KR20200080024A - A recommendation system for product purchase using collaborative filtering algorism and method thereof - Google Patents

Abstract

The present invention relates to a system for making a recommendation related to product purchase using collaborative filter algorithm and a method thereof. The method comprises the steps of: storing product information on a plurality of products; storing existing customer information on an existing customer who has purchased at least one of the plurality of products for a predetermined period; generating purchase data based on whether the existing customer purchases each product and the frequency of purchase, with respect to each of the plurality of products; and calculating a recommendation score corresponding to the purchase possibility of a new customer for a predetermined product based on a result of applying a positive or negative Pearson correlation to the purchase data. Accordingly, when calculating the recommendation score for the predetermined product, the contribution ratio of the data on the Pearson correlation in each of two cases in which the existing customer purchased or did not purchase the product is considered together, thereby increasing the recommendation performance.

Description

A RECOMMENDATION SYSTEM FOR PRODUCT PURCHASE USING COLLABORATIVE FILTERING ALGORISM AND METHOD THEREOF}

The present invention relates to a product purchasing related recommendation system and method for improving recommendation performance in consideration of both a case where an existing customer purchases a product and a case where a product is not purchased when calculating a recommendation score for a given product using a collaborative filter algorithm will be.

The recommendation system has become a factor influencing the competitiveness of service providers in online businesses such as the search field, advertising marketing field, e-commerce field, and multimedia content distribution field. The personalized target recommendation system can significantly reduce the time it takes for customers to reach the desired item online and provide one-on-one marketing to drive consumption, which can have a great impact on corporate performance.

Various recommendation techniques have been proposed so far, but one of the most successful recommendation techniques is collaborative filtering (CF). This is a technique that analyzes the purchase propensity of multiple customers on a large scale through their past purchase records, and recommends items to specific customers by referring to purchase records purchased by other customers who have a high correlation with specific customers. Many successful e-commerce companies use this collaborative filtering recommendation technique to recommend items to customers.

However, since the collaborative filtering recommendation technique recommends based on past purchase records, recommendation performance is deteriorated even when data to derive correlation is sparse because of insufficient purchase records compared to the number of items, and the recommendation score is considered In general, there is a problem that the accuracy of the prediction result for the actual product purchasing possibility may be deteriorated because only the case in which another customer purchases the product is considered.

KR 10-1567684 B1 KR 10-1567217 B1

An object of the present invention is to solve the above problems, and when calculating a recommendation score for a given product, a collaborative filter capable of improving recommendation performance by considering both the case where the existing customer purchases the product and the case where the product is not purchased It is an object of the present invention to provide a recommendation system and method related to product purchase using an algorithm.

A product purchasing-related recommendation method according to an aspect of the present invention for achieving the above object is to store product information for a plurality of products, and to an existing customer who has purchased at least one of the plurality of products for a predetermined period of time. Storing existing customer information about the product; generating purchase data based on purchase and frequency of purchase of each existing product for each of the plurality of products; and applying a positive or negative Pearson correlation to the purchase data. And calculating a recommendation score corresponding to a new customer's purchase possibility for a given product based on the result.

In addition, the product purchase-related recommendation system according to another aspect of the present invention stores storage of product information for a plurality of products and existing customer information for an existing customer who has purchased at least one of the plurality of products for a predetermined period, respectively. Based on the result of applying a positive or negative Pearson correlation to the purchase data, and a purchase data generation unit for generating purchase data based on purchase frequency and purchase frequency of each existing customer for each of the plurality of products. And a recommendation score calculating unit that calculates a recommendation score corresponding to a new customer's purchase possibility for a given product.

According to the present invention, when calculating a recommendation score for a given product, if an existing customer purchases the corresponding product or not, an effect of improving recommendation performance by considering the ratio of data contribution to each Pearson correlation is also obtained. have.

In addition, according to the present invention, it is universally applicable when calculating recommendation scores for various types of products through a simple algorithm that selects the optimal combination according to the expected purchaseability of a new customer based on the purchase frequency and purchase frequency of an existing customer. There is an advantage.

1 is a block diagram showing the configuration of a product purchasing related recommendation system using a collaborative filter algorithm according to an embodiment of the present invention,
FIG. 2 is a view for explaining a process of calculating a recommendation score by the recommendation score calculation unit of FIG. 1,
3 is a flowchart illustrating a product purchasing related recommendation method using a collaborative filter algorithm according to an embodiment of the present invention.

The problems to be solved for the present invention as described above, the means for solving the problems, and specific matters including the effects of the invention are included in the following examples and drawings. Advantages and features of the present invention, and a method of achieving the same will be apparent with reference to embodiments described below in detail together with the accompanying drawings. The same reference numerals refer to the same components throughout the specification.

1 is a block diagram showing the configuration of a product purchasing-related recommendation system using a collaborative filter algorithm according to an embodiment of the present invention, and FIG. 2 illustrates a process of calculating a recommendation score by the recommendation score calculating unit of FIG. 1 It is a drawing for doing.

Hereinafter, a recommendation system related to product purchase using a collaborative filter algorithm according to an embodiment of the present invention will be described with reference to the drawings.

Referring to FIG. 1, a recommendation system related to product purchase according to an embodiment of the present invention is largely configured to include a storage unit 100, a purchase data generation unit 300, and a recommendation score calculation unit 500.

The storage unit 100 may be provided in the form of a predetermined database, product information 101 for a plurality of products, and an existing customer for an existing customer i who has purchased at least one of the plurality of products for a predetermined period of time. Each of the information 102 is stored.

Here, the product information 101 includes information on the product name and product type of each of a plurality of products, and the existing customer information 102 includes information on the customer name and the product name corresponding to the product purchased by the customer. can do.

At this time, the product corresponding to the product information 101 is not limited to an object of a type that can be purchased and owned, and an intangible value that can be viewed or operated through a purchase act that pays a predetermined ticket value. It may also include cultural products (eg, movies, performances, concerts, etc.).

The purchase data generation unit 300 generates purchase data D based on whether the existing customer i purchases by product (Wp) and purchase frequency (Fp) for each of a plurality of products previously stored in the storage unit 100. .

Here, the purchase data generation unit 300 is a predetermined first that sets each of a plurality of products pre-stored in the storage unit 100 as one column and each of a plurality of customers pre-stored in the storage unit 100 as one row. After constructing the matrix M1, the first purchase data D ₁ may be generated by matching a binary value state value of whether a particular customer purchases each product to a corresponding row and column position.

At this time, the status value is a 1-bit binary format, and if a customer corresponding to a corresponding row purchases a product corresponding to a corresponding column, it has a value of '1' and a value of '0' otherwise. Can.

Table 1 below shows an example of the first purchase data D ₁ generated in a predetermined matrix configuration corresponding to the above-described first matrix M1.

Here, when the total number of customer information pre-stored in the storage unit 100 is'n' and the total number of product information is'm', the first matrix M1 is composed of a total of n rows and a total of m columns. It can be configured in the form of an'n×m' matrix.

For example, referring to Table 1 above, in the case of'Customer 1'in the first purchase data D,'Product 1'is purchased and'Product 2'is not purchased, so the first row corresponding to'Product 1' The position of the first column may have a status value of "1", and the position of the second column of the first row corresponding to "Product 2" may have a status value of "0".

In addition, in the case of'customer n'in the first purchase data D,'product 2'is not purchased and'product m'is purchased, so the position of the second column of the n-th row corresponding to'product 2'is "0." It has a status value of ", and may have a status value of "1" at the position of the m-th column of the n-th row corresponding to'product m'.

Here, the purchase data generation unit 300 sets each state value according to whether or not an existing customer (i) for a given product is purchased as one column, and each of the respective states according to whether or not a new customer (a) for the corresponding product is purchased. After constructing the predetermined second matrix M2 having the status value as one row, the purchase frequency (Fp) in the case where the existing customer (i) and the new customer (a) respectively have a predetermined status value and correspond to it May generate the second purchase data D ₂ matched to the corresponding row and column positions.

At this time, the status value is a binary form of a 1-bit method, and the first status value ("0" or "1") for whether or not a new customer (a) corresponding to a predetermined row is purchased and an existing one corresponding to a predetermined column It includes a second state value ("0" or "1") of whether or not the customer i is purchased, and the first state value corresponding to the corresponding row and column, respectively, at a position corresponding to any one specific row and column, and The values sequentially listing the second status values, for example, “00”, “01”, “10”, and “11” may be matched.

Table 2 below shows an example of the second purchase data D generated in a predetermined matrix configuration corresponding to the above-described second matrix M2.

For example, referring to Table 2 above, in the second purchase data D, the first state value when'new customer (a)' does not purchase a predetermined product becomes '0 (non-purchase)' and'existing customer' When (i)' purchases the product, the second state value becomes '1 (purchase)', and the first state value ('0') and the second state value ('1') The frequency for the corresponding row and column is "10".

In addition, the first state value when'new customer (a)' purchases the product becomes '1 (purchase)', and the second state value when'existing customer (i)' purchases the product. When is 1 (purchase), the frequency for rows and columns corresponding to each of the first state value ('1') and the second state value ('1') is "11".

In this case, regardless of whether'new customer (a)' is purchased or not, the sum of the status values when'existing customer (i)' purchases the product corresponds to the sum of "10" and "11" Can have

The recommendation score calculating unit 500 is based on the result of applying a positive or negative Pearson correlation to the purchase data D generated by the purchase data generation unit 300, the possibility of purchasing a new customer a for a given product The recommended score (P _ak ) corresponding to is calculated.

Here, the recommendation score calculating unit 500 is positive when the existing customer (i) purchases the product (v _ij =1) and when the existing customer (i) does not purchase the product (v _ij = 0), respectively. The recommendation score P _ak may be calculated by reflecting all of the similar weighting values (w(a,i), hereinafter referred to as'w') based on the results of applying the Pearson correlation and negative Pearson correlation.

In this case, the Pearson correlation corresponds to a kind of statistical analysis method used to test a linear correlation between two continuous variables, for example, a correlation coefficient is derived between "-1" to "+1". When the coefficient has a positive value close to "+1", it is determined that there is a positive linear correlation between the two variables, and the correlation coefficient has a negative value close to "-1" It can be judged that there is a negative linear relationship between the two variables.

)의 제1 차분값(dv₁)과, 기존 고객(i)의 복수 개의 상품 중 어느 하나의 특정 상품의 구매 여부에 대한 제2 상태값(v_ij)과 상품별 구매 여부에 대한 제2 평균 상태값(

)의 제2 차분값(dv₂)을 산출한 후, 복수 개의 상품 각각에 대하여 상기 제1 차분값(dv₁)과 상기 제2 차분값(dv₂)을 곱해서 합한 값을 분자로 하고, 복수 개의 상품 각각에 대한 상기 제1 차분값(dv₁)의 제곱합과 상기 제2 차분값(dv₂)의 제곱합의 곱의 제곱근값을 분모로 하여 유사가중치값(w)을 산출할 수 있다.Here, the recommendation score calculating unit 500, in calculating the above-described similar weight value (w), a first state value (v _aj ) of whether or not to purchase any one of a plurality of products of the new customer (a) ) And the first average status value for purchase by product (

)'S first difference value (dv ₁ ), and a second state value (v _ij ) of whether or not to purchase one specific product among a plurality of products of the existing customer (i) and a second average state of whether to purchase by product value(

) A second difference value (dv ₂₎ then calculated, and the for each of a plurality of product sum by multiplying said first difference value (dv ₁₎ and the second difference value (dv ₂₎ values in the molecule, a plurality of The pseudo-weight value w may be calculated by using the square root of the product of the square sum of the first difference value dv ₁ and the square sum of the second difference value dv ₂ for each of the products.

In this regard, referring to FIG. 2, when the total number of product information 101 pre-stored in the storage unit 100 is'm' and the total number of existing customer information 102 is'n','i The area'xj' is the matrix data for the state value corresponding to the j-th product purchased by the existing customer (i), and the area'a×j' is the state value corresponding to the j-th product purchased by the new customer (a). It may be matrix data for.

At this time, j may correspond to any one of natural numbers corresponding to 1 to m, and i may correspond to any one of natural numbers corresponding to 1 to n.

In this case,'v _ij 'is a value corresponding to _the'i ×j' area, and indicates a status value of whether to purchase a product corresponding to the j-th order among m products of the existing customer (i), _{and'v aj} 'Is a value corresponding to the'a×j' area, and may indicate a status value of whether to purchase a product corresponding to the j-th order among m products of the new customer a.

), 제2 상태값(v_ij) 및 제2 평균 상태값(

)을 이용하여 수식으로 나타낸 것이다.In Equation 1 below, the above-described pseudo-weighted value w is the first state value v _aj and the first average state value (

), the second state value (v _ij ) and the second average state value (

).

)과 제2 평균 상태값(

)을 각각 수식으로 나타낸 것이다.In addition, Equation 2 and Equation 3 below are the first average state values (

) And the second average state value (

) Respectively.

Here,'m' denotes the total number of products, _{and'v aj} 'denotes a first state value of whether to purchase a product corresponding to the j-th order among m products of the new customer a. _ij 'represents a second status value of whether to purchase a product corresponding to the j-th order among m products of the existing customer i.

Table 3 below summarizes the types of arithmetic operations corresponding to each combination according to the sign of the second state value v _ij and the similar weight value w.

At this time, when the arithmetic operation is'add (+)', when calculating the recommendation score P _ak , the similar weight value w corresponding to the corresponding combination is summed, and the arithmetic operation is'subtract. In the case of (-))', when calculating the recommendation score P _ak , the similar weight value w corresponding to the combination can be subtracted.

For example, referring to Table 3, when the second status value v _ij is “1”, that is, when the existing customer i purchases the j-th product, if the first pseudo-weight value w is If it has a positive value (w> 0), the first pseudo-weight value (w) is summed (+) when calculating the recommended score (P _ak ), and conversely, the second pseudo-weight value (w) is negative (w < If it has a value of 0), when calculating the recommendation score P _ak , the second pseudo-weighted value w may be subtracted (-).

Also, when the second status value v _ij is “0”, that is, when the existing customer i has not purchased the j-th product, the first pseudo-weighted value w is positive (w> 0). If the value of, the first pseudo-weight value (w) is subtracted (-) when calculating the recommended score (P _ak ), the second pseudo-weight value (w) is negative (w <0). If it is, the second pseudo-weighted value w may be summed (+) when calculating the recommendation score P _ak .

Specifically, the recommendation score calculating unit 500, the first similar weight value (w (a, i) _1, based on the result of applying a positive Pearson correlation when the existing customer (i) purchased a predetermined product, _1, or less w ₁ ) and the second pseudo-weighted value (w(a,i) _2, less than w ₂ ) based on the result of applying a negative Pearson correlation when the existing customer (i) has not purchased the product. ), and a third pseudo-weighted value (w(a,i) _3, hereinafter referred to as w ₃ ) based on the result of applying a negative Pearson correlation when an existing customer (i) purchased the product, and an existing customer When (i) does not purchase the product, after calculating the fourth pseudo-weighted value (w(a,i) _4, hereinafter referred to as w ₄ ) based on the result of applying the positive Pearson correlation, respectively, 1 Addition or subtraction depending on whether the pseudo-weight value (w ₁ ), the second pseudo-weight value (w ₂ ), the third pseudo-weight value (w ₃ ), and the fourth pseudo-weight value (w ₄ ) are selected respectively Based on one result, a recommendation score P _ak may be calculated.

Equation 4 below shows the formula for the recommended score (P _ak ) considering the correlation between the purchase of the existing customer (i) and the new customer (a) and Pearson, and the equation (5) is a predetermined gain constant. The equation for k _a'is expressed using the pseudo-weight value (w).

Here, w(a,i) ₁ is the first pseudo-weighted value, w(a,i) ₂ is the second pseudo-weighted value, w(a,i) ₃ is the third pseudo-weighted value, and w(a ,i) ₄ is the fourth pseudo-weighted value.

In addition, l ₁ indicates whether a first pseudo-weight value is selected, l ₂ indicates whether a second pseudo-weight value is selected, l ₃ indicates whether a third pseudo-weight value is selected, and l ₄ The fourth pseudo-weighted value is selected, and the above-described l ₁ to l ₄ may have a value of '1' when all are selected, and a value of '0' if not selected.

At this time, in Equation 4, the first pseudo-weight value (w ₁ ), the second pseudo-weight value (w ₂ ), the third pseudo-weight value (w ₃ ) and the fourth pseudo-weight value (w ₄ ) attached to each The arithmetic operation code ('+'or'-') may be based on Table 3 above.

In this case, the first pseudo-weighted value w ₁ and the second pseudo-weighted value w ₂ corresponding to the case where the existing customer i purchases the j-th product are positive (w> 0) and negative, respectively. As it has a value of (w <0), the first pseudo-weighted value (w) is summed (+) when calculating the recommended score (P _ak ) and the second pseudo-weighted value (w) is the recommended score (P _ak ) It can be subtracted (-) when calculating.

In addition, the third pseudo-weighted value w ₃ and the fourth pseudo-weighted value w ₄ corresponding to the case where the existing customer i does not purchase the j-th product are positive (w> 0) and negative, respectively. As it has a value of (w <0), the third pseudo-weighted value (w ₃ ) is subtracted (-) when calculating the recommended score (P _ak ) and the fourth pseudo-weighted value (w ₄ ) is the recommended score (P) _When calculating _ak ), it can be summed (+).

On the other hand, the recommendation score calculation unit 500, the first similar weight value (w ₁ ) and corresponding to the existing customer (i) and new customer (a) of each state value (v _ij , v _aj ) frequency of purchase The first value multiplied by the first contribution factor based on (Fp), the second pseudo-weighted value (w ₂ ), and the corresponding state values (v _ij ,v _aj ) of each of the existing customers (i) and new customers (a). ), the second value multiplied by the second contribution factor based on the purchase frequency (Fp), and the third pseudo-weighted value (w ₃ ) and the corresponding status values of the existing customers (i) and new customers (a) ( The third value multiplied by the third contribution factor based on the purchase frequency (Fp) according to v _ij ,v _aj ), and the fourth pseudo-weighted value (w ₄ ) and the corresponding existing customer (i) and new customer (a) After calculating the fourth value multiplied by the fourth contribution coefficient based on the purchase frequency Fp according to the respective state values v _ij and v _aj , the first value, the second value, the third value, and The recommendation score _Pak may be calculated based on the sum or subtraction result depending on whether each of the fourth values is selected.

Equation 6 below summarizes the formulas for the recommendation score (P _ak ) considering the correlation between the existing customers (i) and the new customers (a) and the purchase frequency and the contribution and Pearson correlation.

In addition, Equation 7 to Equation 10 below are a summary of equations for each of the first value, the second value, the third value, and the fourth value.

은 제1 기여도 계수를 나타내고,

은 제2 기여도 계수를 나타내고,

은 제3 기여도 계수를 나타내고,

은 제4 기여도 계수를 나타낸다.here,

Denotes a first contribution coefficient,

Denotes the second contribution coefficient,

Denotes the third contribution coefficient,

Indicates the fourth contribution coefficient.

At this time, "00" is the location corresponding to (0,0) where the status value of the existing customer (i) is '0 (non-purchase)' and the status value of the new customer (a) is '0 (non-purchase)'. "11" is the location corresponding to the case where the status value of the existing customer (i) is '1 (purchase)' and the status value of the new customer (a) is '1 (purchase)'. ,1) indicates the frequency of purchase, and "10" corresponds to the case where the status value of the existing customer (i) is '1 (purchase)' and the status value of the new customer (a) is '0 (non-purchase)'. Represents the frequency of purchase for location (1,0), and "01" means that the status value of the existing customer (i) is '0 (non-purchase)' and the status value of the new customer (a) is '1 (purchase)'. It may indicate the purchase frequency for (0,1), which is a corresponding location.

Here, as described above, w(a,i) ₁ is the first pseudo-weighted value, w(a,i) ₂ is the second pseudo-weighted value, and w(a,i) ₃ is the third pseudo-weighted value And w(a,i) ₄ is the fourth pseudo-weighted value.

In addition, l ₁ indicates whether the sum of the first pseudo-weighted values is selected, l ₂ indicates whether the sum of the second pseudo-weighted values is selected, l ₃ indicates whether the sum of the third pseudo-weighted values is selected, l ₄ indicates whether the sum of the fourth pseudo-weighted values is selected, and l ₁ to l ₄ described above may have a value of '1' when all are selected and a value of '0' if not selected.

Hereinafter, an embodiment in which the recommendation score is calculated using the data analysis device related to product purchase using the collaborative filtering algorithm according to the present invention will be described.

Here, the product purchase related data analysis apparatus according to the present invention may include all of the above-described components, that is, the storage unit 100, the purchase data generation unit 300, and the recommendation score calculation unit 500.

Table 4 below shows the customer number, product number, number of purchases, and purchase rate corresponding to each area based on the plurality of areas shown in FIG. 2 by the recommendation score calculating unit 500.

In addition, Table 5 below applies the data based on Table 4 to Equation (6) to calculate the recommended score (P _ak ), but the first pseudo-weighted value (w ₁ ) based on the case where the existing customer (i) purchased ) And the second similar weight value (w ₂ ), the first recommendation score calculated by reflecting only, and the first similar weight value (w ₁ ) and the second similar weight value (w) based on the purchase by the existing customer (i) _{In addition to 2} ), the second recommended score calculated by reflecting both the third similar weight value (w ₃ ) and the fourth similar weight value (w ₄ ) based on the case where the existing customer (i) has not purchased by product (N= 1,2,3,4,5).

Here, the first similar weight value (w ₁ ) is based on the result of applying a positive Pearson correlation when the existing customer (i) purchases a predetermined product, and the second similar weight value (w ₂ ) is the existing customer ( i) is based on the result of applying the negative Pearson correlation when the product is not purchased, and the third pseudo-weighted value (w ₃ ) is negative Pearson's correlation when the existing customer (i) purchases the product. It is based on the result of applying the relationship, and the fourth pseudo-weighted value w ₄ may be based on the result of applying the positive Pearson correlation when the existing customer i has not purchased the product.

For example, when comparing the first recommendation accuracy and the second recommendation accuracy with reference to Table 5, when one product is recommended, two products are recommended, three products are recommended, and four products are recommended In one case, since it can be confirmed that the second recommendation accuracy is higher than the first recommendation accuracy, it can be seen that the recommended performance is highest when considering both the case where the existing customer i purchased and the case where the existing recommendation is not purchased. .

3 is a flowchart illustrating a method for analyzing data related to product purchase using a collaborative filter algorithm according to an embodiment of the present invention.

Hereinafter, a method for analyzing data related to product purchase using a collaborative filter algorithm according to an embodiment of the present invention will be described with reference to FIG. 3 and the aforementioned drawings.

First, product information for a plurality of products is stored (S100).

Here, the product information 101 may include information about a product name and a product type of each of a plurality of products.

At this time, the product corresponding to the product information 101 is not limited to an object of a type that can be purchased and owned, and an intangible value that can be viewed or operated through a purchase act that pays a predetermined ticket value. It may also include cultural products (eg, movies, performances, concerts, etc.).

Next, existing customer information about an existing customer who purchased at least one of a plurality of products previously stored in step S100 for a predetermined period is stored (S200).

Here, the existing customer information 102 may include information on a customer name and a product name corresponding to the product purchased by the customer.

Next, purchase data D based on whether the existing customer i purchases by product (Wp) and purchase frequency (Fp) for each of the plurality of products is generated (S300).

Here, the step (S300) of generating purchase data D is a predetermined first matrix in which each of a plurality of products pre-stored in step S100 is a column and each of a plurality of customers pre-stored in step S200 is a row. After configuring (M1), it may include a step (S320) of generating the first purchase data (D ₁ ) by matching the status values in binary form for whether or not a given customer purchases by product to corresponding row and column positions. have.

At this time, the status value is a 1-bit binary format, and if a customer corresponding to a corresponding row purchases a product corresponding to a corresponding column, it has a value of '1' and a value of '0' otherwise. Can.

In addition, in the step of generating purchase data D (S300), each status value according to whether or not an existing customer i for a given product is purchased is set as one column, and a new customer (a) is purchased for the corresponding product. When a predetermined second matrix (M2) having each status value as one row according to whether or not is configured, and when the existing customer (i) and the new customer (a) each have a predetermined status value, corresponding to this The method may further include generating the second purchase data D ₂ matching the purchase frequency Fp to the corresponding row and column positions (S340 ).

At this time, the status value is a binary form of a 1-bit method, and the first status value ("0" or "1") for whether or not a new customer (a) corresponding to a predetermined row is purchased and an existing one corresponding to a predetermined column It includes a second state value ("0" or "1") of whether or not the customer i is purchased, and the first state value corresponding to the corresponding row and column, respectively, at a position corresponding to any one specific row and column, and The values sequentially listing the second status values, for example, “00”, “01”, “10”, and “11” may be matched.

Next, based on the result of applying the positive or negative Pearson correlation to the purchase data, a recommendation score (P _ak ) corresponding to a new customer's purchase possibility for a predetermined product is calculated (S400).

Here, in the step of calculating the recommendation score P _ak (S400), when the existing customer (i) purchases the product (v _ij =1) and when the existing customer (i) does not purchase the product (v _ij) The recommended score (P _ak ) is reflected by reflecting all the similar weight values (w(a,i), hereinafter referred to as'w') based on the result of applying the positive Pearson correlation and the negative Pearson correlation to =0), respectively. Can be calculated.

In this case, the Pearson correlation corresponds to a kind of statistical analysis method used to test a linear correlation between two continuous variables, for example, a correlation coefficient is derived between "-1" to "+1". When the coefficient has a positive value close to "+1", it is determined that there is a positive linear correlation between the two variables, and the correlation coefficient has a negative value close to "-1" It can be judged that there is a negative linear relationship between the two variables.

)의 제1 차분값(dv₁)과, 기존 고객(i)의 복수 개의 상품 중 어느 하나의 특정 상품의 구매 여부에 대한 제2 상태값(v_ij)과 상품별 구매 여부에 대한 제2 평균 상태값(

)의 제2 차분값(dv₂)을 산출한 후, 복수 개의 상품 각각에 대하여 상기 제1 차분값(dv₁)과 상기 제2 차분값(dv₂)을 곱해서 합한 값을 분자로 하고, 복수 개의 상품 각각에 대한 상기 제1 차분값(dv₁)의 제곱합과 상기 제2 차분값(dv₂)의 제곱합의 곱의 제곱근값을 분모로 하여 유사가중치값(w)을 산출할 수 있다.Here, in the step of calculating the recommendation score (P _ak ) (S400), in calculating the above-mentioned similar weight value (w), whether or not to purchase any one of a plurality of products of the new customer (a) The first state value (v _aj ) and the first average state value for purchase by product (

)'S first difference value (dv ₁ ), and a second state value (v _ij ) of whether or not to purchase one specific product among a plurality of products of the existing customer (i) and a second average state of whether to purchase by product value(

) A second difference value (dv ₂₎ then calculated, and the for each of a plurality of product sum by multiplying said first difference value (dv ₁₎ and the second difference value (dv ₂₎ values in the molecule, a plurality of The pseudo-weight value w may be calculated by using the square root of the product of the square sum of the first difference value dv ₁ and the square sum of the second difference value dv ₂ for each of the products.

In this regard, referring to FIG. 2, when the total number of product information 101 previously stored in steps S100 and S200 is'm' and the total number of existing customer information 102 is'n', ' The i×j' area is matrix data for the state value corresponding to the j-th product purchased by the existing customer (i), and the'a×j' area corresponds to the j-th product purchased by the new customer (a). It may be matrix data for a value.

At this time, j may correspond to any one of natural numbers corresponding to 1 to m, and i may correspond to any one of natural numbers corresponding to 1 to n.

In this case,'v _ij 'is a value corresponding to _the'i ×j' area, and indicates a status value of whether to purchase a product corresponding to the j-th order among m products of the existing customer (i), _{and'v aj} 'Is a value corresponding to the'a×j' area, and may indicate a status value of whether to purchase a product corresponding to the j-th order among m products of the new customer a.

Specifically, in the step of calculating the recommendation score P _ak (S400), when the existing customer i purchases a predetermined product, the first pseudo-weighted value based on the result of applying the positive Pearson correlation (w(a ,i) _1, hereinafter referred to as w ₁ ) and the second pseudo-weighted value (w(a,i) ₂ ) based on the result of applying a negative Pearson correlation when the existing customer (i) has not purchased the product. _, Hereinafter referred to as w ₂ ) and a third pseudo-weighted value (w(a,i) _3, less than w ₃ ) based on the result of applying a negative Pearson correlation when an existing customer (i) purchases the product ) And the fourth pseudo-weighted value (w(a,i) _4, hereinafter referred to as w ₄ ) based on the result of applying a positive Pearson correlation when the existing customer (i) has not purchased the product. After calculation, selection of each of the first pseudo-weighted value w ₁ , the second pseudo-weighted value w ₂ , the third pseudo-weighted value w ₃ and the fourth pseudo-weighted value w ₄ The recommendation score _Pak may be calculated based on the summation or subtraction result depending on whether or not.

In this regard, Equation 4 corresponding to the equation for the recommendation score P _ak considering the correlation between the purchase of the existing customer (i) and the new customer (a) and Pearson may be used, and the equation 4 The detailed contents will be omitted as already described.

In this case, the first pseudo-weighted value w ₁ and the second pseudo-weighted value w ₂ corresponding to the case where the existing customer i purchases the j-th product are positive (w> 0) and negative, respectively. As it has a value of (w <0), the first pseudo-weighted value (w) is summed (+) when calculating the recommended score (P _ak ) and the second pseudo-weighted value (w) is the recommended score (P _ak ) It can be subtracted (-) when calculating.

On the other hand, in the step of calculating the recommendation score P _ak (S400), the first pseudo-weighted value w ₁ and the corresponding state values of the existing customers i and new customers a, respectively v _ij ,v _aj ), the first value (p ₁ ) multiplied by the first contribution factor based on the purchase frequency (Fp), the second similar weight value (w ₂ ), and the corresponding existing customer (i) and new customer (a) The second value (p ₂ ) multiplied by the second contribution factor based on the purchase frequency (Fp) according to each state value (v _ij ,v _aj ), and the third similar weight value (w ₃ ) and the corresponding existing customer (i) and a new customer (a) a third value (p ₃ ) multiplied by a third contribution factor based on the purchase frequency (Fp) according to the respective state values (v _ij ,v _aj ) and a fourth pseudo-weighted value ( w ₄ ) and the fourth value (p ₄₎ multiplied by the fourth contribution factor based on the purchase frequency (Fp) according to the status values (v _ij ,v _aj ) of the existing customers (i) and new customers (a) corresponding thereto. ), and then summed according to whether each of the first value (p ₁ ), the second value (p ₂ ), the third value (p ₃ ), and the fourth value (p ₄ ) is selected or The recommendation score _Pak may be calculated based on the subtracted result.

In this regard, Equation (6) corresponding to the formula for the recommendation score (P _ak ) considering the contribution and Pearson correlation according to the purchase frequency and purchase frequency of the existing customer (i) and the new customer (a) may be used, The detailed description of Equation 6 will be omitted as already described.

Accordingly, according to the present invention described above, when calculating the recommendation score for a given product, the existing customer purchases the product and otherwise, considers the contribution ratio of data for each Pearson correlation together to improve the recommendation performance There is an effect that can be made.

In addition, according to the present invention, it is possible to apply universally when calculating recommendation scores for various types of products through a simple algorithm that selects the optimal combination according to the expected purchaseability of a new customer based on the purchase frequency and purchase frequency of an existing customer. There is an advantage.

As described above, the present invention has been described in detail through preferred embodiments, but the present invention is not limited thereto and may be variously implemented within the scope of the claims.

In particular, the foregoing has described the features and technical strengths of the present invention somewhat broadly to better understand the claims of the invention, which will be described later, so that the concept and specific embodiments of the present invention described above perform similar purposes to the present invention. It should be appreciated by those skilled in the art that it can be used immediately as the basis for designing or modifying other shapes for.

In addition, the above-described embodiments are only one embodiment according to the present invention, and can be implemented in various modified and changed forms within the scope of the technical spirit of the present invention by those skilled in the art. You will understand. Therefore, the disclosed embodiments are to be considered in terms of explanation rather than limitation, and such various modifications and changes are also indicated in the claims of the present invention described above as belonging to the scope of the technical spirit of the present invention, and equivalent scope thereof. All differences within should be construed as being included in the present invention.

100: storage
300: purchase data generation unit
500: recommended score calculation unit

Claims (10)

Storing product information for a plurality of products;
Storing existing customer information about an existing customer who purchased at least one of the plurality of products for a predetermined period;
Generating purchase data based on whether each of the plurality of products is purchased for each product by the existing customer and the frequency of purchase; And
And calculating a recommendation score corresponding to a new customer's purchase possibility for a given product based on a result of applying a positive or negative Pearson correlation to the purchase data. How to recommend purchasing. According to claim 1,
The step of calculating the recommended score,
The recommendation score reflects all the similar weights based on the result of applying the positive Pearson correlation and the negative Pearson correlation, respectively, when the existing customer purchases the product and when the existing customer does not purchase the product Product recommendation-related recommendation method using a collaborative filter algorithm, characterized in that for calculating. According to claim 2,
The step of calculating the recommended score,
A first pseudo-weighted value based on a result of applying a positive Pearson correlation when the existing customer purchases the product, and a result based on applying a negative Pearson correlation when the existing customer does not purchase the product. A second similar weight value, a third similar weight value based on a result of applying a negative Pearson correlation when the product is purchased by the existing customer, and a positive Pearson when the existing customer does not purchase the product After calculating the fourth similar weight value based on the result of applying the correlation,
The recommendation score is calculated based on the sum or subtraction result according to whether each of the first pseudo-weight value, the second pseudo-weight value, the third pseudo-weight value, and the fourth pseudo-weight value is selected. Product purchasing recommendation method using collaborative filter algorithm. According to claim 3,
The step of calculating the recommended score,
A first value multiplied by a first contribution factor based on a purchase frequency according to a status value for each purchase of the first pseudo-weighted value and the corresponding existing customer and the new customer;
A second value multiplied by a second contribution factor based on a purchase frequency according to a status value of whether the second pseudo-weighted value corresponds to the purchase status of each of the existing customers and the new customers corresponding thereto;
A third value multiplied by the third similarity value and a third contribution factor based on the purchase frequency according to the status value of each of the existing customers and the new customers corresponding to the purchase, and
After calculating the fourth similarity value and the fourth value multiplied by the fourth contribution coefficient based on the purchase frequency according to the status value for each purchase of each of the existing customers and the new customers corresponding thereto,
A product using a collaborative filter algorithm, characterized in that the recommendation score is calculated based on a sum or subtraction result depending on whether each of the first value, the second value, the third value, and the fourth value is selected. How to recommend purchasing. According to claim 2,
The step of calculating the recommended score,
Among the plurality of products of the existing customer and a first difference value between a first state value for whether to purchase a specific product among the plurality of products of the new customer and a first average state value for whether to purchase by product, etc. After calculating the second difference value between the second status value for whether or not to purchase one specific product and the second average status value for whether or not to purchase each product,
For each of the plurality of products, a product obtained by multiplying the first difference value and the second difference value is a numerator, and the sum of squares of the first difference value and the second difference value for each of the plurality of products A product-related recommendation method using a collaborative filter algorithm, characterized in that the square weight value of a product is used as a denominator to calculate the similar weight value. A storage unit for storing product information for a plurality of products and existing customer information for existing customers who have purchased at least one of the plurality of products for a predetermined period;
A purchase data generation unit for generating purchase data based on whether each of the plurality of products is purchased by each of the existing customers and purchased frequency; And
And a recommendation score calculating unit that calculates a recommendation score corresponding to a new customer's purchase possibility for a given product based on a result of applying a positive or negative Pearson correlation to the purchase data. Product recommendation system related to product purchase. The method of claim 6,
The recommended score calculation unit,
The recommendation score reflects all the similar weights based on the result of applying the positive Pearson correlation and the negative Pearson correlation, respectively, when the existing customer purchases the product and when the existing customer does not purchase the product Product purchasing-related recommendation system using a collaborative filter algorithm, characterized in that to calculate. The method of claim 7,
The recommended score calculation unit,
A first pseudo-weighted value based on the result of applying a positive Pearson correlation when the existing customer purchased, and a second pseudo-weighted value based on the result of applying a negative Pearson correlation if the existing customer did not purchase , A third pseudo-weighted value based on the result of applying a negative Pearson correlation when the existing customer purchased, and a fourth pseudo-weighted value based on the result of applying a positive Pearson correlation if the existing customer did not purchase After calculating each,
The recommendation score is calculated based on the sum or subtraction result according to whether each of the first pseudo-weight value, the second pseudo-weight value, the third pseudo-weight value, and the fourth pseudo-weight value is selected. Product purchasing related recommendation system using collaborative filter algorithm. The method of claim 8,
The recommended score calculation unit,
A first value multiplied by a first contribution factor based on a frequency of purchase according to the first pseudo-weighted value and the respective state values of the existing customers and the new customers corresponding thereto;
A second value multiplied by the second similarity weight value and a second contribution factor based on the frequency of purchase according to the status value of each of the existing and corresponding new customers;
A third value multiplied by the third similarity value and a third contribution factor based on the frequency of purchase according to the status values of each of the existing customers and the new customers corresponding thereto;
After calculating a fourth value multiplied by a fourth contribution factor based on the frequency of purchase according to each of the fourth pseudo-weighted value and the corresponding state values of the existing customers and the new customers, respectively,
A product using a collaborative filter algorithm, characterized in that the recommendation score is calculated based on a result of adding or subtracting according to whether each of the first value, the second value, the third value, and the fourth value is selected. Purchase-related recommendation system. The method of claim 7,
The recommended score calculation unit,
Among the plurality of products of the existing customer and a first difference value between a first state value for whether to purchase a specific product among the plurality of products of the new customer and a first average status value for whether to purchase by product, etc. After calculating the second difference value between the second status value for whether or not one particular product is purchased and the second average status value for whether each product is purchased,
For each of the plurality of products, a product obtained by multiplying the first difference value and the second difference value is a numerator, and the sum of squares of the first difference value and the second difference value for each of the plurality of products A product recommendation system related to product purchase using a collaborative filter algorithm, characterized in that the square weight value of a product is used as a denominator to calculate the similar weight value.

Images