KR102495011B1 - Method for recommending product using regression analysis according to product code and purchase pattern and apparatus for the same - Google Patents

Abstract

A product recommendation method and apparatus using regression analysis according to product codes and purchase patterns are disclosed. A product recommendation method using regression analysis according to product codes and purchase patterns includes receiving a purchase request from a user terminal; completing a transaction corresponding to the purchase request and obtaining a reference purchase pattern corresponding to a target user of the user terminal; selecting a plurality of candidate users by comparing the reference purchase pattern with purchase patterns of each of a plurality of users previously collected; obtaining a first target purchase detail of the target user and a plurality of first purchase details corresponding to each of the candidate users, based on a purchase time point according to the purchase request; comparing the first target purchase details with the first purchase details and selecting one of the candidate users as a comparison target user; Based on the purchase time according to the purchase request, the second purchase details of the comparison target user and the second target purchase details of the target user are newly acquired, and the second purchase details and the second target purchase details are compared with each other to determine a recommended product; and transmitting a transaction completion message including the recommended product to the user terminal.

Description

Product recommendation method and apparatus using regression analysis according to product code and purchase pattern

The present invention relates to a product recommendation system, and more particularly, to a product recommendation method and apparatus using regression analysis according to a product code and purchase pattern.

The product recommendation system may refer to a system that recommends products to a user using information filtering.

An information system is a technique that obtains a user's information profile by querying the user for personal information, areas of interest, and preferences. Based on this information, a recommendation system recommends or provides information and products suitable for the customer's psychological information and preference information. way to do it It can be applied to searches such as movies, music, news, books, research topics, search queries, and products.

On the other hand, it is common for these recommendation systems to recommend products in consideration of customer preferences and purchase history, but they are limited to recommending products based on matched items by simply comparing preferences and purchase history, so the recommendation accuracy is not high. there is a problem.

In addition, when all purchase details of all users are collectively compared and judged, a lot of computational load is consumed, which causes a delay to immediately check recommended products on the user terminal, which causes inconvenience to the user. cause

Therefore, there is a need for a method of increasing recommendation accuracy by quickly selecting recommended products and providing them to the user, and at the same time providing the user with a product that is estimated to have the highest purchase incentive for the user.

An object of the present invention to solve the above problems is to provide a product recommendation method and apparatus using regression analysis according to product codes and purchase patterns.

One aspect of the present invention for achieving the above object provides a product recommendation method and apparatus using regression analysis according to product codes and purchase patterns.

A product recommendation method using regression analysis according to product codes and purchase patterns includes receiving a purchase request from a user terminal; completing a transaction corresponding to the purchase request and obtaining a reference purchase pattern corresponding to a target user of the user terminal; selecting a plurality of candidate users by comparing the reference purchase pattern with purchase patterns of each of a plurality of users previously collected; obtaining a first target purchase detail of the target user and a plurality of first purchase details corresponding to each of the candidate users, based on a purchase time point according to the purchase request; comparing the first target purchase details with the first purchase details and selecting one of the candidate users as a comparison target user; Based on the purchase time according to the purchase request, the second purchase details of the comparison target user and the second target purchase details of the target user are newly acquired, and the second purchase details and the second target purchase details are compared with each other to determine a recommended product; and transmitting a transaction completion message including the recommended product to the user terminal.

In the obtaining of the reference purchase pattern, a purchase trend line may be obtained by performing a linear regression analysis on a graph showing the number of purchases for each purchase period of the target user.

The selecting of the plurality of candidate users may include determining an upward threshold value larger than a threshold value and a downward threshold value smaller than the threshold value based on an intercept of the y-axis representing the number of purchases after the purchase trend line passes through the graph. step; Among the lines passing through two or more coordinates in the graph, a line whose y-axis intercept falls between the upward threshold and the downward threshold and has the largest slope is selected as the upward threshold line, and the y-axis intercept is the upward threshold. selecting a line between a threshold value and the downward threshold value and having the smallest slope as a downward threshold line; and selecting, as the candidate users, users corresponding to a trend line passing through an area between the upper threshold line and the lower threshold line among trend lines calculated using the number of purchases for each purchase period for each of the users. .

The step of selecting one of the candidate users as a comparison target user includes products included in the first target purchase details and each of the first purchase details based on a product code uniquely assigned to each product in advance. comparing the products to each other to determine similar products; and selecting, as the comparison target user, a candidate user corresponding to first purchase details having a preset number or more of similar products among the first purchase details.

The product code may include a first code indicating a large category of products; A second code indicating the subclass of the product, a third code indicating the brand or producer of the product, a fourth code indicating the characteristics of the product, and a fifth code indicating the raw material of the product.

In the determining of the similar product, a product having the same first code, the second code, and the third code may be determined as the similar product.

Determining the recommended product may include generating product vectors corresponding to each of the products included in the second target purchase details using the product code; generating a product vector matrix having each of the product vectors as a column; calculating an eigenvector of the product vector matrix; and determining one of the products included in the second purchase details as the recommended product by comparing the eigenvector with product vectors corresponding to respective products included in the second purchase details. .

In the case of using the product recommendation method and apparatus using regression analysis according to product codes and purchase patterns according to the present invention as described above, it is possible to quickly determine recommended products because rapid calculation is performed using product codes set for each product in advance. there is.

In addition, since the user's purchase pattern is quantified and compared with other users' purchase patterns with a clearly set indicator, anyone can accurately compare the purchase patterns using the present invention.

In addition, purchase details used to determine the comparison target user may determine a candidate user similar to a recent purchase pattern of the target user by utilizing purchase details at a relatively recent point in time based on the present.

In addition, the recommended product is determined by using the past purchase history, which is considerably past the point of purchase by the target user, so that the product that has not been purchased recently but is similar to the product that has been purchased in the past can be induced to purchase again as a recommended product. there is.

1 is a diagram illustrating an environment in which a product recommendation method according to an embodiment of the present invention is performed.
FIG. 2 is a representative flowchart illustrating an operation of the product recommendation device according to FIG. 1 .
FIG. 3 is a graph for explaining a process of comparing a standard purchase pattern and a purchase pattern of each user in step S120 according to FIG. 2 .
FIG. 4 is a conceptual diagram for explaining a process of comparing target purchase details and purchase details of each candidate user in step S140 according to FIG. 2 .
FIG. 5 is a conceptual diagram for explaining a process of determining a recommended product by comparing purchase details of a comparison target user and target purchase details of a target user in step S150 according to FIG. 2 .
6 is a diagram showing a hardware configuration of a product recommendation device according to an exemplary embodiment.

Since the present invention can make various changes and have various embodiments, specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. Like reference numerals have been used for like elements throughout the description of each figure.

Terms such as first, second, A, and B may be used to describe various components, but the components should not be limited by the terms. These terms are only used for the purpose of distinguishing one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element, without departing from the scope of the present invention. The terms and/or include any combination of a plurality of related recited items or any of a plurality of related recited items.

It is understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, but other elements may exist in the middle. It should be. On the other hand, when an element is referred to as “directly connected” or “directly connected” to another element, it should be understood that no other element exists in the middle.

Terms used in this application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, the terms "include" or "have" are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more other features It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and unless explicitly defined in the present application, they should not be interpreted in an ideal or excessively formal meaning. don't

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram illustrating an environment in which a product recommendation method according to an embodiment of the present invention is performed.

Referring to FIG. 1 , a product recommendation method based on a user's purchase details may be performed by a product recommendation device 100 (which may be abbreviated as a product recommendation device hereinafter) based on the user's purchase details.

The product recommendation device 100 may support purchase of products online by communicating with the user terminal 200 through a wireless or wired network. For example, the product recommendation apparatus 100 receives user information from the user terminal 200, authenticates the user information, permits access of the user terminal 200, and provides access from the user terminal 200 to which access is allowed. Purchase requests can be received.

For example, the user terminal 200 includes a communicable desktop computer, a laptop computer, a notebook, a smart phone, a tablet PC, and a mobile phone. ), smart watch, smart glass, e-book reader, portable multimedia player (PMP), portable game device, navigation device, digital camera, digital multimedia broadcasting (DMB) It may be a player, a digital audio recorder, a digital audio player, a digital video recorder, a digital video player, a personal digital assistant (PDA), and the like.

The product recommendation device 100 may complete a transaction of the product included in the purchase request and notify the user terminal 200 of a message C31 instructing the transaction completion. At this time, the product recommendation device 100 may determine and provide the recommended product C32 to the user terminal 200 together with the message C31 indicating the completion of the transaction.

Also, the product recommendation device 100 may determine the expected arrival time C33 of the recommended product C32 and provide the determined arrival time C33 to the user terminal 200 .

The product recommendation device 100 may store product information corresponding to each of a plurality of pre-registered products. For example, the product information may include a product name, a product code uniquely assigned to each product, unit price, brand name, manufacturer, country of origin, shape, color, and the like.

In addition, the product recommendation device 100 may generate purchase details corresponding to each user of the user terminals 200 by accumulating transaction completion processing results according to purchase requests in conjunction with a plurality of user terminals 200. there is.

Here, the purchase details may include a purchase completion time, the number of purchases, a list of purchased products, and the like.

The product recommendation device 100 determines a recommended product C32 corresponding to a user of a specific user terminal 200 by referring to purchase details and product information, and transmits the determined recommended product C32 to the user terminal 200 By doing so, recommended products can be displayed on the user terminal 200 .

FIG. 2 is a representative flowchart illustrating an operation of the product recommendation device according to FIG. 1 .

Referring to FIG. 2 , the product recommendation device 100 may receive a purchase request from the user terminal 200 (S100).

The product recommendation device 100 may complete a transaction corresponding to the received purchase request and obtain a reference purchase pattern corresponding to the target user of the user terminal 200 (S110).

Here, the reference purchase pattern may be determined based on the number of purchases according to the purchase timing of the target user.

The product recommendation apparatus 100 may select a plurality of candidate users by comparing the reference purchase pattern with the pre-collected purchase patterns of each of the plurality of users (S120). For example, the product recommendation apparatus 100 may select users having purchase patterns within a preset range based on a reference purchase pattern as candidate users.

Here, the candidate users may be users who are determined to have a purchase pattern similar to that of the target user and provide purchase details for determining a recommended product.

The product recommendation apparatus 100 may obtain the target purchase details of the target user and purchase details corresponding to each of the candidate users based on the purchase time according to the purchase request (S130). For example, target purchase details belonging to a past point in time within a preset threshold period (for example, one month) based on the purchase point (current) according to the purchase request of the target user and purchase details corresponding to each of the candidate users can be obtained

The product recommendation apparatus 100 may select one of the candidate users as a comparison target user by comparing target purchase details and purchase details (S140). Here, the comparison target user may provide one of the products included in the comparison target user's purchase details to the target user as a recommended product.

For example, the product recommendation apparatus 100 compares products included in the target purchase details with products included in each of the purchase details, and selects products similar to the products included in the target purchase details as purchase details. A user may be selected as a comparison target user. At this time, the product recommendation device 100 may compare products included in the target purchase details with products included in each of the purchase details based on product codes assigned to each of the products in advance.

The product recommendation device 100 newly acquires the purchase details of the comparison target user and the target purchase details based on the purchase time according to the purchase request, and compares the acquired target purchase details and purchase details with each other to determine a recommended product. It can (S150). For example, the product recommendation device 100 is a comparison target user corresponding to a past time point exceeding a predetermined critical period (eg, 3 months) based on the purchase time (current) according to the purchase request of the target user. Purchase history and target purchase history can be compared with each other.

That is, in an embodiment of the present invention, by determining a user to be compared using the purchase history within the most recent (eg, one month) based on the purchase time (current) according to the purchase request, the target user's recent purchase performance Select a user with similar purchase performance as a user for comparison, but purchase history used to select a new recommended product belongs to a preset critical period based on the purchase time according to the current purchase request. Therefore, there is an advantage in that a product without a recent purchase record can be selected as a recommended product.

When the recommended product is determined, the product recommendation device 100 may provide the user terminal 200 with a transaction completion message including the recommended product.

FIG. 3 is a graph for explaining a process of comparing a standard purchase pattern and a purchase pattern of each user in step S120 according to FIG. 2 .

Referring to FIG. 3 , in one embodiment of the present invention, comparison of purchase patterns is performed using the number of purchases according to the purchase time. In this case, the number of purchases may vary greatly for each user. For example, in the case of users with a high proportion of product purchases, there is a large number of purchases at multiple times, so there is abundant data for determining purchase patterns, whereas in the case of users who purchase small products themselves, there are a small number of purchases at very few times. There are many cases where there are purchases of

The product recommendation device 100 according to an embodiment of the present invention uses the number of purchases per purchase time that has a record to easily derive a purchase pattern even if the user has a relatively small number of purchases according to the purchase time. ) can be calculated.

For example, as shown in FIG. 2, in the case of a user who has one or more purchases in each of the four periods of January 10, July 12, September 18, and December 31, purchases according to the four periods The number of purchases per purchase period may be generated as a graph using only the number of cases, and a purchase trend line RL may be calculated from the generated graph.

Here, the purchase trend line (RL) can be obtained by performing a widely known linear regression analysis on a graph showing the number of purchases for a given purchase period, and the purchase trend line (RL) This can be derived and has the advantage of being suitable for use as a purchase pattern.

Here, the calculated purchase trend line (RL) is a straight line equation representing the number of purchases (y) according to the purchase time (x) (y=ax+b, x is a variable for the purchase time, and y is a variable for the number of purchases) , specified by the slope (a) and the y-axis intercept (b). A purchase trend line (RL) may be used as a meaning representing a standard purchase pattern of a target user.

The product recommendation device 100, based on the y-axis intercept (b) specified by the purchase trend line (RL), an upward threshold (b + th) as large as the threshold (th) and a downward as small as the threshold (th) A threshold value (b-th) can be determined. For example, the interval between the y-axis intercept (b) and the upward threshold value (b+th) and the interval (th) between the y-axis intercept (b) and the downward threshold value (b-th) may be equal to each other.

Next, in the product recommendation device 100, among the lines passing through two or more coordinates in the graph representing the number of purchases by purchase time, the y-axis intercept is an upward threshold value (b+de) and a downward threshold value (b-de) ), and the line with the largest slope is selected as the upward threshold (UL), the y-axis intercept falls between the upward threshold (b + de) and the downward threshold (b-de), and the slope is The smallest line may be selected as the downward critical line (LL).

At this time, the product recommendation device 100, if the y-axis intercept among the lines passing through two or more coordinates does not fall between the upper threshold value (b + de) and the lower threshold value (b-de), Among the lines passing through one coordinate, the y-axis intercept passes through the upward threshold value (b + de) and the line with the largest slope is selected as the upward threshold line (UL), and the y-axis intercept passes through the downward threshold value (b-de ), and a line having the smallest slope may be selected as the downward critical line LL.

In FIG. 2, among the lines passing through two or more coordinates, the y-axis intercept belongs between the upward threshold value (b+de) and the downward threshold value (b-de), and there is a line with the largest slope, so that the upward threshold line (UL). However, among the lines passing through two or more coordinates, there is no line whose y-axis intercept falls between the upward threshold value (b+de) and the downward threshold value (b-de) and has the smallest slope.

Therefore, in FIG. 2 , among lines passing through one coordinate, a line having a y-axis intercept passing through a downward threshold value (b-de) and having the smallest slope is selected as a downward threshold line (LL).

The product recommendation apparatus 100 calculates trend lines corresponding to each of the plurality of users by using the number of purchases according to the purchase time of each of the plurality of users, and among the calculated trend lines, an upper threshold line (UL) and a lower threshold line Users having a trend line passing through an area between the lines LL may be selected as candidate users.

That is, users having an area between the upper threshold line UL and the lower threshold line LL shown in FIG. 2 as a trend line may be selected as candidate users. Candidate users selected here may be determined to be users whose purchase pattern, which is confirmed through the number of purchases according to the purchase time, is within an error range (upper threshold line, lower threshold line) from the target purchase pattern of the target user.

FIG. 4 is a conceptual diagram for explaining a process of comparing target purchase details and purchase details of each candidate user in step S140 according to FIG. 2 .

Referring to FIG. 4 , the target purchase history H1 belonging to a predetermined threshold period (for example, 1 month) based on the purchase time (current) according to the purchase request of the target user (user A) includes jewelry, grapes, and keyboards. , shoes, power drills, and the like.

In addition, the purchase details H2 belonging to a predetermined threshold period based on the present of the candidate user B may include jewelry, a shine musket, a mouse, shoes, a hammer, and the like.

In addition, the purchase details H2 belonging to a preset threshold period based on the present of candidate user C may include jackets, jewelry, handbags, dining tables, blankets, and the like.

In addition, the purchase history H2 belonging to a predetermined threshold period based on the present of candidate user D may include a bed, instant noodles, a watch, jewelry, wine, and the like.

The product recommendation device 100 may compare the target purchase details H1 and the purchase details H2, H3, and H4 of each of the candidate users to select a comparison target user whose number of similar products is equal to or greater than a preset number. there is.

At this time, in order to maximize the calculation speed, the product recommendation device 100 may use a product code uniquely assigned to each of the products in advance. Here, the product code includes a first code (CD1) indicating the major classification of the product, a second code (CD2) indicating the subclassification of the product, a third code (CD3) indicating the brand or manufacturer of the product, and a product characteristic. A fourth code (CD4) indicating the product and a fifth code (CD5) indicating the raw material of the product may be included. These product codes are constructed as a kind of big data by being systematically input by a manager when products are initially registered in the product recommendation device 100 .

Here, the major classification may indicate whether the product belongs to clothing or jewelry, and the subcategory may indicate whether the product belongs to clothing, such as a T-shirt or jeans belonging to clothing. Characteristics of the product may include the color and shape of the product.

Therefore, if the first code CD1 is the same, it can be known that the product is the same clothing or jewelry, and if the first code CD1 and the second code CD2 are the same, it can be known that the product is the same T-shirt or jeans. there is.

The product recommendation device 100 may determine a similar product by referring to the product code assigned in this way. For example, the product recommendation device 100 treats products having the same first to third codes (CD1 to CD3) included in the product codes as similar products, and compares the number of similar products to a preset number or more. Users can be selected.

As such, since the product code of the present invention is not assigned as a randomly set code without meaning, but given in consideration of the classification and characteristics of the product, similar products can be judged by simply comparing only a part of the product code, As a result, there is an advantage in that it can be performed at a very high computational speed when comparing similar products to a large number of candidate users.

FIG. 5 is a conceptual diagram for explaining a process of determining a recommended product by comparing purchase details of a comparison target user and target purchase details of a target user in step S150 according to FIG. 2 .

When the comparison target user is determined, the product recommendation device 100 compares data corresponding to a past point in time exceeding a predetermined threshold period (for example, 3 months) based on the purchase point (current) according to the purchase request of the target user. Purchase details of the target user and target purchase details are newly acquired.

That is, the purchase history used to determine the comparison target user can determine a candidate user similar to the target user's recent purchase pattern by utilizing the purchase history at a relatively recent point in time based on the present. By making a decision using the past purchase details, which have not been purchased recently, but similar to products that have been purchased in the past, it is induced to re-purchase as a recommended product.

Specifically, the product recommendation device 100 may generate a product vector for each product by using a product code corresponding to each product included in the newly acquired target purchase details H11 of the target user.

For example, a product vector having values corresponding to each of the first to fifth codes CD1 to CD5 constituting the product code as component values may be generated. That is, the product vector has a value corresponding to 'A1' as a first component value when the first code CD1 is 'A1', and a value corresponding to 'B8' when the second code CD2 is 'B8'. as a second component value, and if the third code CD3 is 'C1', has a value corresponding to 'C1' as a third component value, and if the fourth code CD4 is 'D3', corresponds to 'D3' is the fourth component value, and if the fifth code CD5 is 'EA', it may have a value corresponding to 'EA' as the fifth component value.

Here, a value corresponding to each code is determined by referring to a code table in which codes and corresponding values are defined in advance, and it is advantageous to define similar values as codes are similar. For example, each of two products whose subclasses are jeans and cotton pants are defined as having similar values such as 3 and 4 corresponding to the second code, and two products whose subclasses are jeans and T-shirts. It is advantageous to define each as a value slightly different from the value corresponding to the second code, such as 3 and 6.

As such, since the product code uniquely assigned to each product can be directly converted into a product vector using the code table, the operation speed is very fast.

In this way, the product recommendation device 100 may generate product vectors x1 to x5 corresponding to each of the products included in the target purchase details H11.

Next, the product recommendation device 100 generates a product vector matrix having each of the generated product vectors (x1 to x5) as a column, and calculates an eigenvector of the generated product vector matrix. can Here, if each of the product vectors is a 5-dimensional vector composed of 5 components, the eigenvector may also be a 5-dimensional vector. Since the method for calculating the eigenvector has already been disclosed through various matrix operations, a detailed description thereof will be omitted.

The product recommendation device 100 may compare the calculated eigenvector with product vectors y1 to y5 generated by converting products included in the comparison target user's purchase history H22 .

Specifically, the product recommendation device 100 selects a product vector having an azimuth angle most similar to an azimuth angle representing an eigenvector in a multidimensional space from among product vectors y1 to y5, and selects a product corresponding to the selected product vector. It can be decided as a recommended product.

In this case, when the eigenvector is a 5-dimensional vector, the eigenvector can be represented on a 5-dimensional space, and a direction angle between two vectors on this multi-dimensional space can be determined through a cosine operation between the two vectors. For example, the direction angle θ between the eigenvector ev and one of the product vectors yi may be determined according to Equation 1 below.

In Equation 1, the numerator part of arccosine is the dot product operation between the eigenvector (ev) and the product vector (yi), and the denominator part is the product of each magnitude of the eigenvector (ev) and product vector (yi).

A product vector having the smallest orientation angle (θ) between the eigenvector (ev) and the product vector (yi) determined according to Equation 1 may be selected, and a product corresponding to the selected product vector may be determined as a recommended product. .

As such, in the present invention, since an eigenvector (ev) is calculated from products purchased by a target user in the past, and a product of a user to be compared is selected based on the calculated eigenvector (ev), it is simply Deviating from the conventional technique of providing products with common products as recommended products, products similar to eigenvectors representing all products purchased by the target user in the past may be provided as recommended products.

6 is a diagram showing a hardware configuration of a product recommendation device according to an exemplary embodiment.

Referring to FIG. 6 , the product recommendation apparatus 100 stores at least one processor 110 and instructions instructing the at least one processor 110 to perform at least one operation. It may include a memory (memory) to.

The at least one operation may include at least one of operations or functions of the product recommendation device described with reference to FIGS. 1 to 5 .

Here, the at least one processor 110 may mean a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor for performing methods according to embodiments of the present invention. can Each of the memory 120 and the storage device 160 may include at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory 120 may be one of a read only memory (ROM) and a random access memory (RAM), and the storage device 160 may be a flash-memory. , a hard disk drive (HDD), a solid state drive (SSD), or various memory cards (eg, a micro SD card).

Also, the product recommendation device 100 may include a transceiver 130 that performs communication through a wireless network. Also, the product recommendation device 100 may further include an input interface device 140 , an output interface device 150 , a storage device 160 , and the like. Each component included in the product recommendation device 100 is connected by a bus 170 to communicate with each other.

For example, the product recommendation device 100 may be a communicable desktop computer, a laptop computer, a notebook, a smart phone, a tablet PC, or a mobile phone. phone), smart watch, smart glass, e-book reader, portable multimedia player (PMP), portable game device, navigation device, digital camera, digital multimedia broadcasting (DMB) ) player, digital audio recorder, digital audio player, digital video recorder, digital video player, personal digital assistant (PDA), and the like.

The methods according to the present invention may be implemented in the form of program instructions that can be executed by various computer means and recorded on a computer readable medium. Computer readable media may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on a computer readable medium may be specially designed and configured for the present invention or may be known and usable to those skilled in computer software.

Examples of computer readable media may include hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions may include not only machine language codes generated by a compiler but also high-level language codes that can be executed by a computer using an interpreter and the like. The hardware device described above may be configured to operate with at least one software module to perform the operations of the present invention, and vice versa.

In addition, the above-described method or device may be implemented by combining all or some of its components or functions, or may be implemented separately.

Although the above has been described with reference to preferred embodiments of the present invention, those skilled in the art will variously modify and change the present invention within the scope not departing from the spirit and scope of the present invention described in the claims below. You will understand that it can be done.

Claims (1)

It is a product recommendation method using regression analysis according to product code and purchase pattern.
Receiving a purchase request from a user terminal;
completing a transaction corresponding to the purchase request and obtaining a reference purchase pattern corresponding to a target user of the user terminal;
selecting a plurality of candidate users by comparing the reference purchase pattern with purchase patterns of each of a plurality of users previously collected;
obtaining a first target purchase detail of the target user and a plurality of first purchase details corresponding to each of the candidate users, based on a purchase time point according to the purchase request;
comparing the first target purchase details with the first purchase details and selecting one of the candidate users as a comparison target user;
Based on the purchase time according to the purchase request, the second purchase details of the comparison target user and the second target purchase details of the target user are newly acquired, and the second purchase details and the second target purchase details are compared with each other to determine a recommended product; and
Transmitting a transaction completion message including the recommended product to the user terminal;
The step of selecting one of the candidate users as a comparison target user,
determining similar products by comparing products included in the first target purchase details with products included in each of the first purchase details based on a product code uniquely assigned to each product in advance; and
Selecting a candidate user corresponding to first purchase details having a preset number or more of similar products among the first purchase details as the comparison target user;
The product code is
A first code indicating a major classification of goods; A second code indicating the subclass of the product, a third code indicating the brand or producer of the product, a fourth code indicating the characteristics of the product, and a fifth code indicating the raw material of the product,
The step of determining the similar product is,
The first code, the second code, and the third code determine a product identical to each other as the similar product;
The step of determining the recommended product,
generating product vectors corresponding to respective products included in the second target purchase details using the product code;
generating a product vector matrix having each of the product vectors as a column;
calculating an eigenvector of the product vector matrix;
The eigenvector and the orientation angle between the product vectors corresponding to each of the products included in the second purchase details are calculated by the following equation

, (ev is the eigenvector, yi is one of the product vectors)
Calculating based on; and
Determining a product corresponding to the product vector having the smallest orientation angle among products included in the second purchase details as the recommended product;
The step of obtaining the reference purchase pattern,
Obtaining a purchase trend line by performing linear regression analysis on a graph showing the number of purchases by purchase time of the target user;
The step of selecting the plurality of candidate users,
determining an upward threshold value that is greater than a threshold value and a downward threshold value that is less than the threshold value based on an intercept of a y-axis representing the number of purchases after the purchase trend line in the graph;
Among the lines passing through two or more coordinates in the graph, a line whose y-axis intercept falls between the upward threshold and the downward threshold and has the largest slope is selected as the upward threshold line, and the y-axis intercept is the upward threshold. selecting a line between a threshold value and the downward threshold value and having the smallest slope as a downward threshold line; and
Selecting users corresponding to a trend line passing through a region between the upper threshold line and the lower threshold line among trend lines calculated using the number of purchases by purchase time for each of the users as the candidate users, How to recommend a product.

Images