JP5696025B2 - Product information recommendation device, method and program - Google Patents

Description

  The present invention relates to a product information recommendation device, method, and program, and more particularly, to a product information recommendation device, method, and program in a user's product purchase provided by an electronic commerce service.

  A number of information recommendation methods have been proposed. For example, in order to recommend a news article, the user is given a five-step evaluation for each news article, the past evaluation history is viewed, the degree of preference similarity between users is measured, and the news article preferred by a user with a similar preference Has been proposed (see, for example, Non-Patent Document 1).

  Also, let the user specify the text of the part of interest, estimate the user's interest based on the appearance frequency of the words contained therein, search the information in the community based on the word representing the user's interest, There is a method of presenting only information of a portion that is of interest to the user (for example, see Patent Document 1).

  In order to maximize user satisfaction, diversification of recommendation lists is effective. For example, when recommending movies, it is known that even if users are sci-fi enthusiasts, users are more satisfied with recommending mixed sci-fi, horror, and action than recommending only sci-fi. (For example, refer nonpatent literature 2).

JP 11-338869 A

Resnick, P. et al., "GroupLens: An open architecture for collaborative filtering of netnews," CSCW '94: Proceedings of the 1994 ACM conference on Computer supported cooperative work, ACM Press New York, 1994. Ziegler C.N. et al, "Improving recommendation lists through topic diversification," Proceedings of the 14th International Conference on World Wide Web, ACM Press New York, 2005

  In general, the products handled by the E-Commerce service are wide, and it is difficult for users to find products that they are really interested in.

  When a product is recommended using the method of Patent Literature 1 described above, a product related to an interesting word designated by the user is recommended. In this case, there is a problem that a word must be specified. Such a system that requires user interaction is inconvenient for the user.

  When a product is recommended by the method described in Non-Patent Document 1, a user's preference is estimated and a product associated with it is recommended, so that only similar products are recommended and the user's satisfaction is lowered. In addition, since it is highly likely that the recommended product is known to the user, it is difficult for the recommended product to lead to a new purchase.

  In addition, if a product is recommended by the method described in Non-Patent Document 2, the satisfaction level of the user is improved, but the product metadata (genre information) is used to calculate the similarity of the product. Not applicable to E-Commerce services for which no data is available.

  The present invention has been made in view of the above points. In the E-Commerce service, by creating a diversified recommendation list using only purchase information, a user can use the product in a situation where no metadata is prepared for the product. It is an object of the present invention to provide a product information recommendation device, method and program capable of making a highly satisfactory recommendation.

In order to solve the above problems, the present invention (Claim 1) is a product information recommendation device that uses user purchase information in an electronic commerce service,
Purchase information acquisition means for acquiring purchase information of each user ID in the electronic commerce service and storing it in a purchase information storage means;
Using the acquired purchase information, a recommendation score that is a probability of purchasing each product of each user is calculated, and a recommendation score calculation unit that is stored in a recommendation score storage unit;
A diversified score calculating unit that calculates a diversification score that is a probability that a user who has purchased a certain product does not purchase other products using the acquired purchase information, and stores the diversified score in a diversified score storage unit;
A recommendation list creating means for creating a recommendation list using the calculated recommendation score and the calculated diversification score;
I have a,
The diversification score calculation means includes
Refer to the purchase information storage means, using the user u, the product i purchased by the user, and the parameters for calculating the diversification score, the combination of products (i, i ′) (where i ′ is higher than i) The product diversification score P (n (i ′) | i) is calculated based on the number of users who have purchased the product i but have not purchased the product i ′ higher than the product i. And means for storing the combination (i, i ′) of the two products in the recommended score storage means .

Further, the present invention (Claim 2 ) provides the recommendation list creation means,
The acquired recommendation score for each item i from recommendation score storage means for said user u, the said diversification score storage means two items set viewed together (i, i ') the diversification score P (n About ( i ′ | i) ) is obtained, and means for determining products from the first place to the L place in order using the weight parameter is included.

  According to the present invention, in an E-Commerce service in which metadata is not attached to a product, recommendation with high user satisfaction can be achieved by diversifying the recommendation list using only purchase information of each user. .

It is a block diagram of the merchandise information recommendation apparatus in one embodiment of this invention. It is an example of the purchase information table in one embodiment of this invention. It is an example of the recommendation score table in one embodiment of this invention. It is an example of the diversification score table in one embodiment of this invention. It is a flowchart of the process at the time of the merchandise information update in one embodiment of this invention. It is a flowchart of the process at the time of the recommendation score update in one embodiment of this invention. It is a flowchart of the detailed process of step 250 of FIG. 6 in one embodiment of this invention. It is a flowchart of the detailed process of step 232 of FIG. 7 in one embodiment of this invention. It is a flowchart of the process at the time of the diversification score update in one embodiment of this invention. The book is a flowchart of processing at the time of user login in one embodiment.

  First, the outline of the present invention will be described.

  An object of the present invention is to provide a highly recommended recommendation list that is diversified using only purchase information in an E-Commerce service in which metadata is not assigned to a product. Specifically, the satisfaction degree of the user is defined as a probability that “when a recommendation list is presented, at least one of them is purchased”. That is, instead of simply arranging products in descending order of probability of purchase by the user as in the prior art, if the length of the recommended list to be provided is L, any one of the L products Create a recommendation list to maximize your purchase probability. The top one is the product with the highest probability of purchase by the user. The second-ranked product is not simply the product with the second highest probability of purchase by the user, but the probability of purchasing using the information that the first-ranked product is not purchased or not liked at the time of selecting the second-ranked item. calculate. In this way, the second-ranked product becomes a product most likely to be purchased by the user in a genre different from the first-ranked product, and when applied recursively, the recommendation list is diversified.

  The explanation will be made more strictly using mathematical expressions. A recommendation list to be presented to user u

とする。１位には、従来の協調フィルタリングと変わらず、

And First place is the same as conventional collaborative filtering,

とする。すなわち該ユーザの購買する確率の最も高い商品を配置する。２位には、従来の協調フィルタリングならば該ユーザの購買する確率が２番目に高いものを配置するが、本発明では、

And That is, the product with the highest probability of purchase by the user is placed. In the second place, the conventional collaborative filtering places the user having the second highest probability of purchase,

とする。ここでn(・)は該商品を購買しない、好きではないという情報を表す。本情報を考慮すれば、２位に配置する商品は１位に配置された商品とは異なるジャンルのものになりやすくなる。ｌ位も再帰的に、

And Here, n (•) represents information that the product is not purchased or is not liked. Considering this information, the product placed in the second place tends to be of a genre different from the product placed in the first place. The 1st place is also recursively,

のように決定する。

Decide like this.

  Next, a method for calculating the right side of Equation (3) will be described. Since it is difficult to obtain the right side directly, the probability P (i | u) that the user u purchases the item i and the probability that the user who does not like purchasing the item i does not purchase items 1 to 1

から求める。前者の確率は通常の推薦スコアであり、後者の確率は推薦リストの多様化に影響を与えるため多様化スコアと呼ぶことにする。

Ask from. The former probability is a normal recommendation score, and the latter probability affects the diversification of the recommendation list, so it is called a diversification score.

  There are several ways to calculate the right side of equation (3) using the recommendation term and the diversification score.

のように定義する方法がある。実施の形態では式(4)を用いて推薦リストを作成する場合について説明する。

There is a way to define as follows. In the embodiment, a case where a recommendation list is created using Expression (4) will be described.

  Specifically, to SF savvy users, the product most likely to be purchased in SF, the product most likely to be purchased in horror, the product most likely to be purchased in action, Will be recommended, and the probability of purchasing any one of the recommendation lists will be higher than the normal recommendation system (the user's satisfaction will be improved).

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

  FIG. 1 shows a configuration of a product recommendation device according to an embodiment of the present invention.

  A product recommendation device 1 shown in FIG. 1 includes a purchase information processing unit 10, a recommendation score calculation unit 20, a diversification score calculation unit 30, a recommendation list creation unit 40, a recording unit 50, a communication unit 60, and an input / output unit 70. The communication unit 60 is connected to the network 3, and the input / output unit 70 is connected to the external device 4 of the input device or the display device. The external terminal 2 corresponds to an EC site that cooperates with the present apparatus, for example.

  The storage unit 50 includes a purchase information table 51, a recommendation score table 52, and a diversification score table 53. Each table will be described below.

<Purchase information table 51>
As shown in FIG. 2, the purchase information table 51 includes a user ID field and a product ID field.

  In the user ID field, an identifier for identifying a user who has purchased the product is set by the purchase information processing unit 10.

  The product ID field corresponds to the identifier of the product that appears in the product information table, and is set by the purchase information processing unit 10.

<Recommended score table 52>
As shown in FIG. 3, the recommendation score table 52 includes a user ID field, a product ID field, and a recommendation score field.

  In the recommendation score field, the recommendation score P (i | u) of the user u for the product i is set by the recommendation score calculation unit 20.

<Diversification score table 53>
As shown in FIG. 4, the diversification score table 53 includes an upper product ID field, a product ID field, and a diversification score field. The diversification score field is literally a score used for diversification, but intuitively explained, the probability that a user who buys the product specified in the product ID field will not buy the product specified in the higher product ID field ( In this embodiment, it should be noted that a user who does not purchase the product specified in the higher-level product ID field does not have the probability of purchasing the product specified in the product ID field.

  In the upper product ID field, the diversification score calculation unit 30 sets the identifier of the product that appears in the product information table.

  In the diversification score field, the diversification score calculation unit 30 sets the probability P (n (i ′) | i) of the user who purchases the product i that he / she does not purchase the top product j.

<When purchasing products>
First, processing at the time of purchasing a product in an embodiment of the present invention will be described in detail. This process is executed when purchase information is sent from the linked EC site via the network 3 and the communication unit 70.

  FIG. 5 is a flowchart of processing at the time of purchasing a product in one embodiment of the present invention.

  Step 110) The purchase information processing unit 10 inserts in the purchase information table 51 a row in which values of the user ID field and the product ID field are set according to the purchased user and the purchased product.

<Recommended score update>
Next, the process at the time of the recommendation score update in one embodiment of this invention is demonstrated. This process may be performed, for example, when the system administrator uses the external device 4 to issue a recommendation score update command to the system, or may be set to be executed automatically every other day. May be.

  FIG. 6 is a flowchart of a process at the time of updating a recommendation score according to an embodiment of the present invention.

  Step 210) The recommendation score calculation unit 20 sets parameters Z, α, and β for calculating the recommendation score. These parameters are set by constants given in advance. However, Z is an integer of 1 or more, and α and β are decimal numbers of 0 or more.

  Step 220) The recommendation score calculation unit 20 refers to the purchase information table 51, and acquires the user u and the product i for each purchase.

  Step 230) The recommendation score calculation unit 20 estimates the topic z of each purchase using Collapsed Gibbs sampling described in Reference 1 (Bishop, M.C., “Pattern recognition and machine learning”, 2006).

  Step 240) The recommendation score calculation unit 20 uses the parameters Z, α, and β set in Step 210 and the topic z of each purchase estimated in Step 230 for each user u as shown in Equation (2). A recommendation score P (i | u) for each product i is calculated.

ここでM_zはトピックｚにアサインされた購買数、M_ziはトピックｚにアサインされ、かつ商品ｉに関する購買の数、M_uzはユーザｕの購買の中でトピックｚにアサインされたものの数、M_uはユーザuの購買数を表す。M_zとM_zi、M_uzはステップ２３０にて推定した各購買のトピックzから求めることができる。なお、Uはユーザの総数を表す。

Where M _z is the number of purchases assigned to topic z, M _zi is the number of purchases assigned to topic z and for item i, M _uz is the number of purchases of user u assigned to topic z, M _u represents the number of purchases of the user u. M _z , M _zi , and M _uz can be obtained from the topic z of each purchase estimated in step 230. U represents the total number of users.

  Step 250) The recommendation score calculation unit 20 obtains the recommendation score of the row where the user ID field value is u and the product ID field value is i in the recommendation score table 52 for each user u and each product i in step 240. Update to (i | u).

  Next, the processing of step 230 in FIG. 6 will be described in detail.

  FIG. 7 is a flowchart of detailed processing of step 230 in FIG. 6 according to the embodiment of the present invention.

Step 231) The recommendation score calculation unit 20 initializes each purchase topic z _m (m = 1,..., M) with an integer random number having a minimum value 1 and a maximum value Z. However, M represents the number of rows in the purchase information table 51, and m corresponds to the number of rows.

Step 232 is repeated a sufficient number of times (eg, 500 times). The number of repetitions may be an arbitrary number set in advance, or may be terminated when the topic z of each purchase converges (when most z _m ceases to change).

Step 232) The recommended score calculation unit 20 updates the topic z _m (m = 1,..., M) of each purchase.

  Next, the processing in step 232 in FIG. 7 will be described in detail.

  FIG. 8 is a flowchart of detailed processing in step 232 of FIG. 7 according to the embodiment of this invention.

  Steps 2321 and 2322 are performed for all purchases m = 1,.

Step 2321) The recommendation score calculation unit 20 calculates a sampling probability assigned to each topic candidate k = 1,..., Z of the topic z _m of purchase by the equation (6).

ここで＼mは該購買を除いたときの回数もしくは変数を表す。つまり、たとえば

Here, \ m represents the number of times or a variable when the purchase is excluded. In other words, for example

は更新前にz_mがｋにアサインされていれば

If z _m is assigned to k before update

を表し、そうでなければ

Represents otherwise

を表す。また、u_m，i_mはそれぞれm番目の購買のユーザと商品を表す。

Represents. Further, u _m, i _m each represent users and items of the m-th purchase.

Step 2322) recommendation score calculation unit 20, each topic candidate k = 1 topic z _m of該購purchase calculated in Step 2321, ..., based on the sampling probability of being assigned to the K, by generating a random number, z _m Update. Specifically, for k = 2,..., K, the sampling probability is updated according to Equation (7).

さらに、k=1，…，Kについて、サンプリングを式(8)のように更新する。

Further, for k = 1,..., K, the sampling is updated as in Expression (8).

その後、最小値０、最大値１の小数の乱数を発生させてその値をサンプリング確率が上回る最小のｋにアサインする。例を挙げて説明すると、今K=3として、式(3)に基づき算出されたサンプリング確率がそれぞれ

Thereafter, a decimal random number having a minimum value of 0 and a maximum value of 1 is generated, and that value is assigned to the minimum k that has a higher sampling probability. For example, assuming that K = 3, the sampling probabilities calculated based on Equation (3) are

であったとする。つまり、それぞれのトピックに４：１：５の割合でサンプリングされるようにしたい。このとき、式(7)と式(8)を用いて

Suppose that In other words, we want to sample each topic at a ratio of 4: 1: 5. At this time, using Equation (7) and Equation (8)

とサンプリング確率を更新し、最小値０、最大値１の小数の乱数を発生させてその値をサンプリング確率が上回る最小のkにアサインすればよい。

And the sampling probability is updated, a random number with a minimum value of 0 and a maximum value of 1 is generated, and that value is assigned to the minimum k that exceeds the sampling probability.

<When updating diversification score>
Next, the process at the time of the diversification score update in one embodiment of this invention is demonstrated. This processing may be performed, for example, when the system administrator uses the external device 4 to issue a diversification score update command to the device 1, or is set to be executed automatically every other day or the like. You may be able to do it.

  FIG. 9 is a flowchart of processing when the diversification score is updated according to the embodiment of this invention.

  Step 310) The diversification score calculation unit 30 sets a parameter δ for diversification score calculation. These parameters are set by constants given in advance. However, δ is a decimal number of 0 or more.

  Step 320) The diversification score calculation unit 30 refers to the purchase information table 51, and acquires the user u and the product i of each purchase.

  Step 330) Using the parameter δ set in Step 310 and the purchase information acquired in Step 320, the diversification score calculation unit 30 uses the parameter (δ) and the combination (i, i ′) of the two products as follows. The calculation score P (n (i ′) | i) is calculated.

ここでU_iは商品iを購買したことのあるユーザ数、U_i,n(i')は商品iを購買したことはあるが商品i'を購買したことはないユーザの数を表す。U_i、U_i,n(i')いずれも購買情報を集計することで簡単に求めることができる。

Here, U _i represents the number of users who have purchased the product i, and U _{i, n (i ′)} represents the number of users who have purchased the product i but have not purchased the product i ′. Both U _i and U _{i, n (i ′)} can be easily obtained by collecting purchase information.

  Step 340) The diversification score calculation unit 30 sets the diversification of the rows in the diversification score table 73 for each combination (i, i ′) of the two products where the value of the upper product ID field is i ′ and the value of the product ID field is i. The update score is updated to P (n (i ′) | i) obtained in step 330.

  Next, processing at the time of product recommendation in one embodiment of the invention will be described. This process may be started when a product recommendation request is received from the linked EC site via the network 3 and the communication unit 70, for example.

<When recommending products>
FIG. 10 is a flowchart of processing at the time of product recommendation in one embodiment of the present invention. The user who is the object of recommendation now is u. This user ID is sent from the linked EC site.

  Step 410) The recommendation list creation unit 40 refers to the recommendation score table 52 and obtains a recommendation score P (i | u) of each product i for the user u.

  Step 420) The recommendation list creation unit 40 refers to the diversification score table 53 and obtains a diversification score P (n (i ′) | i) for each combination (i, i ′) of two products.

  Step 430) The recommendation list creation unit 40 creates a recommendation list in order from the top using the acquired recommendation score and diversification score. For a specific recommendation list creation method, refer to the equations (1) to (4) and (5) described above. However,

は多様化スコアを使って以下のように求める。

Is obtained using the diversification score as follows.

ここでP(i)は全購買情報に対する商品iの購買の割合、Nはアイテムの種類数をそれぞれ表し、購買情報テーブル５１を参照することで簡単に求めることができる。

Here, P (i) represents the ratio of the purchase of the product i with respect to all purchase information, and N represents the number of types of items, which can be easily obtained by referring to the purchase information table 51.

  Step 440) The recommendation list creation unit 40 provides the created recommendation list to the user u. The created recommendation list is sent to the external terminal 2 providing the EC site via the communication unit 70 and the network 3.

  Note that the processing of each component of the product recommendation device can be constructed as a program and installed in a computer used as the product recommendation device for execution, or distributed via a network.

  In addition, the constructed program can be stored in a portable storage medium such as a hard disk, a flexible disk, or a CD-ROM, and can be installed or distributed in a computer.

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made within the scope of the claims.

1 Product recommendation device 2 External terminal 3 Network 4 External device 10 Purchase information processing unit 20 Recommended score calculation unit 30 Diversification score calculation unit 40 Recommendation list creation unit 50 Storage unit 51 Purchase information table 52 Recommendation score table 53 Diversification score table 60 Communication unit 70 Input / output unit

Claims (5)

In an electronic commerce service, a product information recommendation device that uses user purchase information,
Purchase information acquisition means for acquiring purchase information of each user ID in the electronic commerce service and storing it in a purchase information storage means;
Using the acquired purchase information, a recommendation score that is a probability of purchasing each product of each user is calculated, and a recommendation score calculation unit that is stored in a recommendation score storage unit;
A diversified score calculating unit that calculates a diversification score that is a probability that a user who has purchased a certain product does not purchase other products using the acquired purchase information, and stores the diversified score in a diversified score storage unit;
A recommendation list creating means for creating a recommendation list using the calculated recommendation score and the calculated diversification score;
Have
The diversification score calculation means includes
Refer to the purchase information storage means, using the user u, the product i purchased by the user, and the parameters for calculating the diversification score, the combination of products (i, i ′) (where i ′ is higher than i) The product diversification score P (n (i ′) | i) is calculated based on the number of users who have purchased the product i but have not purchased the product i ′ higher than the product i. And a product information recommendation device comprising means for storing a combination (i, i ′) of two products in the recommendation score storage means. The recommendation list creation means includes:
A recommendation score of each product i for the user u is acquired from the recommendation score storage means, and the diversification score P (n (i ′) for the combination (i, i ′) of the two products from the diversification score storage means. 2. The product information recommendation device according to claim 1, further comprising means for obtaining i)) and determining products from the first place to the L place in order using the weight parameter. In an electronic commerce service, a product information recommendation method in a device that uses user purchase information,
Purchasing information acquisition means of the device acquires purchasing information of each user ID in the electronic commerce service, and stores the purchasing information in the purchasing information storage means of the device ; and
A recommendation score calculating unit of the device calculates a recommendation score that is a probability of purchasing each product of each user using the acquired purchase information, and stores the recommendation score in a recommendation score storage unit of the device ; ,
The diversification score calculation means of the device uses the acquired purchase information to calculate a diversification score that is a probability that a user who has purchased a product will not purchase another product , and diversifies the device A diversification score calculation step stored in the score storage means;
A recommendation list creating means for creating a recommendation list using the calculated recommendation score and the calculated diversification score;
And
In the diversification score calculation step,
Refer to the purchase information storage means, using the user u, the product i purchased by the user, and the parameters for calculating the diversification score, the combination of products (i, i ′) (where i ′ is higher than i) The product diversification score P (n (i ′) | i) is calculated based on the number of users who have purchased the product i but have not purchased the product i ′ higher than the product i. A method for recommending product information. In the recommendation list creation step,
A recommendation score of each product i for the user u is acquired from the recommendation score storage means, and the diversification score P (n (i ′) for the combination (i, i ′) of the two products from the diversification score storage means. | i)) is acquired, and the product information recommendation method according to claim 3, wherein the products from the first place to the L place are determined in order using the weight parameter. Computer
The product information recommendation program for functioning as each means of the product information recommendation device according to claim 1.

Images