JP6932919B2 - Program and product extraction system - Google Patents

Description

The present invention relates to programs, product extraction systems and product recommendation systems.

For example, in Patent Document 1, in the item recommendation degree presenting device in the recommendation system that extracts and presents recommended items to the target user based on the usage history data of all users, the evaluation value for each item of each user is used in the usage history. Collected as data, calculate the predicted evaluation value for the unevaluated item of the target user from the usage history data, extract the recommended item for the target user based on the predicted evaluation value, refer to the usage history data, and each of the recommended items It is disclosed that the recommendation level for the item is determined and the recommended item is presented together with the recommendation level.

Japanese Unexamined Patent Publication No. 2005-122369

Products may be recommended to consumers based on the consumer's preference distribution and purchase history. When recommending products to consumers in this way, for example, when focusing only on the number of products purchased in the past and recommending products with a large number of purchases, the consumer does not have to recommend them in the first place. There is a possibility of recommending products that you would like to purchase. On the other hand, even if the number of products purchased in the past is small, it is conceivable that the consumer will purchase the product by recommending it to the consumer.
An object of the present invention is to enhance the effect of a recommendation to make a consumer purchase a product, as compared with a case where a product is recommended based only on the number of products purchased in the past.

The invention according to claim 1 has a function of acquiring information on sales history of a product in a predetermined period on a computer, a function of selecting a user to be recommended from a plurality of users, and purchasing the product. A function to calculate a customer recommendation index, which is an index that increases as the number of users increases and increases as the number of purchases by the recommended users decreases, for each product included in the acquired sales history. It is a program that realizes a function of extracting products having a high customer recommendation index from the acquired products included in the sales history and recommending them to the recommended user.
According to a second aspect of the invention, the ability to the calculation, when each product is classified into one of grouping destinations, based on the number of purchase of one product grouping destinations classified as Ru goods, a claim 1 Symbol placement program and estimates the cart of the one of the product.
According to a third aspect of the invention, the ability to the calculated, in response to said time period of sales history information, according to claim 1 or claim 2, wherein the correcting the cart of goods It is a program.
The invention according to claim 4 purchases the product, an acquisition unit that acquires information on the sales history of the product in a predetermined period, a selection unit that selects a user to be recommended from a plurality of users, and the product. A calculation unit that calculates a customer recommendation index, which is an index that increases as the number of users increases and increases as the number of purchases by the recommended users decreases, for each product included in the acquired sales history. It is a product extraction system including a recommendation unit that extracts products having a high customer recommendation index from the acquired products included in the sales history and recommends them to the users to be recommended.

According to the invention of claim 1, the effect of the recommendation to make one consumer purchase a product that has been purchased as compared with the case of recommending a product based only on the number of products purchased in the past. Can be enhanced.
According to the invention of claim 2 , even when the amount of sales history data related to one product is small, the degree of recommendation of one product can be estimated.
According to the invention of claim 3, the accuracy of the recommendation effect can be improved even for a product whose number of purchases varies depending on the season or the like.
According to the invention of claim 4 , the effect of the recommendation to make one consumer purchase the product that has been purchased as compared with the case of recommending the product based only on the number of products purchased in the past. Can be enhanced.

It is a figure which shows the hardware configuration example of the product recommendation apparatus to which this embodiment is applied. It is a block diagram which showed the functional structure example of the product recommendation apparatus in Embodiment 1. FIG. It is a figure for demonstrating the concept of a recommendation index. It is a figure for demonstrating the concept of a recommendation index. It is a figure for demonstrating the relationship between a customer recommendation index and the number of sales increase of a product by recommendation. It is a figure which shows an example of the sales history. It is a figure which shows an example of the classification destination of a product. It is a figure which shows an example of the provisional recommendation product list. (A) and (b) are diagrams showing an example of a recommended product list. FIG. 5 is a flowchart showing an example of a processing procedure in which the product recommendation device outputs a recommended product list from the sales history in the first embodiment. (A) is a diagram showing an example of the number of weekly sales of four products, and (b) is a diagram showing an example of a sales correction coefficient obtained based on the number of weekly sales of four products. It is a block diagram which showed the functional structure example of the product recommendation apparatus in Embodiment 2. FIG. 2 is a flowchart showing an example of a processing procedure in which the product recommendation device outputs a product list from the sales history in the second embodiment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
<Embodiment 1>
[Hardware configuration example of product recommendation device]
FIG. 1 is a diagram showing a hardware configuration example of the product recommendation device 100 to which the present embodiment is applied.
A product recommendation device 100 as an example of a product extraction system or a product recommendation system is a computer device used for recommending a product to consumers (purchasers, users), for example, in a retail business. As shown in the figure, the product recommendation device 100 includes a CPU (Central Processing Unit) 101 as a calculation means, a main memory 102 as a storage means, and a magnetic disk device 103.

Here, the CPU 101 executes various programs such as an OS (Operating System) and an application, and realizes various functions of the product recommendation device 100. Further, the main memory 102 is a storage area for storing various programs and data used for executing the various programs. Further, the magnetic disk device 103 is a storage area for storing input data for various programs, output data from various programs, and the like.

Further, the product recommendation device 100 includes a communication interface (communication I / F) 104 for communicating with the outside, a display mechanism 105 including a video memory and a display, and an input device 106 including a keyboard and a mouse. Be prepared.

[Example of functional configuration of product recommendation device]
FIG. 2 is a block diagram showing a functional configuration example of the product recommendation device 100 according to the first embodiment.
The product recommendation device 100 uses the sales history storage unit 111 that stores the sales history of the product, the sales history acquisition unit 112 that acquires the sales history of the product, and the sales history of the product, and uses the product sales history for each user and each product. It is provided with a customer recommendation index calculation unit 113 for calculating an index (hereinafter, referred to as a customer recommendation index) indicating the degree of recommendation. Further, the product recommendation device 100 includes a provisional recommendation product list generation unit 114 that generates a list (hereinafter referred to as a provisional recommendation product list) in which the customer recommendation index is shown in a list, and a customer recommendation shown in the provisional recommendation product list. A recommended product list generation unit 115 that generates a list of products recommended to each user (hereinafter referred to as a recommended product list) based on an index, a product information storage unit 116 that stores various information about products, and product information. A product information output unit 117 that outputs product information shown in the recommended product list is provided based on the information stored in the storage unit 116.

The sales history storage unit 111 stores the sales history of the product. This sales history shows a sales record for each of a plurality of products that have been sold, and includes information such as the number of products purchased by a user and the number of users who have purchased each product. I'm out. More specifically, for example, in a supermarket that sells products at stores or a retailer that sells products by mail order on the Internet, it is possible to exemplify those showing sales records of various products such as groceries and daily necessities. The sales history may be that of one store, but may be a record of sales at a plurality of stores or a plurality of different companies.

The sales history acquisition unit 112 as an example of the acquisition unit acquires the sales history of the product from the sales history storage unit 111. Here, the sales history acquisition unit 112 may acquire all the sales history stored in the sales history storage unit 111, but for example, the sales history for the past one year or the sales for one month in April last year. The sales history may be acquired by specifying a period such as history. When the sales history is acquired by designating the period, for example, the period may be set in advance, or the period may be set by the input of the user of the product recommendation device 100 or the like.

The customer recommendation index calculation unit 113 as an example of the extraction unit or the output unit calculates the customer recommendation index for each user and each product based on the sales history acquired by the sales history acquisition unit 112. The details of the customer recommendation index will be described later.

The provisional recommendation product list generation unit 114 generates a provisional recommendation product list by using the sales history acquired by the sales history acquisition unit 112 and the customer recommendation index calculated by the customer recommendation index calculation unit 113. In this provisional recommendation product list, various information including a customer recommendation index is shown for each product included in the sales history acquired by the sales history acquisition unit 112. Details of the provisionally recommended product list will be described later.

The recommended product list generation unit 115 as an example of the recommendation unit generates a recommended product list based on the customer recommendation index shown in the provisional recommended product list. Here, the recommended product list generation unit 115 compares the customer recommendation indexes of each product, and extracts a product whose customer recommendation index satisfies a predetermined standard as a product recommended to a certain user. Then, the recommended product list generation unit 115 generates a recommended product list by listing the extracted products. As a product in which the customer recommendation index satisfies a predetermined standard, for example, a certain number of products can be exemplified from those having a large customer recommendation index. Further, for example, a product whose customer recommendation index is equal to or higher than a predetermined threshold value can be exemplified.

Further, the user or the like of the product recommendation device 100 may check the customer recommendation index shown in the provisional recommendation product list and select a product to be recommended to the user. In this case, the recommended product list generation unit 115 extracts the product selected by the user of the product recommendation device 100 or the like as a product whose customer recommendation index satisfies a predetermined standard, and generates a recommended product list. .. The procedure for extracting products recommended to users and the details of the recommended product list will be described later.

The product information storage unit 116 stores various information related to the product. More specifically, in the product information storage unit 116, for each product, for example, a product name, a product description, a product price, a product image, etc., regarding products that have been sold or products that are scheduled to be sold in the future. Information is stored. In addition, the product information storage unit 116 may store information such as a product code, a major classification code, a middle classification code, and a minor classification code, which will be described later, for each product.

The product information output unit 117 outputs information on the products shown in the recommended product list. Here, the product information output unit 117 acquires the information stored in the product information storage unit 116 for the product shown in the recommended product list. Then, the product information output unit 117 transmits the product information acquired from the product information storage unit 116 to, for example, a terminal device or the like possessed by the user who is the target of creating the recommended product list. In this way, the product information is transmitted to the user's terminal device or the like, so that the product is recommended to the user. In this embodiment, since the product is recommended using the customer recommendation index obtained for each user, the recommended product may differ for each user.

Each functional unit constituting the product recommendation device 100 shown in FIG. 2 is realized by the cooperation of software and hardware resources. Specifically, the CPU 101 implements a program that realizes a sales history acquisition unit 112, a customer recommendation index calculation unit 113, a provisional recommendation product list generation unit 114, a recommended product list generation unit 115, a product information output unit 117, and the like, for example, magnetically. These functional units are realized by reading from the disk device 103 into the main memory 102 and executing the process. Further, the sales history storage unit 111 and the product information storage unit 116 are realized by, for example, a magnetic disk device 103. The program executed by the CPU 101 can be provided not only by communication means but also by storing it in a recording medium such as a CD-ROM.

[Concept of recommendation index]
Next, the concept of the recommendation index in the present embodiment will be described.
In retail businesses such as supermarkets, for example, there are top-selling products such as rice and milk that are purchased regularly, and products that are occasionally purchased by a small number of consumers. The best-selling products are considered to be purchased without being recommended to consumers. In addition, products that are occasionally purchased by a small number of consumers are in low demand, and even if they are widely recommended to consumers, the number of products purchased is considered to be small. That is, even if such a product is recommended to consumers, it cannot be expected that sales will be significantly improved. Therefore, in the present embodiment, many consumers are the target of purchase, but the total number of products purchased is smaller than that of the best-selling products such as rice and milk, and the products should be recommended to the consumers.

3 and 4 are diagrams for explaining the concept of the recommendation index.
First, the concept of the recommendation index will be described with reference to FIG. In the graph shown in FIG. 3, the vertical axis represents the total number of products purchased, that is, the total number of products purchased by consumers for each product. The horizontal axis is the product ranking based on the total number of purchases, that is, the products are arranged in order from the left to the right in the figure in descending order of the total number of purchases.

Here, the area A1 is an area in which a product having a large total number of purchases exists among all the products. For example, the above-mentioned hot-selling products such as rice and milk fall under this category. Further, the area A3 is an area in which a product having a small total number of purchases exists among all the products. For example, the above-mentioned products that are occasionally purchased by a small number of consumers are applicable. As described above, the products existing in these areas A1 and A3 cannot be expected to significantly improve their sales even if they are recommended to consumers.

On the other hand, the area A2 is an area in which the total number of purchased products is in the middle of all the products. In this area A2, for example, there are products that many people purchase, but the total number of products purchased is smaller than that of the best-selling products. In other words, the area A2 includes products that should be recommended to consumers in the present embodiment, as described above. However, it is considered that the area A2 also includes, for example, products purchased in large quantities by a specific consumer. Such products purchased in large quantities by a specific consumer are unlikely to be purchased by consumers other than the specific consumer, and even if they are widely recommended to consumers, it cannot be expected that sales will be significantly improved. Therefore, it is required to exclude the products that a specific consumer purchases in large quantities from the products existing in the area A2 and extract the products that should be recommended to the consumers.

Further description will be continued with reference to FIG.
In the graph shown in FIG. 4, the vertical axis is the value obtained by dividing the total number of purchases of the product by the number of purchases of the product, that is, the number of purchases per purchaser. The horizontal axis is the number of people who purchased the product.
Here, the area B1, which is the first quadrant of the four divisions of the coordinate plane, is an area to which a product having a large number of purchases and a large number of purchasers per person belongs. In addition, the product in region A1 shown in FIG. 3 is likely to be included in region B1. Further, the area B3, which is the third quadrant, is an area to which a product having a small number of purchases and a small number of purchasers per person belongs. In addition, the product in region A3 shown in FIG. 3 is likely to be included in region B3. Therefore, in the present embodiment, the products included in the area B1 and the area B3 cannot be said to be products that should be recommended to consumers.

Then, the product in the region A2 shown in FIG. 3 is likely to be included in either the region B2 or the region B4. Here, the area B2, which is the second quadrant, is an area to which a product having a small number of purchases but a large number of purchases per person belongs. In other words, the product in area B2 is likely to be a product purchased in large quantities by a specific consumer, and such a product cannot be said to be a product that should be recommended to consumers.

On the other hand, the area B4, which is the fourth quadrant, is an area to which a product having a large number of purchases belongs, although the number of purchases per person is small. In other words, the products in area B4 are targeted for purchase by many consumers, but are likely to be purchased in a smaller total number than the top-selling products, and are recommended to consumers in the present embodiment. It can be said that it is a product that should be used.

Further, for the products in the region B4 in the fourth quadrant, there are consumers who purchase more and less than the average value of the number of purchases. Here, even if the product in the area B4 is recommended to consumers who purchase more than the average value, it cannot be expected that the sales will be significantly improved. Therefore, it is considered desirable to recommend the products in area B4 to consumers who purchase less than the average value in order to improve sales.

Therefore, in the present embodiment, as an index for identifying the products existing in the area B4 and the area close to the area B4 and recommending the products to consumers (users) whose purchase number is relatively low, the following The value calculated from the formula (1) (customer recommendation index) is used. Further, the formula (2) shows the denominator X in the formula (1), and the formula (3) shows the numerator Y in the formula (1).

As shown in the formula (1), the customer recommendation index is "the ratio of the number of users who purchased the product (i) to all users" (Y), and "the user (u) purchases the product (i)". It is obtained by dividing by "frequency" (X). Further, as shown in the formula (2), the denominator X in the formula (1), that is, "the frequency with which the user (u) purchases the product (i)" is "the user (u) purchases the product (i)". It is obtained by dividing "the total number of purchases" (X2) by "the total number of purchases of the product (i) by all users" (X1). Further, the numerator Y in the formula (1), that is, "the ratio of the number of users who purchased the product (i) to all users" is "the number of users who purchased the product (i)" (Y2) and "total use". It is obtained by dividing by "number of persons" (Y1).

As is clear from the formulas (1) to (3), the value of the customer recommendation index becomes large when the "frequency of the user (u) purchasing the product (i)" which is the denominator X is "low". , And / or the numerator Y, "the ratio of the number of users who purchased the product (i) to all users" is "high". More specifically, the value of the customer recommendation index becomes large when "the number of users who purchased the product (i)" (Y2) is "large" and "all users purchase the product (i)". When the "total number of purchases" (X1) is "large", the "total number of purchases by the user (u) of the product (i)" (X2) is "small". Therefore, the customer recommendation index of the present embodiment means that the larger the value, the higher the degree of recommendation.

As is clear from the formulas (1) to (3), when the user (u) has never purchased the product (i), in other words, "the total number of the user (u) purchased the product (i)". When X2 is 0, the customer recommendation index diverges. Therefore, in the present embodiment, X2 is estimated when X2 is 0. The details of this estimation of X2 will be described later.

The customer recommendation index in this embodiment is used in fields such as information retrieval and document summarization, and can be grasped as the reciprocal of TFIDF (Term Frequency Inverse Document Frequency) for finding the word that characterizes this document from the document. .. TFIDF is expressed as word frequency (TF) divided by document frequency (DF), but if the document corresponds to the user and the word corresponds to the product, the customer recommendation index is the reciprocal of TFIDF. It is understood that it can be grasped as.

It should be noted that FIGS. 3 and 4 are for explaining the product recommendation index used in the second embodiment described later, and strictly speaking, the concept of the customer recommendation index of the present embodiment is completely different. It does not match. Therefore, even if the products belong to the areas A1 and A3 shown in FIG. 3 and the areas B1 to B3 shown in FIG. 4, the numerical value of the customer recommendation index may be large.

[Relationship between customer recommendation index and increase in product sales due to recommendation]
FIG. 5 is a diagram for explaining the relationship between the customer recommendation index and the number of sales increase of products due to the recommendation.
The vertical axis of the graph shown in FIG. 5 is the number of increases in product sales before and after product recommendation. Further, the horizontal axis shows the customer recommendation index, and here, a plurality of users (100 people) are grouped by 10 people in descending order of the customer recommendation index, so that a total of 10 groups are formed. The numerical values shown on the horizontal axis in the figure are the average values of the customer recommendation indexes of the 10 users constituting each group. Then, FIG. 5 shows this product by sending the information of this product to, for example, the terminal device of each user for a specific one product (here, natto) for 100 users. It is a sales result obtained when is recommended.

For example, in the case of the group with the largest value of the customer recommendation index (the group with the average value of the customer recommendation index of 353.74) located on the leftmost side of the horizontal axis, the number of sales increase is 53 before and after the product is recommended. You can see that it is. On the other hand, in the case of the group with the smallest value of the customer recommendation index (the group with the average value of the customer recommendation index of 5.50) located on the rightmost side of the horizontal axis, the number of sales increases by 2 before and after recommending the product. You can see that it is. In addition, it is understood that the larger the average value of the customer recommendation index is, the larger the number of sales increases, that is, there is a positive correlation between the customer recommendation index and the number of sales increases. NS.

[Sales history]
Next, the sales history stored in the sales history storage unit 111 will be described.
FIG. 6 is a diagram showing an example of the sales history.
FIG. 6 illustrates the sales record of the product for 6 times. For such a sales history, for example, a user or the like of the product recommendation device 100 acquires sales history data from the outside and stores it in the sales history storage unit 111. Further, for example, the sales history may be stored by transmitting the sales record of each product from the store via a network (not shown) by the POS (Point of Sales) system.

Here, the "purchase date" indicates the date on which the product was purchased. The "user code" is a number given to a user to identify the user who purchased the product. "Product name" indicates the name of the product. The "major classification code", "middle classification code", and "minor classification code" are numbers assigned to identify the classification destination to which the product is classified. The "major classification code" is a code indicating a major classification, the "middle classification code" is a code indicating a middle classification, and the "minor classification code" is a code indicating a minor classification. The "product code" is a number assigned to a product to identify the product. "Number of purchases" indicates the number of purchased products.

In the example shown in FIG. 6, in order from the top, the product names "Company A unadjusted milk", "Company A coffee milk", "Company B donut", "Company A unadjusted milk", "Company C frozen hamburger", It is recorded that the product of "Company D silken tofu" was purchased. For example, it is recorded that a user with a user code "12345467" purchased one product with the product name "Company A unadjusted milk" on January 21, 2016. Further, for this product, it is recorded that the major classification code is "01", the middle classification code is "01", the minor classification code is "01", and the product code is "098765". Furthermore, the user code is "12345667" for the product with the product name "Company A coffee milk", and it can be said that the user who purchased the product "Company A unadjusted milk" also purchased "Company A coffee milk". .. Furthermore, it is also recorded that another user, that is, a user with the user code "7357542", purchased one product with the product name "Company A unadjusted milk" on January 21, 2016. ..

The "user code" is, for example, a card that is given as information necessary for login when a user logs in to a website and purchases a product, or is presented when a user purchases a product at a store. It is given to such as. Even if the same user makes a purchase, for example, if the purchase is made without logging in to the website or without showing the card at the store, it is not possible to identify the same user. , A different user code may be assigned for each purchase, or the user code itself may not be assigned.

In addition, the "major classification code", "middle classification code", "minor classification code", and "product code" may be assigned different numbers for each store or each company that sells the same product. be. When different numbers are assigned to each store and each company for the same product, the records of the same product can be recorded for all stores and all by associating the numbers for each store and each company using, for example, a recombination table. It will be possible for companies to handle them collectively.

Further, in the example shown in FIG. 6, only the sales record of the product for 6 times (5 types, 8 pieces) is shown, but it is assumed that the information is added to the sales history every time the product is purchased.
Further, although not shown in FIG. 6, the sales history may include data other than the information shown in FIG. For example, information on consumption such as the cost price of a product and the selling price of a product, user attributes such as the gender and age of the user, and information at the time of sale such as the cash register number used when selling the product are sold. It may be included in the history.

[Product classification destination]
Next, the classification destination to which the products are classified will be described in detail.
FIG. 7 is a diagram showing an example of a product classification destination. Here, the classification destination will be described by taking a "beverage" product as an example.
In the example shown in FIG. 7, products are managed in three layers of "major classification", "middle classification", and "minor classification". "Milk beverages" and "soy milk" are provided as major categories. The major categories of "milk drinks" and "soy milk" are further divided into middle categories. For example, "milk drinks" are classified into "milk", "milk drinks", and "yogurt". The middle classification of "milk", "milk drink", and "yogurt" is further divided into sub-classifications. For example, "milk" is subdivided into "unadjusted milk", "low-fat milk", and "calcium milk". Then, each product is classified into one of the sub-categories. For example, a product with the product name "Company A unadjusted milk" is classified into a sub-category "unadjusted milk". The product with the product name "Company A unadjusted milk" is classified into the minor category "unadjusted milk", thereby being classified into "milk" in the middle category and "milk beverage" in the major category.

In addition, in the example shown in FIG. 7, for example, the major classification codes of "milk drink" and "soy milk" are "01" and "02", respectively. Further, for example, the middle classification codes of "milk", "milk drink", and "yogurt", which are the middle classifications of "milk drink", are "01", "02", and "03", respectively. Further, for example, the sub-classification codes of "unadjusted milk", "low-fat milk", and "calcium milk", which are sub-classifications of "milk", are "01", "02", and "03", respectively. Further, for example, the sub-classification codes of "coffee milk" and "fruit milk", which are sub-classifications of "milk beverage", are "01" and "02", respectively.

Such product classifications may be common even if the stores and companies that sell them are different, or they may be classified differently for each store or company. For example, a product with the product name "Company C frozen hamburger" may be classified as "frozen food" in some stores, while it may be classified as "processed meat" instead of "frozen food" in other stores. .. In this way, when the same product is classified into different classification destinations for each store and each company, the same product can be recorded by associating the same products belonging to different classification destinations using, for example, a recombination table. Will be able to be handled collectively by all stores and all companies.

In the example shown in FIG. 7, the case where the classification destination layer is divided into three layers has been described, but the classification destination layer is not limited to three layers, and may be less than three layers or more than three layers. In some cases.

[Temporary Recommended Product List]
Next, the provisional recommendation product list generated by the provisional recommendation product list generation unit 114 will be described.
FIG. 8 is a diagram showing an example of a provisionally recommended product list. FIG. 8 exemplifies a list of provisionally recommended products generated for the user code “1234567”, and information on five types of products is recorded.
Here, the "user code", "product name", "major classification code", "medium classification code", "minor classification code", and "product code" are the same as the items shown in FIG. The explanation is omitted. The "customer recommendation index" is a customer recommendation index calculated by the customer recommendation index calculation unit 113.

In the example shown in FIG. 8, the products with the product names "A company unadjusted milk", "A company coffee milk", "B company donut", "C company frozen hamburger steak", and "D company silken tofu" are listed in order from the top. It has been recorded. Then, the customer recommendation indexes of "Company A unadjusted milk", "Company A coffee milk", "Company B donut", "Company C frozen hamburger steak", and "Company D silken tofu" for the user code "1234567" are They are "1.23", "12.5", "210", "140", and "91.2", respectively. Comparing these five types of products, the customer recommendation index of "Company B donuts" is the highest for the user code "1234567", and the user (user code "1234567") is the highest among the five types of products. It is a product that should be recommended to users who have been granted.

In the example shown in FIG. 8, only information on five types of products is shown, but since information on each product recorded as a sales history is actually shown, six or more types of products are sold. If so, information on six or more types of products is shown.

[Recommended product list]
Next, the recommended product list generated by the recommended product list generation unit 115 will be described.
9 (a) and 9 (b) are diagrams showing an example of a recommended product list. In FIG. 9A, three types of products recommended to the user to which the user code "12345667" is assigned are selected. On the other hand, in FIG. 9B, two types of products recommended to the user to which the user code "12345467" is assigned are selected. The "user code", "product name", "major classification code", "medium classification code", "minor classification code", "product code", and "customer recommendation index" are shown in FIG. 6 or FIG. Since it is the same as the item, the description is omitted here.

First, in the example shown in FIG. 9A, three types of products are selected: "C company frozen hamburger steak", "B company donut", and "D company silken tofu".
Here, the recommended product list generation unit 115, for example, compares the customer recommendation indexes shown in the provisional recommended product list and extracts the top three products to obtain the recommended product list shown in FIG. 9A. Generate. In other words, the recommended product list generation unit 115 has "C company frozen hamburger" and "Company C frozen hamburger", which are the products of the top three customer recommendation indexes in the provisional recommendation product list, as the products whose customer recommendation index meets the predetermined criteria. Extract "B company donuts" and "D company silken tofu" to generate a recommended product list. The number of products to be included in the recommended product list shall be determined based on, for example, the budget required for recommendation and the layout of the website.

Further, in the example shown in FIG. 9B, two products, "Frozen hamburger steak of company C" and "Donut of company B", are selected.
Here, the recommended product list generation unit 115 compares, for example, the customer recommendation indexes shown in the provisional recommended product list, and extracts products (the number is not limited) that exceeds a predetermined threshold value. The recommended product list shown in FIG. 9B is generated. In other words, the recommended product list generation unit 115 is a product whose customer recommendation index is equal to or higher than a threshold value (for example, 100) in the provisional recommended product list as a product whose customer recommendation index satisfies a predetermined standard. "Company frozen hamburger" and "Company B donut" are extracted and a recommended product list is generated. The size of the threshold value of the customer recommendation index may be, for example, predetermined or may be determined by the user who uses the product recommendation device 100.

[Processing procedure to output recommended product list from sales history]
Next, a processing procedure in which the product recommendation device 100 according to the present embodiment outputs a recommended product list from the sales history will be described.
FIG. 10 is a flowchart showing an example of a processing procedure in which the product recommendation device 100 outputs a recommended product list from the sales history.

First, the sales history acquisition unit 112 acquires the sales history of the period to be analyzed (for example, for the latest one year) from the sales history storage unit 111 (step 101). Here, the sales history acquisition unit 112 acquires the sales history, for example, triggered by an operation by a user or the like of the product recommendation device 100.

Next, the customer recommendation index calculation unit 113 selects one user (u) who wants the customer recommendation index (step 102). The user (u) selected in step 102 corresponds to one purchaser. In this example, the customer recommendation index calculation unit 113 actually selects one user code from the sales history acquired in step 101. However, the present invention is not limited to this, and for example, the user code may be selected from the user list in which the user code is listed.

Subsequently, the customer recommendation index calculation unit 113 selects one product (i) for which the customer recommendation index is sought (step 103). In this example, the customer recommendation index calculation unit 113 actually selects one product code from the sales history acquired in step 101. However, the present invention is not limited to this, and for example, the product code may be selected from the product list in which the product code is listed.

Next, the customer recommendation index calculation unit 113 counts the total number of users (Y1) from the sales history acquired in step 101 (step 104), and counts the number of users (Y2) who purchased the product (i). (Step 105). Further, the customer recommendation index calculation unit 113 counts the total number (X1) of all users purchasing the product (i) from the sales history acquired in step 101 (step 106), and the user (u) is the product (u). i) Count the total number of purchases (X2) (step 107). The order of steps 104 to 107 may be changed.

Then, the customer recommendation index calculation unit 113 determines whether or not the total number (X2) of the user (u) purchasing the product (i) acquired in step 107 is 0 (step 108). If a negative determination (NO) is made in step 108, the process proceeds to step 110, which will be described later.

On the other hand, when a positive judgment (YES) is made in step 108, the customer recommendation index calculation unit 113 uses, for example, a known collaborative filtering method to use another user whose purchasing tendency is similar to that of the user (u). The total number of purchases (X2) of the product (i) by the user (u) is estimated from the sales history (step 109), and the process proceeds to step 110. The estimation method of X2 is not limited to collaborative filtering, and various methods may be adopted.

Then, the customer recommendation index calculation unit 113 uses the values acquired in steps 104 to 106 and steps 107 or 109, and uses the above-mentioned equations (1) to (3) to obtain the user (i) for the product (i). u) Calculate the customer recommendation index (step 110).

After the customer recommendation index is calculated in step 110, the provisional recommendation product list generation unit 114 adds the information of the product (i) for which the customer recommendation index is calculated to the provisional recommendation product list of the user (u) (step). 111). Here, if the tentatively recommended product list of the user (u) has not been generated yet, the tentatively recommended product list generation unit 114 newly generates the tentatively recommended product list of the user (u), and the first one. Add information about product (i). Then, the customer recommendation index calculation unit 113 determines whether or not all the products have been selected in the sales history acquired in step 101 (step 112). If a negative determination (NO) is made in step 112, that is, if there is a product that has not yet been selected, the process proceeds to step 103, and the process relating to the next product (i (= i + 1)) is continued.

On the other hand, when a positive determination (YES) is made in step 112, the recommended product list generation unit 115 uses the recommended product list generation unit 115 based on the temporary recommendation product list of the user (u) generated by the temporary recommendation product list generation unit 114. A list of recommended products of the person (u) is generated (step 113). Here, the recommended product list generation unit 115 extracts products that satisfy predetermined criteria based on the customer recommendation index of each product shown in the provisional recommended product list of the user (u), and recommends products. Generate a list. More specifically, the recommended product list generation unit 115 generates a recommended product list by extracting products having a higher customer recommendation index or extracting products having a customer recommendation index equal to or higher than a threshold value. After that, the product information output unit 117 acquires the information of each product shown in the recommended product list of the user (u) from the product information storage unit 116, and outputs the acquired information (step 114), and a series of series. End the process.

If you want to create a recommended product list for another user (u), you may repeat the above procedure and select this other user (u) in step 102.

(First modification)
In the processing procedure shown in FIG. 10, the customer recommendation index was calculated for each product (such as "Company A unadjusted milk"), but the present invention is not limited to this. For example, the total number of purchases of products classified into the classification destination (for example, major classification) to which the product (referred to as product Z) belongs may be estimated as the total number of purchases of product Z as it is. In this case, the customer recommendation index calculation unit 113 calculates the customer recommendation index corresponding to the major classification to which the product Z belongs, for example, using the number of users who purchased the product classified into the major classification to which the product Z belongs. .. Then, the calculated customer recommendation index is estimated as the customer recommendation index of the product Z. In other words, the customer recommendation index calculation unit 113 estimates the customer recommendation index corresponding to the major classification to which the product Z belongs as it is as the customer recommendation index of the product Z. The customer recommendation index corresponding to the middle classification to which the product Z belongs and the customer recommendation index corresponding to the minor classification to which the product Z belongs may be calculated and estimated as the customer index of the product Z.

Further, as another example of calculating the customer recommendation index of the product Z, for example, the customer recommendation index of another product classified into the classification destination (for example, the major classification) to which the product Z belongs is used as it is as the customer recommendation index of the product Z. May be used as. In this case, the customer recommendation index calculation unit 113 estimates, for example, the recommendation index of another product classified into the major classification to which the product Z belongs as the recommendation index of the product Z as it is. Further, the customer recommendation index calculation unit 113 classifies the customer recommendation index of the product Z into, for example, a customer recommendation index of another product classified into the middle category to which the product Z belongs, or a subclass to which the product Z belongs. The customer recommendation index of the product may be estimated as the customer recommendation index of the product B.

(Second modification)
If a product is sold, for example, for a limited time, it is possible that there are few sales records for this product. In the calculation of the recommendation index, it is preferable that the sales record is larger than the predetermined reference amount, for example, the record for one month or more. Therefore, the customer recommendation index calculation unit 113 may estimate the total number of purchases (X1 and X2) of this product by using the sales record of the product related to this product.

For example, assuming that the product (i) sold for a limited time is sold annually, the customer recommendation index calculation unit 113 estimates X1 using the formula (4) shown below, and also formula (5). Is used to estimate X2.

In the formula (4), the "total number (annual) of all users purchasing the higher-class products including the product (i)" refers to the products classified into the classification destination (for example, the major classification) to which the product (i) belongs. , The total number of annual purchases by all users. The "total number of purchases of the product (i) by all users (for a limited time)" is the total number of purchases of the product (i) by all users during the sales period of the product (i). Furthermore, the "total number of all users purchasing higher-class products including product (i) (for a limited time)" is the total use of products classified into the classification destination (for example, major classification) to which product (i) belongs. It is the total number of purchases of the product (i) by the person during the sales period.

In the formula (5), the "total number (annual) of the user (u) purchasing the higher-class products including the product (i)" is the product classified into the classification destination (for example, major classification) to which the product (i) belongs. Is the total number of annual purchases by the user (u). Further, the "total number of purchases of the product (i) by the user (u) (for a limited time)" is the total number of purchases of the product (i) during the sales period of the product (i) by the user (u). Further, the "total number of users (u) purchasing higher-class products including the product (i) (for a limited time)" is the product classified into the classification destination (for example, major classification) to which the product (i) belongs. This is the total number of purchases made by the user (u) during the sales period of the product (i).
The customer recommendation index calculation unit 113 estimates X1 and X2 in this way, and calculates the customer recommendation index of the product (i) by the formula (1).

(Third variant)
The sales volume of vegetables and seafood, whose yield changes depending on the season, and the ingredients of ice cream and hot pot, which are affected by the climate, fluctuate throughout the year.
FIG. 11A is a diagram showing an example of weekly sales of four products (Chinese cabbage, cabbage, ice cream and mackerel).
The vertical axis of the graph shown in FIG. 11A shows the number of sales of each product, and the horizontal axis shows a certain year, that is, from January 1st week to December 4th week on a weekly basis.

As is clear from FIG. 11A, each of these four products has a period in which the number of sales increases and a period in which the number of sales decreases. In addition, the time when the number of sales peaks differs depending on the product. For example, cabbage is sold in large numbers in May and October. In addition, for example, Chinese cabbage has been sold in large numbers since September.

In the processing procedure shown in FIG. 10, the customer recommendation index is calculated based on the sales history of the period to be analyzed. Therefore, when the period is set to less than one year, the seasonal fluctuation of the number of products sold has a great influence on the customer recommendation index.

Therefore, in order to reduce the influence of such seasonal fluctuations, the total number of purchased products may be corrected.
FIG. 11B is a diagram showing an example of a sales correction coefficient obtained based on the number of weekly sales of the four products shown in FIG. 11A.
The sales correction coefficient shown in FIG. 11B is smoothed by averaging the number of sales of each product shown in FIG. 11A for each week together with the two weeks before and after that (5 weeks in total), and further, each Obtained by dividing by the annual average of the number of products sold. Then, the value obtained by dividing the total number of purchased the target products (X1 and X2) by the sales correction coefficient set for the products may be used in the calculation of the customer recommendation index.

<Embodiment 2>
In the first embodiment, the customer recommendation index is calculated for all the products existing in the acquired sales history. On the other hand, in the present embodiment, the products to be recommended to all users are selected in advance from all the products existing in the acquired sales history, and the customers for each user are selected for the selected products to be recommended. The recommendation index is calculated. In the present embodiment, the same reference numerals as those in the first embodiment are designated by the same reference numerals, and detailed description thereof will be omitted.

[Example of functional configuration of product recommendation device]
FIG. 12 is a block diagram showing a functional configuration example of the product recommendation device 100 according to the second embodiment. The hardware configuration of the product recommendation device 100 in this embodiment is the same as that in the first embodiment (see FIG. 1).

The product recommendation device 100 includes a sales history storage unit 111, a sales history acquisition unit 112, a customer recommendation index calculation unit 113, a provisional recommended product list generation unit 114, a recommended product list generation unit 115, a product information storage unit 116, and a product information output unit. In addition to 117, the product recommendation index calculation unit 118 that calculates an index (hereinafter referred to as a product recommendation index) indicating the degree of product recommendation for each product using the product sales history, and the product recommendation index are listed. A list of products recommended to all users based on the temporary product list generation unit 119 that generates the indicated list (hereinafter referred to as the temporary product list) and the product recommendation index shown in the temporary product list (hereinafter referred to as the temporary product list). It includes a product list generation unit 120 that generates (referred to as a product list).

The product recommendation index calculation unit 118 calculates the product recommendation index for each product based on the sales history acquired by the sales history acquisition unit 112. Details of the product recommendation index will be described later.

The temporary product list generation unit 119 generates a temporary product list by using the sales history acquired by the sales history acquisition unit 112 and the product recommendation index calculated by the product recommendation index calculation unit 118. In this provisional product list, various information including a product recommendation index is shown for each product included in the sales history acquired by the sales history acquisition unit 112. Details of the temporary product list will be described later.

The product list generation unit 120 generates a product list based on the product recommendation index shown in the provisional product list. Here, the product list generation unit 120 compares the product recommendation indexes of each product, and extracts the products whose product recommendation indexes satisfy the predetermined criteria as the products to be recommended to the consumer. Then, the product list generation unit 120 generates a product list by listing the extracted products. As a product whose product recommendation index satisfies a predetermined standard, for example, a certain number of products can be exemplified from those having a large product recommendation index. Further, for example, a product whose product recommendation index is equal to or higher than a predetermined threshold value can be exemplified.

[Explanation of product recommendation index]
Next, the product recommendation index will be described.
As described in the first embodiment, the products in the fourth quadrant region B4 shown in FIG. 4 are targeted for purchase by many consumers, but the total number of products purchased is smaller than that of the best-selling products. It can be said that it is a product that should be recommended to consumers (users).

Therefore, in the present embodiment, a value (product recommendation index) calculated from the following formula (6) is used as an index for specifying the region B4.

As shown in the formula (6), the product recommendation index sets the number of purchasers (= Y2), which is the value on the horizontal axis of the graph shown in FIG. 4, to the number of purchases per purchaser, which is the value on the vertical axis of the graph (= Y2). = X1 / Y2) to calculate.
In addition, in the graph shown in FIG. 4, the area where the value on the horizontal axis is large and the value on the vertical axis is small, that is, the product in the area B4 is the value of the product recommendation index as compared with the products in the other areas B1 to B3. Is likely to increase. On the other hand, in the region where the value on the horizontal axis is large and the value on the vertical axis is also large, that is, the product in the region B1 is likely to have a smaller value of the product recommendation index than the product in the region B4. Further, in the region where the value on the horizontal axis is small and the value on the vertical axis is small, that is, the product in the region B3 is likely to have a smaller value of the product recommendation index than the product in the region B4. Further, in the region where the value on the horizontal axis is small and the value on the vertical axis is large, that is, the product in the region B2 may have a smaller value of the product recommendation index than the products in the other regions B1, B3, and B4. high.

The product recommendation device 100 of the present embodiment identifies the product included in the area B4 based on the size of the product recommendation index, and determines the product to be recommended to all users. Further, the product recommendation device 100 of the present embodiment should be recommended to each user based on the size of the customer recommendation index from the products to be recommended to all users obtained in this way. Determine the product.

The temporary product list generated by the temporary product list generation unit 119 is the provisional product list (see FIG. 8) described in the first embodiment in which the "customer recommendation index" is replaced with the "product recommendation index". Become. Further, the product list generated by the product list generation unit 120 is obtained by replacing the "customer recommendation index" with the "product recommendation index" in the recommended product list (see FIG. 9) described in the first embodiment.

[Processing procedure to output the product list from the sales history]
Next, a processing procedure in which the product recommendation device 100 according to the present embodiment outputs a product list from the sales history will be described.
FIG. 13 is a flowchart showing an example of a processing procedure in which the product recommendation device 100 outputs a product list from the sales history in the second embodiment.

First, the sales history acquisition unit 112 acquires the sales history of the period to be analyzed (for example, for the latest one year) (step 201).

Next, the product recommendation index calculation unit 118 selects one product (i) for which the product recommendation index is to be obtained from the sales history (step 202). Subsequently, the product recommendation index calculation unit 118 counts the total number (X1) of all users purchasing the product (i) from the sales history (step 203), and the number of users purchasing the product (i) (Y2). ) Is counted (step 204). The order of steps 203 and 204 may be reversed.

Then, the product recommendation index calculation unit 118 calculates the product recommendation index for the product (i) by the above-mentioned formula (6) using the respective values acquired in step 203 and step 204 (step 205).

After the product recommendation index is calculated in step 205, the temporary product list generation unit 119 adds the information of the product (i) for which the product recommendation index is calculated to the temporary product list (step 206). Here, if the temporary product list has not been generated yet, the temporary product list generation unit 119 newly generates the temporary product list and adds the information of the first product (i). Then, the product recommendation index calculation unit 118 determines whether or not all the products have been selected in the sales history acquired in step 201 (step 207). If a negative determination (NO) is made in step 207, that is, if there is a product that has not yet been selected, the process proceeds to step 202 to continue the process related to the next product (i).

On the other hand, when a positive determination (YES) is made in step 207, the product list generation unit 120 generates a product list based on the temporary product list generated by the temporary product list generation unit 119 (step 208). Here, the product list generation unit 120 generates a product list by extracting products satisfying a predetermined standard based on the product recommendation index of each product shown in the provisional product list. More specifically, the product list generation unit 120 generates a product list by extracting products having a higher product recommendation index or extracting products having a product recommendation index equal to or higher than a threshold value. After that, the product list generation unit 120 outputs the generated product list to the customer recommendation index calculation unit 113 (step 209), and ends a series of processes.

After that, the customer recommendation index calculation unit 113, the provisional recommendation product list generation unit 114, the recommended product list generation unit 115, and the product information output unit 117 basically perform processing according to the processing procedure shown in FIG. However, in step 103, the customer recommendation index calculation unit 113 selects one product (i) from the products recorded in the product list generated by the product list generation unit 120. Further, in step 112, the customer recommendation index calculation unit 113 determines whether or not all the products have been selected in the product list, not in the sales history.

<Others>
In the first and second embodiments, the product recommendation device 100 is realized by one device, but it may be realized by dividing it into a plurality of devices. For example, various functional units including the customer recommendation index calculation unit 113 are realized by the client device, and various storage units including the sales history storage unit 111 and the product information storage unit 116 are realized by the server device connected to the client device via the network. It does not matter if it is realized.

100 ... Product recommendation device, 101 ... CPU, 102 ... Main memory, 103 ... Magnetic disk device, 104 ... Communication I / F, 105 ... Display mechanism, 106 ... Input device, 111 ... Sales history storage unit, 112 ... Sales history acquisition Department, 113 ... Customer recommendation index calculation unit, 114 ... Temporary recommended product list generation unit, 115 ... Recommended product list generation unit, 116 ... Product information storage unit, 117 ... Product information output unit, 118 ... Product recommendation index calculation unit, 119 … Temporary product list generation unit, 120… Product list generation unit

Claims (4)

On the computer
A function to acquire information on product sales history for a predetermined period, and
A function to select users to be recommended from multiple users,
A customer recommendation index, which is an index that increases as the number of users who have purchased the product increases and increases as the number of purchases by the users to be recommended decreases, is calculated for each product included in the acquired sales history. Function to do and
A program that realizes a function of extracting products having a high customer recommendation index from the acquired products included in the sales history and recommending them to the recommended user. The function of calculating, when each product is classified into one of grouping destinations, based on the number of purchase of one product grouping destinations classified as Ru goods, estimate the number of purchases of the one of the product claim 1 Symbol placement program characterized by. The function for calculating, the depending on the time of duration of the sales history information, according to claim 1 or claim 2, wherein the program and corrects the cart of goods. An acquisition department that acquires product sales history information for a predetermined period,
A selection unit that selects users to be recommended from multiple users,
A customer recommendation index, which is an index that increases as the number of users who have purchased the product increases and increases as the number of purchases by the users to be recommended decreases, is calculated for each product included in the acquired sales history. Calculation unit and
A product extraction system including a recommendation unit and a recommendation unit that extracts products having a high customer recommendation index from the acquired products included in the sales history and recommends them to the users to be recommended.

Images