KR20140056731A - Purchase recommendation service system and method - Google Patents

Purchase recommendation service system and method Download PDF

Info

Publication number
KR20140056731A
KR20140056731A KR1020120122108A KR20120122108A KR20140056731A KR 20140056731 A KR20140056731 A KR 20140056731A KR 1020120122108 A KR1020120122108 A KR 1020120122108A KR 20120122108 A KR20120122108 A KR 20120122108A KR 20140056731 A KR20140056731 A KR 20140056731A
Authority
KR
South Korea
Prior art keywords
purchase
period
unit
user
similarity
Prior art date
Application number
KR1020120122108A
Other languages
Korean (ko)
Inventor
김민성
Original Assignee
에스케이플래닛 주식회사
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 에스케이플래닛 주식회사 filed Critical 에스케이플래닛 주식회사
Priority to KR1020120122108A priority Critical patent/KR20140056731A/en
Publication of KR20140056731A publication Critical patent/KR20140056731A/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0631—Item recommendations

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Data Mining & Analysis (AREA)
  • Development Economics (AREA)
  • General Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Mathematical Physics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • General Engineering & Computer Science (AREA)
  • Algebra (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Game Theory and Decision Science (AREA)
  • Pure & Applied Mathematics (AREA)
  • Software Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The present invention relates to a purchase recommendation system and method, and more particularly, to a purchase history database in which purchase histories including at least two users and purchases of respective users and purchase dates are stored. A purchase history extracting device for each period of dividing a total purchased purchase period arbitrarily into N unit purchase periods and extracting purchase histories in each unit purchase period; And calculating the average value in the entire purchase period by calculating the degree of similarity between users or the preference degree of each user for each of the N unit purchase periods, calculating a product preference for each user according to the calculated average value, And a collaboration filtering processing device for generating recommendation information. Accordingly, by dividing the purchasing history of the users into sub-periods and generating the recommendation information by reflecting the recent purchasing history and the old purchasing history relatively different in weight, it is possible to generate recommendation information that is sensitive to the trend, To a purchase recommendation system and method capable of providing information.

Description

[0001] PURCHASE RECOMMENDATION SERVICE SYSTEM AND METHOD [0002]

The present invention relates to a purchase recommendation system and method, and more particularly, to a system and method for purchasing recommendation information by dividing purchase histories of users into sub-periods and generating recommendation information by reflecting recent purchasing histories and relatively- The present invention relates to a purchase recommendation system and method capable of providing reliable recommendation information by reflecting past purchasing history while being sensitive to fashion.

Due to the development of large-capacity data transmission and processing technology, the amount of content that users can access is increasing exponentially. In particular, with the introduction of high-speed communication networks and the increasing capacity of various multimedia devices such as smart phones, tablets, netbooks, and IP TVs, users can enjoy a lot of contents regardless of time and place.

However, as the amount of content increases, the time and effort required for the user to find the desired content increases. In order to solve this problem, a method of selecting and recommending contents satisfying the user has appeared.

Such a conventional recommendation method is a collaborative filtering technique. Collaborative filtering is a technology that recommends items that are likely to be purchased for each user by using purchase records of multiple users. This collaborative filtering extracts a recommendation list by using the similarity between users, and especially calculates a weighted value for a customer similar to the user.

According to the conventional recommendation method, the entire purchasing history for a certain period is used in calculating the similarity between users based on the purchase history. However, if the period of collecting purchase histories is changed, the user's purchase history is different, so that similarity between users is calculated differently, and different recommendation results are generated depending on the period of collecting purchase histories.

In addition, in the case of the conventional recommendation method, the entire purchasing history for the set period is used, but the difference in purchase time is not reflected. For example, the similarity between users who bought A, B, and C 10 months ago and users who bought A, B, and C a week ago is calculated to be the same.

Accordingly, there is a problem that completely different values can be recommended to each user depending on the definition of the purchase period. In addition, when the similarity is calculated in the same way without considering the purchase timing of the purchasing history of products with high sensitivity to fashion, such as an app or a popular song, there is a problem that the reliability of the recommendation information may deteriorate.

Korean Patent Laid-Open No. 10-2011-0043369; Relationship analysis method for music recommendation

Disclosure of Invention Technical Problem [8] Accordingly, the present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to provide a method and apparatus for generating recommendation information by dividing purchasing history of users into detailed periods, The present invention also provides a purchase recommendation system and method that can provide reliable recommendation information while reflecting past purchasing history.

According to an aspect of the present invention, there is provided a purchase history database storing purchase histories including at least two users and purchases of each user and purchase dates; A purchase history extracting device for each period of dividing a total purchased purchase period arbitrarily into N unit purchase periods and extracting purchase histories in each unit purchase period; And calculating the average value in the entire purchase period by calculating the degree of similarity between users or the preference degree of each user for each of the N unit purchase periods, calculating a product preference for each user according to the calculated average value, There is provided a purchase recommendation system including a collaboration filtering processing device for generating recommendation information.

Here, the collaborative filtering processing apparatus may calculate a mean value in the entire purchase period by assigning a relatively high weight to the similarity of users in the N unit purchase periods or the product preference of each user in the latest purchase period .

According to another aspect of the present invention for solving the above-mentioned problems, there is provided a method for calculating a similarity degree per unit purchase period by using a purchase history of each unit purchase period extracted by dividing an entire purchase period into N unit purchase periods, Output module; A user similarity average calculating module for calculating a similarity average in the entire purchasing period by summing the similarities for each unit purchase period; And a collaborative filtering module for collaboratively filtering the user similarity average and purchasing history in the entire purchasing period to generate recommendation information in descending order of product preference for each user.

The apparatus may further include a period-specific weight setting module for assigning a relatively high weight to the similarity of users in the N unit purchase periods to the latest purchase period to generate a similarity for each unit purchase period reflecting the weight.

The user similarity calculation module for each period can calculate the purchase history for each unit purchase period by the jacquard similarity calculation method according to [Equation 1].

[Equation 1]

Figure pat00001

Ji (A, B) is the jacquard similarity between user A and user B in period pi, and A (p, i) It means a set of goods purchased in pi.

The user similarity average calculating module may calculate the similarity average in the entire purchase period by summing the similarities of the unit purchase periods according to Equation (2).

&Quot; (2) "

Figure pat00002

In addition, the user similarity calculation module for each period can calculate the purchase history for each unit purchase period by the calculation method of the cosine similarity according to [Equation 3].

&Quot; (3) "

Figure pat00003

Here, p1, p2, ..., pN are the unit purchasing periods, Cospi (A (Pi), B (pi)) is the cosine similarity between user A and user B in period pi,

Figure pat00004
Denotes the dot product of the vector.

The user similarity average calculation module may calculate the similarity average in the entire purchase period by summing the similarities of the unit purchase periods according to Equation (4).

&Quot; (4) "

Figure pat00005

According to another aspect of the present invention for solving the above problems, there is provided a service providing method of a service providing system, comprising: a user who is a user who calculates a degree of similarity per unit purchase period by using the purchase history of each unit purchase period extracted by dividing the entire purchase period into N unit purchase periods; A degree of similarity calculation module; A collaborative filtering module for each period for collectively filtering a similarity and a purchasing history in the unit purchase period to calculate a product preference per user in each unit purchase period; And a collaborative filtering average calculation module for calculating a preference average in the entire purchase period by summing product preferences of each user in each unit purchase period and generating recommendation information in order of product preference per user according to the calculated average value, And a cooperative filtering processing unit.

The apparatus may further include a period weight setting module for assigning a relatively high weight to the product preference of each user in the N unit purchase periods to generate a product preference for each user per unit purchase period in which the weight is reflected, .

The collaborative filtering average calculation module may calculate the average of preferences in the entire purchase period by summing the per-user product preferences in each unit purchase period according to Equation (7).

&Quot; (7) "

Figure pat00006

Here, p1, p2, ..., pN are the unit purchasing periods, Pot i is the commodity preference in the period i, Ai is the purchase history of the unit purchasing period, and Si is the user similarity matrix do.

According to another aspect of the present invention for solving the above-mentioned problems, (a) dividing a total purchasing period arbitrarily set into N unit purchasing periods, and calculating a purchasing item and a purchase date of each user in each unit purchasing period Extracting an included purchase history; (b) calculating a degree of similarity for each unit purchasing period using the purchasing history in each unit purchasing period; (c) calculating a similarity average in the entire purchasing period by summing the similarities of the unit purchasing periods; And (d) collaboratively filtering the user similarity average and purchase history in the entire purchase period to generate recommendation information in the order of higher product preference per user.

The step (b) may include calculating the similarity for each unit purchase period by processing the purchase history in each unit purchase period using a jacquard similarity calculating method or a cosine similarity calculating method.

The method may further include, during the steps (b) and (c), assigning relatively high weights to the similarity degree of each of the N unit purchase periods as the latest purchase period.

According to another aspect of the present invention for solving the above problems, there is provided a method for purchasing a product, comprising: dividing an entirely purchased purchase period into N unit purchase periods, Extracting a history; Calculating a degree of similarity for each unit purchase period using purchase histories in each unit purchase period; Calculating a similarity average in the entire purchasing period by summing the similarities of the unit purchasing periods; And a step of collaboratively filtering the user similarity average and the purchase history in the entire purchase period to generate recommendation information in the order of higher product preference per user is recorded as a program and recorded on a recording medium readable by an electronic device Is provided.

According to another aspect of the present invention for solving the above-mentioned problems, (a) dividing a total purchasing period arbitrarily set into N unit purchasing periods, and calculating a purchasing item and a purchase date of each user in each unit purchasing period Extracting an included purchase history; (b) performing a collaborative filtering process on the similarity and the purchase history in the unit purchase period to calculate product preference per user in each unit purchase period; (c) calculating a preference average in the entire purchase period by summing product preferences of each user in each unit purchase period; And (d) generating recommendation information in descending order of product preference per user according to the calculated average value.

Between step (b) and step (c), a relatively high weight may be added to the product preference of each user in the N unit purchase periods as the latest purchase period is performed.

According to another aspect of the present invention for solving the above problems, there is provided a method for purchasing a product, comprising: dividing an entirely purchased purchase period into N unit purchase periods, Extracting a history; Performing collaborative filtering processing of the degree of similarity and the purchase history in the unit purchase period to calculate product preference per user in each unit purchase period; Calculating a preference average in the entire purchase period by summing product preferences of each user in each unit purchase period; And a step of generating recommendation information in the order of higher product preference per user according to the calculated average value, and a recording medium readable by the electronic apparatus is provided.

As described above, the purchase recommendation system and method according to the present invention generate recommendation information by dividing purchase histories of users into sub-periods and reflect purchase histories of recent purchases and purchases that are relatively old, It is possible to provide reliable recommendation information by reflecting past purchasing history.

In addition, the purchase recommendation system and method of the present invention recommends a product based on the degree of similarity among users whose preferences are clearly evident by reflecting the similarity of purchase histories for each period. Therefore, regardless of the purchase period or time, Can be accessed by subdividing the relationships between users with similarity

In addition, the purchase recommendation system and method of the present invention can generate recommendation information that is sensitive to recent fashion while reflecting the past purchasing history according to the life cycle of a pandemic product or a product by assigning a weight for each period.

1 is a configuration diagram of a purchase recommendation system according to an embodiment of the present invention;
2 is a control block diagram of a collaborative filtering processing apparatus according to an embodiment of the present invention,
3 is a control block diagram of a cooperative filtering processing apparatus according to another embodiment of the present invention,
4 is a flowchart of a purchase recommendation method according to an embodiment of the present invention;
5 is a flowchart of a purchase recommendation method according to another embodiment of the present invention;
FIG. 6 through FIG. 10 illustrate processing data of the purchase recommendation system according to the embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description with reference to the accompanying drawings, the same or corresponding components will be denoted by the same reference numerals, and redundant description thereof will be omitted.

1 is a configuration diagram of a purchase recommendation system according to an embodiment of the present invention.

As shown in FIG. 1, the purchase recommendation system of the present invention includes a purchase history database 100, a purchase history extracting device 200 for each period, and a collaboration filtering processing device 300.

The purchase history database 100 stores a purchase history including a product purchased by each user and a purchase date.

The purchase history extracting apparatus 200 for each period divides the entire purchase period into N pieces and extracts the user purchase history of each purchase period. In the case of the existing method, since the total purchase history of the last one to three months is used, the customer's preference reflected in the previous data can not be used at all. However, the purchase history extracting apparatus 200 according to the present invention, History can be segmented into a plurality of unit purchase periods and extracted.

Here, the entire purchasing period does not necessarily mean accumulated cumulative purchase history data. For example, if you look at the last 12 months of purchase history, 12 months is the total purchase period. Therefore, the total purchase period can be set to the last 5 months, or the last 12 months depending on the characteristics of the commodity.

The purchase history extracting apparatus 200 for each period can divide the entire purchase period into N unit purchase periods. The method of dividing by N unit purchase periods can be set differently according to the application subject, such as the number of divisions and the size of the unit purchase period.

For example, the entire purchase period can be evenly divided. If the total purchase period is one year, N = 4 can be divided equally by three months, or divided into dates corresponding to spring, summer, autumn and winter. In addition, it is possible to divide the unit purchasing period arbitrarily by subdividing the conditions by dividing the total purchase period into the latest six months, dividing by month or week, or starting the Monday or starting the Sunday if the unit is week. It is also possible to divide it into unequal cycles in consideration of the seasonal characteristics of the service, the purchase history of the promotional user, and the like. For example, it is also possible to divide an arbitrary date in which the service is changed into unit purchase periods.

Since the unit purchase period becomes shorter as N becomes larger, the number of users who purchase the same product at the same time decreases, and the similarity between the users as a whole may be reduced. However, when extracting the recommendation information, So that there is no problem. However, if N is too large, the unit purchasing period becomes too short, and when the similarity becomes close to zero, recommendation information may not be extracted. Therefore, if the average number of purchases of products is small, For a product having a large average number of purchases, the total purchase period can be set relatively short and N can be set to a large value.

The collaborative filtering processing apparatus 300 recommends items that are likely to be purchased for each user according to the similarity histories of purchase histories of users for N periods. The collaborative filtering processing unit 300 extracts recommendation information based on the purchase history similarity for each unit purchase period. Here, considering the purchased product as well as the purchase time, Can be calculated to have a higher degree of similarity between them.

According to such a configuration, the present invention specifies the purchase period as X months, subdividing the purchasing history into N pieces, and reflects not only the purchased products but also the same product at the time of purchase, thereby calculating the degree of similarity between users, can do.

Hereinafter, a method of generating recommendation information by calculating the degree of similarity between purchase histories of users for each period in which the collaboration filtering processing apparatus 300 is divided into N is as follows.

FIG. 2 is a control block diagram of an apparatus 300 for processing collaborative filtering according to an embodiment of the present invention. The similarity degree between users is calculated for N purchase periods, and the similarity average for the entire purchase period is calculated As shown in FIG.

The collaborative filtering processing apparatus 300 according to an embodiment of the present invention includes a user similarity calculation module 310 for each period, a user similarity average calculation module 312, a weighted weight setting module 316, and a collaboration filtering module 314 .

The user similarity calculation module 310 for each period calculates the similarity between users for the purchase histories of N unit purchase periods. Here, the Jaccard similarity calculation method and the cosine similarity calculation method can be applied as the similarity calculation method between users using the purchase history.

The method of calculating purchase history for each unit purchase period by the jacquard similarity calculating method is as follows. First, when the unit purchasing period is set to p1, p2, ..., pN, the jacquard similarity degree between the unit purchasing periods of the two users A and B is expressed by Equation (1).

Figure pat00007

Here, A (p, i) denotes a set of goods purchased by the user A in the period pi.

The user similarity average calculating module 312 calculates the similarity average for the entire purchase period using the similarity of N unit purchase periods.

Thus, if the jacquard similarity degree for each unit purchase period is obtained according to Equation (1), the jacquard similarity degree Jp (A, B) for the entire purchase period reflecting the purchase history similarity level for each unit purchase period is calculated through the following Equation (2) .

Figure pat00008

On the other hand, a method of calculating purchase histories for each unit purchase period by the calculation method of the cosine similarity is as follows. First, when the unit purchasing period is set to p1, p2, ..., pN, the cosine value of the unit purchasing period of the two users A and B is expressed by the following Equation (3).

Figure pat00009

here,

Figure pat00010
Denotes the dot product of the vector.

Accordingly, if the cosine similarity for each unit purchase period is obtained according to Equation (3), the cosine similarity Cosp (A, B) for the entire purchase period reflecting the purchase history similarity for each unit purchase period is calculated through the following Equation (4) .

Figure pat00011

Here, the jacquard similarity calculated by [Equation 2] and the cosine similarity calculated by [Equation 4] are the same weighted values applied to each unit purchase period.

However, it can be judged that the purchasing history of the most recent period mainly expresses the user's preference more strongly depending on the life cycle of the commodity being released and consumed and the next commodity being released, or the service change. Accordingly, the weight setting module 316 for each unit purchase period can assign a higher weight to the similarity of purchase histories in the period close to the current point of time to calculate the similarity.

The period weight setting module 316 sets the total sum of the weights to be 1 in consideration of the ratio of the unit purchasing period to the total purchase period to calculate the similarity for the entire period.

In the case of calculating the jacquard similarity by setting the weight, the applied equation is as shown in the following equation (5).

Figure pat00012

Figure pat00013

When the weight is set to calculate the cosine similarity, the applied equation is as shown in Equation (6) below.

Figure pat00014

Figure pat00015

The collaborative filtering module 314 calculates preference for each product according to the jacquard similarity degree for the entire purchase period calculated by the periodic user similarity calculation module 310 to extract recommendation information.

FIG. 3 is a control block diagram of a collaborative filtering processing apparatus 300 according to another embodiment of the present invention. In FIG. 3, the degree of similarity is calculated for N purchase periods, and the degree of preference in the corresponding period is calculated by performing collaborative filtering. And calculating a preference average for the entire purchase period.

The collaborative filtering processing apparatus 300 according to an embodiment of the present invention includes a user similarity calculation module 310 for each period, a collaboration filtering module 320 for each period, a weight setting module 316 for each period, a collaboration filtering average calculation module 322, .

The user similarity calculation module 310 for each period calculates the similarity between users for the purchase histories of N unit purchase periods. Here, as the method of calculating the degree of similarity between users using the purchase history, it is possible to calculate the jacquard similarity or to calculate the cosine similarity. The jacquard similarity calculation method calculates the number of items purchased by users A and B at the same time when the user A and the user B purchase (1) a specific product (0) Divided by the number. That is, a value obtained by dividing the number of intersections of products purchased by users A and B by the number of unions is calculated as J (user A, user B) between the user A and the user B.

The period-by-period collaborative filtering module 320 performs collaborative filtering using the degrees of similarity calculated for each of the N unit purchase periods. When collaborative filtering is performed independently using one purchase history of one user and K users for N unit purchase periods, the preference score for the same product is calculated for one user in different periods. That is, if N is equal to 5, if the user A in the first period of time for the user A, or a user whose degree of similarity to the user A is not equal to the user A in the same period, purchases the product a, the potential matrix of A includes A non-zero score is calculated. Likewise, if there is a user who has purchased the product a and the similarity to the user A is not 0 in another period, a preference score for the product a is generated in the user A's preference matrix. Accordingly, the period-by-period collaborative filtering module 320 generates a potential matrix for each user in each unit purchase period.

The collaborative filtering average calculation module 322 calculates an average of the similarity calculated for each of the N unit purchase periods. The collaborative filtering average calculation module 322 may calculate the average preference matrix by summing up all the user's preference matrices in each unit purchase period generated by the periodical collaboration filtering module 320 and then dividing it into N pieces.

If the purchase history in the unit purchase period is Ai and the similarity matrix between users in the unit purchase period is Si, the preference matrix of the unit purchase period is calculated as Si * Ai. Therefore, the average preference matrix is calculated by adding each matrix to the preference matrix of the unit purchase period and then dividing by the number N of the total unit purchase periods. This can be expressed by Equation (7).

Figure pat00016

The period weight setting module 316 sets the total sum of the weights to be 1 in consideration of the ratio of the unit purchasing period to the total purchase period to calculate the similarity for the entire period.

In the case of calculating the average preference matrix by setting the weights, the applied formula is as shown in [Equation 8].

Figure pat00017

Figure pat00018

On the other hand, the weight per unit purchasing period can take the following form to satisfy the condition that all the weights are equal to one.

In the first embodiment, the weight Wi can be set so that the sum of the weights for all N unit purchase periods is 1. At this time,

Figure pat00019
. ≪ / RTI >

As a second embodiment, the weight Wi can be expressed as an exponential decay function according to a time difference between a current time and a past period, that is, a start date (end date, intermediate day) Can be used. At this time, the weight Wi

Figure pat00020
. ≪ / RTI > Here, (t-ti) represents the difference between the present day and the specific unit purchasing period, and the unit of time may be set to minute, hour, day, month, etc. according to the characteristics of the target commodity.

As a third embodiment, the weight Wi may be indexed as 1,2,3 ...... n for each unit purchase period to define a weight for each unit purchase period. At this time,

Figure pat00021
. ≪ / RTI > Here, the smaller the value of the index, the reciprocal of the index value of the number of periods n can be taken to have a higher weighting value.

FIG. 4 is a flowchart of a purchase recommendation method according to an embodiment of the present invention. Referring to FIG. 4, there is shown a method of generating recommendation information using the apparatus 300 for collaborative filtering according to an embodiment of the present invention will be.

In order to generate the recommendation information, the purchase history extracting apparatus 200 for each period divides the entire purchase period into N unit purchase periods (S110), and extracts the user purchase history of each unit purchase period (S120). Here, the total purchasing period can be set to the latest 5 months, the last 12 months, etc. according to the characteristics of the commodity. The method of dividing the unit purchasing period into N unit purchasing periods can be divided into a number of divisions, Can be set. FIG. 6 illustrates purchase history data for each unit purchase period. Purchase history data for each unit purchase period shown in FIG. 6 is purchase history data of users # 1, # 2, and # 3, and unit purchase periods are set to five (periods # 1 to # 5). According to the purchase history data, both the user # 1 and the user # 2 purchased the article # 2 in the period # 1. It can be seen that in the case of purchased item # 3, user # 1 has purchased in period # 3, and that user # 2 and user # 3 have purchased in period # 4.

When the user purchase histories of the unit purchasing periods are extracted, the user similarity calculating module 310 for each period calculates the similarities between users for the purchasing histories of N unit purchasing periods (S130). When the purchasing history of FIG. 6 is processed according to the Jacquard degree of similarity calculation method by applying equations (1) and (2), the degree of similarity W_p between users is calculated as shown in FIG.

Thereafter, by performing collaborative filtering using the calculated inter-user similarity, a preference matrix of goods is generated (S150). The preference matrix can be calculated by multiplying the user similarity (W_p) shown in FIG. 7 by the purchase history matrix for the entire purchase period.

Thereafter, the recommendation information is generated and provided in the order of high preference on the preference matrix (S160).

FIG. 5 is a flowchart of a recommendation method according to another embodiment of the present invention. FIG. 5 illustrates a method of generating recommendation information using the collaboration filtering processing apparatus 300 according to another embodiment of the present invention illustrated in FIG. 3 .

In order to generate recommendation information, the purchase history extracting apparatus 200 for each period divides the entire purchase period into N unit purchase periods (S210), and extracts the user purchase history of each unit purchase period (S220). Purchase history Ai of each user in the specific unit purchase period can be displayed as shown in FIG.

Thereafter, the user similarity degree calculation module 310 for each period calculates the similarity degree Si between users for the purchase histories of N unit purchase periods (S230). Here, as the method of calculating the degree of similarity between users using the purchase history, it is possible to calculate the jacquard similarity or to calculate the cosine similarity. Thus, the similarity Si calculated by the Jacquard method with respect to the purchase history Ai in FIG. 8 can be calculated as shown in FIG. The similarity (Si) calculated by the jacquard similarity calculation method is a value obtained by dividing the number of products purchased simultaneously by the users A and B during the i-unit purchase period by the number of all the products purchased by the users A and B.

When the degree of similarity of the unit purchase period is calculated, the preference Pi of each item is calculated (S240). The preference Pi of the product in the unit purchase period can be calculated by multiplying the purchase history Ai in the unit purchase period i by the similarity Si. Thus, the preference matrix W * A can be calculated as shown in Fig. 10 by multiplying the purchase history Ai in Fig. 8 by the similarity Si in Fig.

Thereafter, an average of preferences of each unit purchasing period is calculated (S250). The collaborative filtering average calculation module 322 calculates the average preference matrix by summing the matrices of the matrices of the respective unit purchase periods and dividing the matrices by N, which is the total number of unit purchase periods, according to [Equation 7] .

When the average preference matrix is calculated, recommendation information is generated in the order of products having a high preference value (S260).

According to another aspect of the present invention, there is provided a method comprising: dividing an arbitrarily set entire purchasing period into N unit purchasing periods, and extracting a purchasing history including a user and purchases of each user and a purchase date in each unit purchasing period; Calculating a degree of similarity for each unit purchase period using purchase histories in each unit purchase period; Calculating a similarity average in the entire purchasing period by summing the similarities of the unit purchasing periods; And a step of collaboratively filtering the user similarity average and the purchase history in the entire purchase period to generate recommendation information in the order of higher product preference per user is recorded as a program and recorded on a recording medium readable by an electronic device Lt; / RTI >

Such purchase recommendation methods can be written in a program, and the codes and code segments constituting the program can be easily deduced by a programmer in the field. In addition, the program relating to the purchase recommendation method may be stored in an information storage medium (Readable Media) readable by the electronic apparatus, and read and executed by the electronic apparatus.

According to still another aspect of the present invention, there is provided a method of purchasing a product, the method comprising: dividing an arbitrarily set entire purchasing period into N unit purchasing periods, and extracting purchase histories including purchasers and purchase dates of users and users in each unit purchasing period; Performing collaborative filtering processing of the degree of similarity and the purchase history in the unit purchase period to calculate product preference per user in each unit purchase period; Calculating a preference average in the entire purchase period by summing product preferences of each user in each unit purchase period; And a step of generating recommendation information in a descending order of product preference per user according to the calculated average value, can be recorded by a program and recorded in a recordable medium readable by an electronic apparatus.

Such purchase recommendation methods can be written in a program, and the codes and code segments constituting the program can be easily deduced by a programmer in the field. In addition, the program relating to the purchase recommendation method may be stored in an information storage medium (Readable Media) readable by the electronic apparatus, and read and executed by the electronic apparatus.

Thus, those skilled in the art will appreciate that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. It is therefore to be understood that the embodiments described above are to be considered in all respects only as illustrative and not restrictive. The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

The present invention divides purchase histories of users into sub-periods and generates recommendation information by reflecting recent purchasing histories and relatively old purchasing histories to different ratios, Lt; RTI ID = 0.0 > and / or < / RTI > providing information.

100: Purchase history database
200: Purchase history extracting device for each period
300: Collaborative filtering processing device
310: user similarity calculation module for each period
312: user similarity average calculation module
314: Collaborative Filtering Module
316: Period weight setting module
320: Collaboration Filtering Module by Period
322: Collaborative filtering average calculation module

Claims (18)

A purchase history database storing purchase histories including at least two users and purchases of each user and purchase dates;
A purchase history extracting device for each period of dividing a total purchased purchase period arbitrarily into N unit purchase periods and extracting purchase histories in each unit purchase period; And
Calculating the average value in the entire purchase period by calculating the similarity degree among users or the product preference degree per user for the N unit purchase periods, calculating the product preference per user according to the calculated average value, And a collaborative filtering processing device for generating information. The method according to claim 1,
Wherein the collaborative filtering processing apparatus comprises:
Wherein the average value in the entire purchase period is calculated by assigning a relatively high weight to the similarity of the users in the N unit purchase periods or the product preference of each user in the latest purchase period. A user similarity degree calculation module for each period in which the total purchase period is divided into N unit purchase periods and the similarity for each unit purchase period is calculated using the purchase histories of the extracted unit purchase periods;
A user similarity average calculating module for calculating a similarity average in the entire purchasing period by summing the similarities for each unit purchase period; And
And a collaborative filtering module that performs collaborative filtering processing of an average of user similarities in the entire purchase period and purchasing history to generate recommendation information in descending order of product preference per user. The method of claim 3,
Further comprising a period weight setting module for assigning a relatively high weight to the similarity of the users of the N unit purchase periods to the latest purchase period to generate a similarity for each unit purchase period in which weights are reflected, Device. The method of claim 3,
The user similarity degree calculation module for each period includes:
Wherein the purchasing history of each unit purchase period is calculated by a Jacquard degree-of-similarity calculating method according to Equation (1).
[Equation 1]
Figure pat00022

Jpi (A, B) is the jacquard similarity between user A and user B in period pi, and A (p, i) is the purchase order of user A in period pi. Means a set of goods. 6. The method of claim 5,
Wherein the user similarity average calculation module comprises:
And calculates a similarity average in the entire purchase period by summing the similarities of the unit purchase periods according to Equation (2).
&Quot; (2) "
Figure pat00023
 The method of claim 3,
The user similarity degree calculation module for each period includes:
And a purchase history for each unit purchase period is calculated by a calculation method of cosine similarity according to the following equation (3).
&Quot; (3) "
Figure pat00024

In this case, p1, p2, ..., pN are unit purchasing periods, and Cospi (A (Pi), B (pi)) is the cosine similarity between user A and user B in period pi,
Figure pat00025
Denotes the dot product of the vector. 8. The method of claim 7,
Wherein the user similarity average calculation module comprises:
And calculates a similarity average in the entire purchase period by summing the similarities of the unit purchase periods according to Equation (4).
&Quot; (4) "
Figure pat00026
 A user similarity degree calculation module for each period in which the total purchase period is divided into N unit purchase periods and the similarity for each unit purchase period is calculated using the purchase histories of the extracted unit purchase periods;
A collaborative filtering module for each period for collectively filtering a similarity and a purchasing history in the unit purchase period to calculate a product preference per user in each unit purchase period;
Calculating a preference average in the entire purchase period by summing product preferences of each user in each unit purchase period and generating recommendation information in order of product preference per user according to the calculated average value; Wherein the cooperative filtering processing unit comprises: 10. The method of claim 9,
And a weight setting module for each of the N pieces of unit purchasing periods to generate a user preference for each user for each unit purchase period in which the weight is reflected by applying a relatively high weight to the product preference for each user in the latest purchase period Collaborative filtering processing unit. 10. The method of claim 9,
Wherein the collaborative filtering average calculation module comprises:
And calculates a preference average in the entire purchase period by summing product preferences of each user in each unit purchase period according to Equation (7).
&Quot; (7) "
Figure pat00027

Here, p1, p2, ... pN denotes a unit purchase period, Pot i denotes a commodity preference in a period i, Ai denotes a purchasing history of a unit purchasing period, and Si denotes a similarity matrix between users in a unit purchasing period. (a) dividing an entirely purchased purchase period arbitrarily into N unit purchase periods, and extracting a purchase history including a purchase item of each user and a purchase date of each user in each unit purchase period;
(b) calculating a degree of similarity for each unit purchasing period using the purchasing history in each unit purchasing period;
(c) calculating a similarity average in the entire purchasing period by summing the similarities of the unit purchasing periods; And
(d) collaboratively filtering the user similarity average and purchase history in the entire purchase period to generate recommendation information in the order of higher product preference per user. 13. The method of claim 12,
The step (b)
And processing the purchasing history in each unit purchasing period by a jacquard similarity calculating method or a cosine similarity calculating method to calculate the similarity for each unit purchasing period. 13. The method of claim 12,
Between step (b) and step (c)
Further comprising the step of assigning a relatively high weight to the similarity degree of each of the N unit purchase periods with respect to the latest purchase period. Dividing the arbitrarily set total purchasing period into N unit purchasing periods, and extracting purchase histories including the purchasers of each user and each purchase date of each user in each unit purchasing period; Calculating a degree of similarity for each unit purchase period using purchase histories in each unit purchase period; Calculating a similarity average in the entire purchasing period by summing the similarities of the unit purchasing periods; And a step of collaboratively filtering the user similarity average and the purchase history in the entire purchase period to generate recommendation information in the order of higher product preference per user is recorded as a program and recorded on a recording medium readable by an electronic device . (a) dividing an entirely purchased purchase period arbitrarily into N unit purchase periods, and extracting a purchase history including a purchase item of each user and a purchase date of each user in each unit purchase period;
(b) performing a collaborative filtering process on the similarity and the purchase history in the unit purchase period to calculate product preference per user in each unit purchase period;
(c) calculating a preference average in the entire purchase period by summing product preferences of each user in each unit purchase period; And
(d) generating recommendation information in descending order of product preference per user according to the calculated average value. 17. The method of claim 16,
Between step (b) and step (c)
Further comprising the step of assigning a relatively high weight to the product preference of each of the N unit purchase periods for the latest purchase period. Dividing the arbitrarily set total purchasing period into N unit purchasing periods, and extracting purchase histories including the purchasers of each user and each purchase date of each user in each unit purchasing period; Performing collaborative filtering processing of the degree of similarity and the purchase history in the unit purchase period to calculate product preference per user in each unit purchase period; Calculating a preference average in the entire purchase period by summing product preferences of each user in each unit purchase period; And generating recommendation information in descending order of product preference for each user according to the calculated average value, wherein the recommendation method is recorded as a program and readable in an electronic device.
KR1020120122108A 2012-10-31 2012-10-31 Purchase recommendation service system and method KR20140056731A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1020120122108A KR20140056731A (en) 2012-10-31 2012-10-31 Purchase recommendation service system and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020120122108A KR20140056731A (en) 2012-10-31 2012-10-31 Purchase recommendation service system and method

Publications (1)

Publication Number Publication Date
KR20140056731A true KR20140056731A (en) 2014-05-12

Family

ID=50887817

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020120122108A KR20140056731A (en) 2012-10-31 2012-10-31 Purchase recommendation service system and method

Country Status (1)

Country Link
KR (1) KR20140056731A (en)

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104346476A (en) * 2014-11-20 2015-02-11 西安电子科技大学 Personalized object recommending method based on object similarity and network structure
KR20150144916A (en) * 2014-06-17 2015-12-29 오은석 system and method providing a suited shopping information by customer profiling
KR20160095704A (en) * 2015-02-03 2016-08-12 이화여자대학교 산학협력단 Method and system for construction of compound network database and drug virtual screening
KR20160118644A (en) * 2015-04-02 2016-10-12 주식회사 지케이21글로벌 System and Method for recommending item, and device thereof
CN109064261A (en) * 2018-07-12 2018-12-21 北京京东金融科技控股有限公司 The method, apparatus and computer readable storage medium of Recommendations
KR20200088132A (en) * 2019-01-14 2020-07-22 카페24 주식회사 Item recommendataion method and apparatus between group member and group item
KR20200111046A (en) * 2019-03-18 2020-09-28 동아대학교 산학협력단 Method and apparatus for calculating two-way recommended scores based on data of purchase
KR20200116615A (en) * 2019-04-02 2020-10-13 동아대학교 산학협력단 Method and apparatus for calculating recommendation score based on conditional probability
CN111815351A (en) * 2020-05-29 2020-10-23 杭州览众数据科技有限公司 Cooperative filtering and association rule-based clothing recommendation method
CN112182411A (en) * 2020-11-09 2021-01-05 广州酷狗计算机科技有限公司 Multimedia data recommendation method and device, computer equipment and readable storage medium
CN112925993A (en) * 2021-04-08 2021-06-08 国网电子商务有限公司 Collaborative filtering recommendation method and device, storage medium and equipment
KR102284432B1 (en) * 2020-01-31 2021-08-03 카페24 주식회사 Product ranking fluctuation indicator calculation method of shopping mall related to e-commerce and apparatus using said method
KR20210098289A (en) * 2020-01-31 2021-08-10 카페24 주식회사 Product display update method of shopping mall related to e-commerce and apparatus using said method
KR20220001618A (en) * 2020-06-30 2022-01-06 카페24 주식회사 Method, Apparatus and System for Recommendation in Groups Using Bigdata
KR20220040903A (en) * 2020-09-24 2022-03-31 이대근 Apparatus of alcoholic liquors recommendation using purchase history
KR102415887B1 (en) * 2021-10-05 2022-07-05 우재희 Method of operating furniture rental platform and apparatus therefor
CN116452304A (en) * 2023-06-16 2023-07-18 深圳迅销科技股份有限公司 Cross-domain green consumption scene integration and preferential recommendation method

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20150144916A (en) * 2014-06-17 2015-12-29 오은석 system and method providing a suited shopping information by customer profiling
CN104346476A (en) * 2014-11-20 2015-02-11 西安电子科技大学 Personalized object recommending method based on object similarity and network structure
CN104346476B (en) * 2014-11-20 2017-07-04 西安电子科技大学 Personalized item recommendation method based on article similarity and network structure
KR20160095704A (en) * 2015-02-03 2016-08-12 이화여자대학교 산학협력단 Method and system for construction of compound network database and drug virtual screening
KR20160118644A (en) * 2015-04-02 2016-10-12 주식회사 지케이21글로벌 System and Method for recommending item, and device thereof
CN109064261A (en) * 2018-07-12 2018-12-21 北京京东金融科技控股有限公司 The method, apparatus and computer readable storage medium of Recommendations
KR20200088132A (en) * 2019-01-14 2020-07-22 카페24 주식회사 Item recommendataion method and apparatus between group member and group item
KR20200111046A (en) * 2019-03-18 2020-09-28 동아대학교 산학협력단 Method and apparatus for calculating two-way recommended scores based on data of purchase
KR20200116615A (en) * 2019-04-02 2020-10-13 동아대학교 산학협력단 Method and apparatus for calculating recommendation score based on conditional probability
KR102284432B1 (en) * 2020-01-31 2021-08-03 카페24 주식회사 Product ranking fluctuation indicator calculation method of shopping mall related to e-commerce and apparatus using said method
KR20210098289A (en) * 2020-01-31 2021-08-10 카페24 주식회사 Product display update method of shopping mall related to e-commerce and apparatus using said method
CN111815351A (en) * 2020-05-29 2020-10-23 杭州览众数据科技有限公司 Cooperative filtering and association rule-based clothing recommendation method
KR20220001618A (en) * 2020-06-30 2022-01-06 카페24 주식회사 Method, Apparatus and System for Recommendation in Groups Using Bigdata
KR20220040903A (en) * 2020-09-24 2022-03-31 이대근 Apparatus of alcoholic liquors recommendation using purchase history
CN112182411A (en) * 2020-11-09 2021-01-05 广州酷狗计算机科技有限公司 Multimedia data recommendation method and device, computer equipment and readable storage medium
CN112925993A (en) * 2021-04-08 2021-06-08 国网电子商务有限公司 Collaborative filtering recommendation method and device, storage medium and equipment
KR102415887B1 (en) * 2021-10-05 2022-07-05 우재희 Method of operating furniture rental platform and apparatus therefor
CN116452304A (en) * 2023-06-16 2023-07-18 深圳迅销科技股份有限公司 Cross-domain green consumption scene integration and preferential recommendation method
CN116452304B (en) * 2023-06-16 2023-09-01 深圳迅销科技股份有限公司 Cross-domain green consumption scene integration and preferential recommendation method

Similar Documents

Publication Publication Date Title
KR20140056731A (en) Purchase recommendation service system and method
CN103823908B (en) Content recommendation method and server based on user preference
CN104915734B (en) Commodity popularity prediction method based on time sequence and system thereof
US20220207542A1 (en) Machine learning for marketing of branded consumer products
CN107230098A (en) Method and system is recommended in a kind of timesharing of business object
Yadav et al. Retailer's optimal policy under inflation in fuzzy environment with trade credit
JP6027679B2 (en) Method and device for determining hot selling goods for holidays
CN109658206A (en) Information recommendation method and device
US20200005366A1 (en) System and method for integrating retail price optimization for revenue and profit with business rules
CN102282551A (en) Automated decision support for pricing entertainment tickets
CN103678518A (en) Method and device for adjusting recommendation lists
Tanaka A sales forecasting model for new-released and nonlinear sales trend products
CN109493113A (en) A kind of providing method and device of favor information
CN110880127B (en) Consumption level prediction method and device, electronic equipment and storage medium
CN112330358A (en) Method and apparatus for product sales prediction, storage medium, and electronic device
Tan et al. A discrete-in-time deteriorating inventory model with time-varying demand, variable deterioration rate and waiting-time-dependent partial backlogging
CN112579876A (en) Information pushing method, device and system based on user interest and storage medium
CN110570233A (en) User buyback time prediction method and device for e-commerce platform
CN110689402A (en) Method and device for recommending merchants, electronic equipment and readable storage medium
KR102097045B1 (en) Method and apparatus to recommend products reflecting characteristics of users
Hemalatha Market basket analysis–a data mining application in Indian retailing
US20220414579A1 (en) Salesperson evaluation apparatus, salesperson evaluation method, and salesperson evaluation program
CN111127074A (en) Data recommendation method
Li et al. Optimal pricing for a short life‐cycle product when customer price‐sensitivity varies over time
Sivashankari et al. Production inventory model for two levels production with defective items and incorporating multi-delivery policy

Legal Events

Date Code Title Description
A201 Request for examination
E902 Notification of reason for refusal
E90F Notification of reason for final refusal
E601 Decision to refuse application