the bias of the release time for each product category, correct the degree of association for each product category based on the calculated bias of the release time, and determine the recommended product based on the corrected degree of association And present.
This makes it possible to recommend products that are biased in release time, such as seasonal products, at an appropriate time prior to the actual purchase behavior of the user, and improve the sales promotion effect.
In addition, since the degree of relevance is corrected using the bias of the release date (release date) that the manufacturer considers to be optimal in terms of sales strategy, it is possible to present reliable recommendation information with a high sales promotion effect.

In the first aspect of the present invention, the apparatus further comprises a bias determination unit that determines whether the bias of the release time calculated by the bias calculation unit is greater than a predetermined threshold, and the correction unit includes the bias determination unit, For a product category for which it is determined that the release time bias is greater than a predetermined threshold, it is desirable to correct the relevance based on the release time bias.
As a result, product categories with a large release time bias are recommended at an appropriate time prior to the user's actual purchase behavior, while product categories with a small release time bias are recommended based on the user's actual purchase behavior. Therefore, it is possible to make a product recommendation flexibly corresponding to the product to be handled.

In the first invention, it is preferable that the correction unit corrects the relevance level based on a bias of a release time in a period corresponding to a date and time when the selected product is selected.
As a result, the recommended product can be determined with an appropriate bias according to the selection date and time of the product, and the product can be recommended without a time lag with respect to the actual purchase behavior of the user.

In addition, in the first invention, when the date and time when the selected product is selected corresponds to a period in which the number of products sold in the product category tends to increase, the correction unit is configured to perform the correction in the period in which the number of products sold is large. It is desirable to correct the relevance based on the bias in release time.
As a result, when a product is selected when the number of sales is increasing or likely to increase, the recommended product is determined using the bias in the release time after the increase, which is different from the bias in the actual release time at the selection date and time. As a result, it is possible to recommend products that will attract attention in the future ahead of the release date, and to effectively support the user's purchase plan.

  A second invention is a product recommendation system in which a user terminal and a server are communicatively connected via a network, and information related to a product is exchanged between the plurality of user terminals and the server. Based on the degree-of-association calculating means for calculating the degree of association between the selected item selected from the group of items by the user terminal and other items based on the purchase history information of a plurality of users, and the release date information of each item , A bias calculating means for calculating the bias of the release time for each product category, and the degree of relevance calculated by the relevance calculating means based on the bias of the sales time calculated by the bias calculating means for each product category A quotient comprising: correcting means for correcting; and presenting means for determining a recommended product based on the degree of association corrected by the correcting means and presenting the recommended product to the user terminal. It is a recommendation system.

According to the product recommendation system of the second invention, for a user terminal that is communicatively connected via a network, the server relates the selected product selected from the product group by the user terminal to another product. Degree based on the purchase history information of multiple users, based on the release time information of each product, calculate the bias of release time for each product category, and based on the calculated bias of release time, The degree is corrected for each product category, a recommended product is determined based on the corrected degree of association, and presented to the user terminal.
As a result, it is possible to recommend products that are biased in release time, such as seasonal products, at an appropriate time prior to the actual purchase behavior of the user, such as seasonal product, in the EC site product recommendation service established on the network. It becomes possible, and it becomes possible to improve the sales promotion effect. In addition, since the manufacturer corrects the degree of association using the release date (release date) that is considered as the optimal release time in terms of sales strategy, it is possible to present the user with highly recommended recommendation information that is highly effective for sales promotion.

  According to a third aspect of the present invention, there is provided a degree-of-association calculating step of calculating a degree of association between a selected item selected from a group of items and another item based on purchase history information of a plurality of users, and release date information of each item. Based on the above, the bias calculation step for calculating the bias of the release time for each product category, and the relevance calculated in the relevance calculation step based on the calculated bias for the sales time is corrected for each product category And a presenting step of determining and presenting a recommended product based on the corrected degree of relevance.

According to the product recommendation method of the third invention, the degree of association between the selected product selected from the product group and other products is calculated based on the purchase history information of a plurality of users, and the release time of each product Based on the information, calculate the bias of the release time for each product category, correct the degree of association for each product category based on the calculated bias of the release time, and determine the recommended product based on the corrected degree of association And present.
This makes it possible to recommend products that are biased in release time, such as seasonal products, at an appropriate time prior to the actual purchase behavior of the user, and improve the sales promotion effect. In addition, since the manufacturer corrects the degree of relevance using the release date (release date) that is considered as the optimal release time in terms of sales strategy, it is possible to present recommended information that is highly effective in sales promotion and reliable.

A fourth invention is a program described in a computer-readable format, and the degree of association between a selected product selected from a product group and other products is based on purchase history information of a plurality of users. The relevance calculation step, the bias calculation step for calculating the bias of the release time for each product category based on the release time information of each product, and the relevance calculation based on the calculated bias of the release time For causing a computer to execute processing including: a correction step for correcting the degree of association calculated in each step for each product category; and a presentation step for determining and presenting a recommended product based on the corrected degree of association It is a program.
According to the fourth invention, it is possible to cause a computer to function as the product recommendation device of the first invention.

  ADVANTAGE OF THE INVENTION According to this invention, the goods recommendation apparatus etc. which can recommend an appropriate goods at an appropriate time without missing the time which should be recommended can be provided.

1 is an overall configuration diagram of a product recommendation system 1 according to the present invention. Hardware configuration diagram of a computer used as the product recommendation server 2, the EC site management server 7, or the user terminal 8 The figure which shows an example of the information which purchase log DB3 hold | maintains The figure which shows an example of the information which goods DB4 hold | maintains The figure which shows the example of the list 5A of the recommended product which the recommended product DB5 holds The figure which shows an example of the information which weighting DB6 classified by period for every goods category hold | maintains The figure which shows the example of the list | wrist 5B of the corrected recommendation goods which the recommendation goods DB5 hold | maintains Flowchart explaining the flow of recommended product determination processing Flowchart for explaining the flow of processing for calculating the weight for each period for each product category The figure which shows an example of the sale product number distribution 7 according to period for every goods category

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.
First, the configuration of the present invention will be described with reference to FIGS.

FIG. 1 is a diagram showing a system configuration of a product recommendation system 1 according to the present invention.
As shown in FIG. 1, the product recommendation system 1 includes a product recommendation server 2, a purchase log DB 3, a product DB 4, an EC site management server 7, and a user terminal 8 that are connected to each other via a network 9. The product recommendation server 2 includes a recommended product DB 5 and a period-specific weight DB 6 for each product category.

  In the present embodiment, as an example, a product recommendation service at an EC site or the like established on the network 9 will be described. That is, the user browses the EC site provided by the EC site management server 7 on the user terminal 8 and acquires desired product information through the EC site, or makes a purchase application or settlement of the product. The EC site management server 7 acquires information on recommended products from the product recommendation server 2 via the network 9 and distributes the information to appropriate user terminals 8.

In addition, in this invention, goods are not limited to goods, but include general goods that are traded and distributed in the same manner as goods, such as electronic contents and services.
Further, the product recommendation server 2 may have a function as the EC site management server 7 together with a function of determining a recommended product. In this case, in FIG. 1, the EC site management server 7 is omitted, and the user terminal 8 communicates with the product recommendation server 2 without going through the EC site management server 7.

  In FIG. 1, a product recommendation server 2 is a computer that determines and presents recommended products related to a product selected by a user terminal 8 from a product group. Information on the product selected by the user terminal 8 is input to the product recommendation server 2 via the EC site management server 7. The product recommendation server 2 executes a process (recommended product determination process; see FIGS. 8 to 9) for determining an appropriate recommended product for the product input from the EC site management server 7 or the user terminal 8. The recommended product determination process will be described later.

  The EC site management server 7 is a computer that manages an EC (Electronic Commerce) site. An EC site is a site that sells products and services on a network 9 such as the Internet, and is an electronic shopping street (“Internet shopping street”, “online mall”, etc.) that combines multiple shopkeeper pages into one site. Also included). The EC site management server 7 obtains product information corresponding to the request from the user terminal 8 from the product DB 4 or the product recommendation server 2 and presents it to the user terminal 8 or purchases purchase history information of the user at the EC site in the purchase log. Register in DB3.

  The user terminal 8 is a computer that browses EC sites. A WEB browser is installed in the user terminal 8, and the EC site management server 7 is accessed in response to an instruction operation from the user, and each page data in the EC site is acquired and displayed so as to be viewable. The user terminal 8 can be realized by a digital television, a mobile phone, a mobile terminal, a game machine equipped with a communication function, or the like instead of a general-purpose computer.

The product recommendation server 2, the EC site management server 7, and the user terminal 8 are realized by a computer shown in FIG.
FIG. 2 is a hardware configuration diagram of a computer used as the product recommendation server 2, the EC site management server 7, or the user terminal 8.
As shown in FIG. 2, the computer includes a control unit 11, a storage unit 12, a media input / output unit 13, a communication control unit 14, an input unit 15, a display unit 16, a peripheral device I / F unit 17, etc. It is connected and configured.

The control unit 11 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like.
The CPU calls a program stored in the storage unit 12, ROM, recording medium or the like to a work memory area on the RAM and executes it, and drives and controls each unit connected via the bus 18. The ROM permanently holds a computer boot program, a program such as BIOS, data, and the like. The RAM temporarily holds the loaded program and data, and includes a work area used by the control unit 11 to perform various processes.

  The storage unit 12 is an HDD (hard disk drive), and stores a program executed by the control unit 11, data necessary for program execution, an OS (operating system), and the like. These program codes are read by the control unit 11 as necessary, transferred to the RAM, and read and executed by the CPU.

The media input / output unit 13 is a media input / output device such as an HD (hard disk) drive, a floppy (registered trademark) disk drive, a PD drive, a CD drive, a DVD drive, or an MO drive, and performs data input / output.
The communication control unit 14 includes a communication control device, a communication port, and the like, is a communication interface that mediates communication with the network 9, and performs communication control.

The input unit 15 is an input device such as a keyboard, a pointing device such as a mouse, or a numeric keypad, and outputs input data to the control unit 11.
The display unit 16 includes a display device such as a liquid crystal panel or a CRT monitor, and a logic circuit (video adapter or the like) for executing display processing in cooperation with the display device, and is input under the control of the control unit 11. Display information is displayed on a display device.

The peripheral device I / F (interface) unit 17 is a port for connecting a peripheral device to the computer, and the computer transmits and receives data to and from the peripheral device via the peripheral device I / F unit 17. The peripheral device I / F unit 17 is configured by USB, IEEE 1394, RS-232C, or the like, and usually includes a plurality of peripheral devices I / F. The connection form with the peripheral device may be wired or wireless.
The bus 18 is a path that mediates transmission / reception of control signals, data signals, and the like between the devices.

FIG. 3 is a diagram showing an example of product purchase history information held in the purchase log DB (database) 3 of FIG. As shown in FIG. 3, the purchase log DB 3 stores past product purchase history information of a plurality of users. The product purchase history information stored in the purchase log DB 3 includes a user ID 31, a purchased product name 32, a purchase Day 33 etc. are included.
The merchandise purchase history information is registered by the EC site management server 7. Moreover, when the manager of the EC site has an actual store in addition to the EC site, it may include product purchase history information collected by a POS server (not shown) of the store. The product purchase history information stored in the purchase log DB 3 is referred to by the product recommendation server 2 and used for determining recommended products.

  FIG. 4 is a diagram illustrating an example of product information held in the product DB (database) 4. As shown in FIG. 4, the product DB 4 stores information related to products such as a product name 41, a product category 42, and a release date 43 of a plurality of products. The product information stored in the product DB 4 is referred to by the product recommendation server 2 and used to determine the recommended product.

  1 shows an example in which the purchase log DB 3 and the product DB 4 are configured by a database server accessible from the network 9, but the present invention is not limited to this configuration example, and the purchase log DB 3 and the product DB 4 are a product recommendation server. 2 or the storage unit 12 of the EC site management server 7 may be used.

FIG. 5 is a diagram showing an example of a recommended product list 5A held in the recommended product DB (database) 5 of the product recommendation server 2. As shown in FIG. In the recommended product list 5A, the degree of association of other products with the selected product is stored. The degree of relevance of each product is calculated by the product recommendation server 2.
Specifically, the recommended product list 5A includes the name of the selected product selected by the user terminal 8 (selected product name 51), the selected product category 52, and the product names of other products related to the selected product (others). Product name 53) and relevance 54 are stored.
The recommended product list 5A in FIG. 5 is calculated using a normal collaborative filtering method that does not take into account the bias of the release date, and is a target of correction in a recommended product determination process described later.

  The selected product name 51 is product identification information such as a product name or product code selected by the user and input to the product recommendation server 2.

The selected product category 52 is information indicating which product category the selected product 51 belongs to. The merchandise category may be the merchandise category used on the EC site itself, or may be a lower or higher category of the merchandise category on the EC site. Moreover, it is good also as a category classified separately from the merchandise category of EC site.
The other product name 53 is product identification information such as the name or product code of another product different from the selected product 51.

  The degree of relevance 54 is obtained by quantifying the degree of relevance between the selected product and other products based on past product purchase history information of a plurality of users in a recommended product determination process (FIG. 8) described later. Here, it is obtained as a value normalized to a numerical value of 0 to 1. The larger the value of the degree of association, the stronger the relationship between the products, and it is presented to the top as something to recommend.

  In the example of FIG. 5, other product names related to the selected product name “printer 01” include “ink 001”, “postcard creation 1”, and “postcard creation 2”. Then, the degree of association between “printer 01” and “ink 001” is “0.9”, the degree of association between “printer 01” and “postcard creation 1” is “0.7”, “printer 01” and “postcard” The degree of association with “Create 2” is obtained as “0.6”.

FIG. 6 is a diagram showing an example of the weight information held in the period-specific weight DB (database) 6 for each product category of the product recommendation server 2.
The product category 61 in FIG. 6 is a category defined in the same manner as the selected product category 52 in the recommended product list 5A described above, and includes, for example, “printer body”, “ink”, “postcard software”, and the like.
The period-specific weight 62 represents the bias of the release date in each period as a numerical value, and is obtained for each product category 61. In the example of FIG. 6, the period is monthly, but the period is not limited to this, and a period according to the sales cycle of the product such as every season or every day of the week may be used.

  In the present invention, the degree of deviation of the release time (release date, release month, release time, etc.) of each product accumulated for each product category is digitized for each predetermined period and used as a weight. For example, the deviation value of the number of sales per month is set as the weight 62 for each period. The product release date information is information that all products have and is reliable information that the manufacturer considers in terms of sales strategy, and is therefore suitable for use in correcting the recommendation time.

  In addition, a standard value is uniformly given as a weight for each period for products with little bias in release time throughout the year. For example, in the example of FIG. 6, since the deviation value is used as the weight, the reference value is “50”. Since the product category “ink” has a small bias in release time, a reference value “50” is assigned as a weight for all periods.

For example, products that belong to the product category “postcard software” generally have an uneven (or increased) product release date from autumn to winter (October to December). Month to December) or a period slightly before the release date (a period in which the number of releases tends to increase) is preferable.
That is, it is known that attention will be paid in the future, but it is also desirable that new products with a small actual purchase number are recommended higher in the period when the number of releases is high, that is, the period to be recommended.

  In the present invention, the product recommendation server 2 corrects the relevance 54 of the recommended product list 5A based on the bias in the release time of the period corresponding to the date and time when the product was selected (periodic weight 62 of the corresponding period). Alternatively, when the date and time when the product is selected falls within a period in which the number of products sold in the product category is increasing, the product recommendation server 2 determines the bias in the release time (the weight for each period) in a period in which the number of sales is high. 62), the relevance 54 of the recommended product list 5A is corrected.

  The corrected degree of relevance 55 is stored in the recommended product DB 5 as the corrected recommended product list 5B, and is transmitted to the user terminal 8 via the EC site.

FIG. 7 is a diagram illustrating an example of the corrected recommended product list 5B.
The corrected recommended product list 5B stores a corrected related degree 55 instead of the related degree 54 in the recommended product list 5A of FIG.

For example, when the product selection date corresponds to the period “December”, the recommended product list 5A in FIG. 5 is referred to by referring to the weight 62 in “December” in the period weight DB 6 by product category in FIG. Each relevance 54 is corrected.
The relevance 54 before correction of the other product name “ink 001” in the recommended product list 5A is “0.9”, but the weight “50” of “December” of the corresponding product category “ink” is multiplied. As a result, the corrected association degree 55 becomes “45”. Further, the relevance 54 before correction of the other product name “postcard creation 1” in the recommended product list 5A is “0.7”, but the weight “83” of “December” of the corresponding product category “postcard creation”. ] And the corrected relevance 55 becomes “58.1”. Similarly, the relevance 54 before correction of the other product name “postcard creation 2” is “0.6”, but the weight “83” of “December” of the corresponding product category “postcard creation” is multiplied, The corrected association degree 55 is “49.8”.
As described above, the relevance level 54 of the recommended product is corrected using the weight 62 for each period, so that an appropriate recommended product is associated with the upper rank at the recommended time and recommended.

In addition, as described above, when the date and time when the product is selected corresponds to a period in which the number of products sold in the product category tends to increase, the product recommendation server 2 determines the release time in the period in which the number of products sold is large. The relevance 54 of the recommended product list 5A may be corrected based on the bias (periodic weight 62).
For example, if the number of products sold in October corresponding to the product category “postcard creation” is greater than in September, even if the product selection date by the user is “September 30”, “9 The product ranking corresponding to “postcard creation” has a higher rank when multiplying by the weight by period 62 of “October” than by multiplying by weight 62 by period of “Month”. In this way, if the relevance level 54 is corrected using the weight of the period with a large number of sales during the period when the number of sales is increasing, the user can receive information about the product before the actual sales date increases. It becomes possible, and it becomes possible to make a purchase plan efficiently.

  Next, the flow of the product recommendation process executed by the product recommendation server 2 will be described with reference to FIGS.

  The product recommendation server 2 executes the following processing in order to present product recommendation information in response to a request from the EC site management server 7. The EC site management server 7 cooperates with the product recommendation server 2 and transmits product recommendation information, information about other products, and the like in response to a request from the user terminal 8.

  When product identification information such as a product name or product code is input from the user terminal 8 or the EC site management server 7 to the product recommendation server 2 (step S11), the control unit 11 of the product recommendation server 2 cooperates. Filtering is executed to obtain a product related (similar) to the input product (step S12). In step S12, the control unit 11 of the product recommendation server 2 acquires product purchase history information of a plurality of users from the purchase log DB 3, and determines the degree of association between each purchased product and the input product input in step S11. The recommended product list 5A as shown in FIG. 5 is calculated. At this time, in addition to the product purchase history information, user attribute information, product attribute information, and the like may be considered. The control unit 11 of the product recommendation server 2 stores the calculated recommended product list 5A in the recommended product DB 5. Items having a relatively high degree of relevance may be extracted and stored as a recommended product list 5A.

Next, the control unit 11 of the product recommendation server 2 calculates the weight 62 for each period for each product category (step S13).
With reference to FIG. 9, the calculation of the weight 62 for each period for each product category in step S13 will be described.

The control unit 11 of the product recommendation server 2 calculates the release product number distribution 7 by period based on the release date information 43 of each product registered in the product DB 4 (step S21).
FIG. 10 shows an example of the distribution of the number of products sold by period 7.
The period-by-period released product number distribution 7 shown in FIG. 10 is obtained by accumulating the number of released products 72 for each period (month) for each product category 71. The product recommendation server 2 acquires the release date information 43 of each product information stored in the product DB 4 and totals it for each period (for example, month) to which the release date belongs. In the case of the product category “postcard software”, “6” in January, “0” in February and March, “1” in April, “16” in November, “14” in December Products will be released.

Next, the control unit 11 of the product recommendation server 2 calculates the standard deviation of the period-by-period released product number distribution 7 for each product category 71 (step S22).
Furthermore, the control unit 11 of the product recommendation server 2 determines whether or not the standard deviation calculated in step S22 is larger than a predetermined threshold (for example, “5”) for each product category 71.

  For the product category 71 whose standard deviation is larger than the predetermined threshold (step S23; Yes), the control unit 11 of the product recommendation server 2 calculates the weight 62 for each period. The weight 62 by period is, for example, a deviation value of the number of released products 72 in each period of the period-by-period released product number distribution 7 (step S24).

  For the product category 71 whose standard deviation is equal to or smaller than the predetermined threshold (step S23; No), the control unit 11 of the product recommendation server 2 does not calculate the weight 62 for each period, and uniformly gives a predetermined reference value for each period. . When the deviation value is used as the weight 62 for each period, the predetermined reference value may be “50” (step S25).

When the period-specific weight 62 for each product category shown in FIG. 6 is calculated as described above, the process proceeds to step S14 in FIG.
The control unit 11 of the product recommendation server 2 obtains the corrected recommended product list 5B by multiplying the relevance 54 of the recommended product list 5A calculated in step S12 by the weight 62 according to the period of the corresponding product category (step 5B). S14). At this time, correction may be made by using the weight 62 for each period corresponding to the current date and time (product selection date), and the date and time when the product is selected tends to increase the number of products sold in the product category. If the period is a certain period, correction may be made by using the period-specific weight 62 in a period with a large number of releases.

  Thereafter, the control unit 11 of the product recommendation server 2 outputs, as a recommended product, the one with the highest corrected relevance 55 in the corrected recommended product list 5B (step S15). The user terminal 8 can acquire recommended product information via the EC site management server 7.

For example, in the collaborative filtering of the conventional method for the selected product “printer 01”, as shown in the recommended product list 5A shown in FIG. 5, the relevance degree is in the order of “ink 001”, “postcard creation 1”, and “postcard creation 2”. Is recommended in that order. For this reason, “ink 001” is always recommended to the top regardless of the period.
On the other hand, according to the present invention, for example, when the recommended product list 5A is corrected using the period-specific weight 62 of “December”, the rank of the corrected relevance level 55 (corrected recommended product list 5B) becomes “postcard creation 1 ”,“ Postcard creation 2 ”,“ Ink 001 ”,. For this reason, “postcard creation 1” is recommended higher than “ink 001” in December, which is a busy period for sales of “postcard software”.
As described above, according to the present invention, a product with a bias in release date can be recommended to a higher rank during an appropriate period to be recommended.

  As described above, in the product recommendation system 1 according to the present embodiment, the product recommendation server 2 stores the degree of association 54 between the selected product selected from the product group and other products in the purchase log DB 3. It is calculated based on product purchase history information of a plurality of users. Further, the product recommendation server 2 calculates the bias of the release time for each product category based on the release time information (release date) of each product. Then, using this bias as the weight 62 for each period, the relevance level 54 of each product is corrected for each corresponding product category, and a product with a high corrected relevance level 55 is presented as a recommended product.

For this reason, when determining recommended products using collaborative filtering based on past product purchase history information of multiple users, even if the product has a small actual purchase history, it is at a time when the number of releases is high or close to that time As a result, it will be recommended as a recommended product.
This makes it possible to recommend products that are biased in release time, such as seasonal products, at an appropriate time prior to the actual purchase behavior of the user, and improve the sales promotion effect.
In addition, since the manufacturer corrects the degree of relevance using the bias of the release date (release date) that is considered as the optimal release time in terms of sales strategy, it is possible to present highly recommended recommendation information with high sales promotion effect.
In addition, for products that are not biased by period, a uniform reference value is given as a weight, thereby enabling efficient calculation processing.

Furthermore, if the degree of relevance is corrected based on the bias in the release period of the period corresponding to the date and time when the product was selected, the recommended product can be determined with an appropriate bias according to the date and time of selection of the product, and the user's actual purchase It becomes possible to recommend products with no time lag for actions.
In addition, if the product is selected when the number of sales is increasing or likely to increase, the recommended product should be determined using a bias in the release time after the increase that is different from the actual sales time bias in the selection date and time. For example, products that will attract attention in the future can be recommended prior to the release date, and the user's purchase plan can be effectively supported. For example, products that can be reserved and purchased should be recommended prior to the release date.

In the present embodiment, the deviation value indicating the bias of the release time by period is used as the weight (period-by-period weight 62) used for the correction of the degree of association, but is not limited to the deviation value. Further, although the weight reference value is “50”, the value is not limited to this value. Further, instead of the word “relevance”, the similarity between products may be used.
In addition, it is obvious that those skilled in the art can come up with various changes and modifications within the scope of the technical idea disclosed in the present application, and these naturally belong to the technical scope of the present invention. It is understood.

1 ... Product recommendation system 2 ... Product recommendation server 3 ... Purchase log DB
4 ... Product DB
5 ... Recommended Product DB
5A: List of recommended products 5B: List of recommended recommended products after correction 6: Weight DB by period for each product category
7 ... EC site management server 8 ... User terminal 9 ... Network

Claims (7)

Relevance calculating means for calculating the relevance between the selected product selected from the product group and other products based on the purchase history information of a plurality of users;
Based on the release date information of each product, a bias calculation means for calculating the bias of release time for each product category,
Correction means for correcting the relevance calculated by the relevance calculation means for each of the product categories based on the bias of the release time calculated by the bias calculation means;
Presenting means for determining and presenting recommended products based on the degree of association corrected by the correcting means;
A product recommendation device comprising: A bias determination means for determining whether or not the bias of the release time calculated by the bias calculation means is greater than a predetermined threshold;
The correction means corrects the relevance of the product category determined by the bias determination means that the bias of the release time is larger than a predetermined threshold based on the bias of the release time. Item recommendation device according to Item 1. The correction means includes
The product recommendation apparatus according to claim 1, wherein the degree of association is corrected based on a bias in a sales period in a period corresponding to a date and time when the selected product is selected. When the date and time when the selected product is selected falls within a period in which the number of products sold in the product category tends to increase,
The product recommendation device according to claim 1, wherein the correction unit corrects the degree of association based on a bias of the release time in a period with a large number of releases. A user terminal and a server are communicably connected via a network, and a product recommendation system for exchanging information about products between the plurality of user terminals and the server,
The server
Relevance calculating means for calculating the relevance between the selected product selected from the product group by the user terminal and other products based on purchase history information of a plurality of users;
Based on the release date information of each product, a bias calculation means for calculating the bias of release time for each product category,
Correction means for correcting the relevance calculated by the relevance calculation means for each of the product categories based on the bias of the release time calculated by the bias calculation means;
Presenting means for determining a recommended product based on the degree of association corrected by the correcting means and presenting the recommended product on the user terminal;
A product recommendation system comprising: A relevance calculating step for calculating the relevance between the selected product selected from the product group and other products based on the purchase history information of a plurality of users;
A bias calculation step for calculating the bias of release time for each product category based on the release date information of each product;
A correction step for correcting the relevance calculated in the relevance calculation step for each of the product categories, based on the calculated bias in the release time;
A presentation step for determining and presenting a recommended product based on the corrected degree of association;
A product recommendation method comprising: A program written in a computer-readable format,
A relevance calculating step for calculating the relevance between the selected product selected from the product group and other products based on the purchase history information of a plurality of users;
A bias calculation step for calculating the bias of release time for each product category based on the release date information of each product;
A correction step for correcting the relevance calculated in the relevance calculation step for each of the product categories, based on the calculated bias in the release time;
A presentation step for determining and presenting a recommended product based on the corrected degree of association;
A program for causing a computer to execute processing including

Images