KR102157026B1 - System and method for products recommendation service, and apparatus applied to the same - Google Patents

Abstract

The present invention discloses a product recommendation service system, a method thereof, and an apparatus applied thereto. That is, among at least two association rules, the best association product for each association rule, which is the association product with the highest association degree, is selected from a specific association rule, and in association rules different from the specific association rule among the at least two association rules. It is determined whether there is a relative optimal related product, which is a related product whose degree of association in the other association rules is larger than that of the optimal associated product, and the optimal associated product for each association rule is determined as a recommended product, or By determining both the optimally associated product and the relative optimally associated product as the recommended product, a related product having a degree of association that can satisfy all of the association rules can be determined as the recommended product.

Description

Product recommendation service system, its method, and apparatus applied thereto {SYSTEM AND METHOD FOR PRODUCTS RECOMMENDATION SERVICE, AND APPARATUS APPLIED TO THE SAME}

The present invention relates to a method for determining as a recommended product a related product having a degree of association that can satisfy all of the plurality of association rules when determining a recommended product by applying a plurality of association rules.

Internet or mobile portal sites provide a product recommendation service that recommends related products through a service page corresponding to a first page or a detailed page related to a specific reference product when a buyer (or a customer who has visited for purchase) visits.

Here, as a method for recommending related products, association rule mining, which is a technique for recommending products with high possibility of simultaneous purchase or click possibility for each product, is applied using purchase records of multiple users.

In applying such association rule mining, it is more important than anything else to decide which association rule to select and provide among a number of association rules for recommending related products.

In this regard, in the case of the existing method to which association rule mining is applied, only one of association rules such as support, confidence, and lift is selected as a criterion.

Through this, among a plurality of related products identified based on the association rule selected as a criterion, a related product having a degree of association with the determined reference product equal to or greater than a preset threshold is determined as a recommended product.

However, since the association rule mining is based on the actual purchase history, the result may vary according to the period of the purchase history and the distribution of product consumption (download/click) in the actual purchase history.

Therefore, in the case of the existing method of selecting and applying only one association rule or selecting a related product based on a preset threshold, the variable characteristics of the above-described association rule mining cannot be actively reflected. It can even lead to a decrease in the satisfaction of the buyer's service.

The present invention was created in view of the above circumstances, and an object of the present invention is to provide an optimal associated product for each association rule, which is a related product whose degree of association is the largest in a specific association rule among at least two association rules. Screening, and determining whether there is a relative optimal associated product, which is a related product whose degree of association in the other association rule is larger than the optimal associated product in association rules different from the specific association rule among the at least two association rules, and , By determining the best associated product for each association rule as a recommended product, or determining both the best associated product for each association rule and the relative optimal associated product as the recommended product, having a degree of association that can satisfy all of the association rules It is in determining related products as recommended products.

A service apparatus according to a first aspect of the present invention for achieving the above object comprises: a confirmation unit for checking related products having a correlation relationship with a reference product in relation to at least two association rules; A calculation unit for calculating a degree of association with the reference product based on each of the at least two association rules for each of the related products; And when the related products with the greatest degree of association are the same in each of the at least two association rules, a related product corresponding to the largest calculated value is determined as a recommended product, and in each of the at least two association rules, the If the related products with the greatest degree of relevance are not the same, the relevant product with the highest degree of relevance in any one of the at least two relevance rules is different from the one of the relevance rules. It characterized in that it comprises a determination unit for determining the related product with the highest correlation in the rule as a recommended product.

More specifically, when there are at least two related products having the same size of the association in any one of the at least two association rules, the determination unit may include, among the at least two related products, one of the association rules. In the association rule different from that, a related product for which the degree of association is largely calculated is determined as the recommended product.

More specifically, the recommended product is determined at each of at least two selection levels for selecting the optimal associated product for each association rule, and the determination unit determines the recommended product at a specific selection level among the at least two selection levels. Then, at a lower selection level adjacent to the specific selection level, the best related products for each of the related rules are selected for the remaining related products excluding the related products determined as the recommended products at the specific selection level.

More specifically, the service device further includes an update unit for accumulating and updating the number of the recommended products determined at each of the at least two selection levels, and the determination unit is accumulated to a specific selection level among the at least two selection levels. When the updated number of recommended products is less than the set number, the recommended products are determined at a lower screening level adjacent to the specific screening level.

More specifically, the service device further includes a valuation unit for setting a recommendation degree for the recommended product determined at each of the at least two selection levels, and the valuation unit is the same for the recommended product determined at the same selection level. A recommendation degree of a size is determined, and the recommendation degree for a recommended product determined at a higher selection level is set to be larger than a recommendation product determined at a neighboring lower selection level.

More specifically, at least one of the at least two association rules is a merge association rule in which at least two similar association rules with correlation exist, and the calculation unit is a recommendation determined based on the at least two similar association rules. The recommendation degree determined by the planning unit for the product is calculated as the association degree for each of the related products corresponding to the recommended product in the merger association rule.

A product recommendation service method according to a second aspect of the present invention for achieving the above object comprises: a requesting step of requesting, by a terminal device, a recommended product related to a reference product; A confirmation step of confirming, by the service device, related products having a related relationship with the reference product in relation to at least two related rules; A calculation step of calculating, by the service device, a degree of association with the reference product based on each of the at least two association rules for each of the related products; And the service device determines a related product corresponding to the largest calculated value as a recommended product, when the related product with the greatest degree of association is the same in each of the at least two association rules, and the at least two associations If the related products with the greatest degree of correlation are not the same in each of the rules, the related product with the largest degree of correlation in any one of the at least two association rules, and any one of the associations And a determining step of determining as a recommended product a related product having the greatest degree of association calculated in a related rule different from the rule.

In order to achieve the above object, a method of operating a service device according to a third aspect of the present invention includes: a confirmation step of confirming related products having a correlation relationship with a reference product in relation to at least two association rules; A calculating step of calculating a degree of association with the reference product based on each of the at least two association rules for each of the related products; And when the related products with the greatest degree of association are the same in each of the at least two association rules, a related product corresponding to the largest calculated value is determined as a recommended product, and in each of the at least two association rules, the If the related products with the greatest degree of relevance are not the same, the relevant product with the highest degree of relevance in any one of the at least two relevance rules is different from the one of the relevance rules. And a determining step of determining a related product for which the degree of association is the largest in the rule as a recommended product.

More specifically, in the case where there are at least two related products having the same degree of association in any one of the at least two association rules, the determining step is, among the at least two related products, any one of the related products. In a related rule different from the rule, a related product for which the degree of correlation is largely calculated is determined as the recommended product.

More specifically, the recommended product is determined at each of at least two screening levels for selecting the best associated product for each association rule, and the determining step includes: the recommended product at a specific screening level among the at least two screening levels. When determined, at a lower selection level adjacent to the specific selection level, the best related products for each related rule are selected for the remaining related products excluding the related products determined as the recommended products at the specific selection level. .

More specifically, the method further includes an updating step of accumulating and updating the number of recommended products determined at each of the at least two selection levels, and the determining step includes: up to a specific selection level among the at least two selection levels. When the accumulated and updated number of recommended products is less than the set number, the recommended products are determined at a lower selection level adjacent to the specific selection level.

More specifically, the method further includes a determining step of determining a degree of recommendation for the recommended product determined at each of the at least two screening levels, wherein the determining step includes the same size for the recommended product determined at the same screening level. The recommendation level of is determined, and the recommendation level for the recommended product determined at the upper selection level is set higher than the recommended product determined at the neighboring lower selection level.

More specifically, at least one of the at least two association rules is a merge association rule in which at least two similar association rules with correlation exist, and the calculating step is determined based on the at least two similar association rules. And calculating the recommendation degree determined by the planning unit for the recommended product as the association degree for each of the related products corresponding to the recommended product in the merger association rule.

The computer-readable medium according to the fourth aspect of the present invention for achieving the above object comprises: a confirmation step of confirming related products having a correlation relationship with a reference product in relation to at least two association rules; A calculating step of calculating a degree of association with the reference product based on each of the at least two association rules for each of the related products; And when the related products with the greatest degree of association are the same in each of the at least two association rules, a related product corresponding to the largest calculated value is determined as a recommended product, and in each of the at least two association rules, the If the related products with the greatest degree of relevance are not the same, the relevant product with the highest degree of relevance in any one of the at least two relevance rules is different from the one of the relevance rules. It is characterized in that it includes instructions for performing a determination step of determining a related product with the greatest degree of correlation calculated as a recommended product in the rule.

A product recommendation service system according to a fifth aspect of the present invention for achieving the above object includes: a terminal device for requesting a recommended product related to a reference product; And, for each of the related products, a degree of association with the reference product based on each of the at least two association rules is calculated by checking the related products having a relationship with the reference product in relation to at least two association rules. And, in each of the at least two association rules, when the associated product with the greatest degree of association is the same, the associated product corresponding to the greatest calculated value is determined as a recommended product, and in each of the at least two association rules If the related products with the greatest degree of relevance are not the same, the relevant product with the highest degree of relevance in any one of the at least two association rules is different from the one of the association rules. It characterized in that it comprises a service device for determining as a recommended product a related product for which the degree of association is the largest in the association rule.

Accordingly, according to the product recommendation service system and method of the present invention, and the apparatus applied thereto, by proposing a method for determining a related product having a degree of association that can satisfy all of a plurality of association rules as a recommended product, the buyer It is possible to recommend more optimized products to customers, which increases service satisfaction and at the same time increases the profits of product sellers.

1 is a schematic configuration diagram of a product recommendation service system according to an embodiment of the present invention.
2 is a diagram for explaining a problem in a process of determining a recommended product according to an embodiment of the present invention.
3 is a schematic configuration diagram of a service device according to an embodiment of the present invention.
4 to 6 are diagrams for explaining a process of determining a recommended product according to an embodiment of the present invention.
7 and 8 are diagrams for explaining an association rule application structure according to an embodiment of the present invention.
9 is a schematic flowchart illustrating an operation flow in a product recommendation service system according to an embodiment of the present invention.
10 is a schematic flowchart illustrating an operation flow in a service device according to an embodiment of the present invention.

Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

1 is a diagram illustrating a product recommendation service system according to an embodiment of the present invention.

As shown in FIG. 1, a product recommendation service system according to an embodiment of the present invention includes a terminal device 100 for receiving a recommended product associated with a reference product from the service device 200, and a related product associated with the reference product. It has a configuration including a service device 200 that determines a recommended product from among and provides it to the terminal device 100.

Here, the terminal device 100 refers to a user device for accessing the service device 200 to check recommended products, and, for example, a smartphone, a personal computer, a notebook, and a tablet PC, etc. All devices that can be connected to the service device 200 may be included, but are not limited thereto.

The service device 200 is intended to refer to a server for providing recommended products to buyers who access the service page through the terminal device 100, and provides products to the purchaser who accesses the service page through the Internet or mobile web. It may be configured as a separate server for providing recommended products while selling or providing only recommended products.

Here, in the case of products sold to buyers, product categories can be classified into, for example, application products and content products according to their characteristics, and further, various product categories (eg, games, learning, etc.) according to the detailed characteristics of the product. Movies, dramas, and books, etc.).

On the other hand, according to an embodiment of the present invention, the service device 200 determines a recommended product from among related products that have a correlation relationship with a reference product in relation to at least two association rules and provides it to the terminal device 100. It provides product recommendation service.

For example, in the case of determining a recommended product by applying a confidence association rule and a lift association rule, check the related products that have a relationship with the reference product in relation to each association rule, It is a product with a high degree of relevance and is decided as a recommended product.

In addition, by arranging the related products based on the Confidence association rule, the top K related products with high degree of relevance (reliability) are determined as recommended products, or the related products based on the Lift association rule are sorted and the degree of relevance The top K related products with high (level of improvement) can be determined as recommended products.

Referring to FIG. 2, when the reliability association rule and the lift association rule are viewed as a two-dimensional plane with X and Y axes, respectively, the actual recommended product satisfies only a single association rule. In terms of the related rules, it can be seen that related products that are not highly related are determined as recommended products.

Here, in the case where the confidence association rule is the standard, related products with only high reliability but not high improvement are determined as recommended products, even when the lift association rule is the standard. The same is true.

As another example, if two users purchase product A and product B at the same time, and there is no such pattern for the other users, the confidence to buy B when buying A is 100 as shown in [Equation] below. Can be %.

[Equation 1]

P(A)= 2/M, P(A∩B)=2/M, confidence(A -> B)= P(A∩B)/P(A)= 1.0(=100%)

Therefore, if a recommended product is determined based only on the Confidence related rule, the likelihood that such a related product will be determined as a recommended product increases.

Of course, by applying an arbitrary threshold, only the pattern corresponding to the case where X or more people purchase the product, or the relational rules where the P(A) value or P(A∩B) value is more than a certain value. Based on the extraction of related products, it is difficult to sufficiently prevent the bias or shift of related products caused by setting the threshold.

In addition, since the threshold value itself is a value that continuously changes according to the purchase history of the purchaser, it is difficult to change it by performing an analysis every time, and in most cases, a service is performed while a predetermined threshold value is fixed.

Meanwhile, as a solution, there is a method of determining a recommended product by setting a weight (weight_#) to each applied association rule as shown in [Equation 2] below.

[Equation 2]

weight_1 x Association Rule_1 + weight_2 x Association Rule_2 +… + weight_p x Association rule_p

However, in this case, it is difficult to establish a criterion for setting a weight for each of the association rules, and when the same weight is applied to each association rule, it is difficult to guarantee the recommendation quality for the determined recommended product.

In addition to the generally applied association rules such as support, confidence, and lift, in association rule mining, various association rules exist, and the superiority and inferiority of all these association rules are measured and corresponding It will be said that it is practically impossible to set weights.

Accordingly, in an embodiment of the present invention, in the case of determining a recommended product by applying a plurality of association rules, a method for determining a related product having a degree of association that can satisfy all of the plurality of association rules as a recommended product. It intends to propose a method, and it will be described in detail below.

The terminal device 100 accesses the service device 200 and performs a function of requesting a recommended product.

More specifically, the terminal device 100 connects to the service device 200 through the Internet or a mobile web and requests a recommended product by clicking, for example, a detailed page related to a reference product, thereby causing the service device 200 to The related products related to the reference product are checked, and a related product having a high degree of correlation with the reference product is determined as a recommended product and provided to the terminal device 100.

The service device 200 performs a function of checking related products having a related relationship with the reference product.

More specifically, when a recommendation product related to a reference product is requested from the terminal device 100, the service device 200 is distributed and stored in a distributed system corresponding to a plurality of data nodes (Data Node #1 ~ Data Node #N). A number of associations mapped to the identification information (ID) of the reference product and stored in order to calculate the degree of association between a reference product and a related product that has a correlation relationship with the reference product and the relationship between the reference product and the related product. Collect the rules.

At this time, the service device 200 selects at least two association rules for determining a recommended product from among the collected plurality of association rules, so that the related products having a correlation with the reference product in relation to the selected at least two association rules are selected. Will confirm.

In addition, the service device 200 calculates a degree of association for the identified related product.

More specifically, when the service device 200 confirms related products having a correlation with a reference product in relation to at least two selected association rules, each of the at least two association rules for each of the related products The degree of association with the reference product based on is calculated.

In addition, the service device 200 performs a function of selecting an optimal associated product for each association rule.

More specifically, when the degree of relevance to the related products in each of the selected at least two association rules is calculated, the service device 200 determines the relationship in a specific association rule among the at least two association rules based on the calculated association degree. By selecting the best related products for each related rule, which is the related product with the greatest degree of correlation, the related products selected as the best related products for each related rule can be determined as recommended products.

In addition, the service device 200 performs a function of determining a relative optimal associated product.

More specifically, when the optimal associated product for each association rule is selected, the service device 200 is configured in a different association rule than the optimal associated product in a different association rule than the specific association rule in which the greatest association degree is calculated. By determining whether there is a relative optimal related product, which is a related product with a large degree of correlation of, a related product determined as a relative optimal related product can be determined as a recommended product.

Here, the service device 200 determines the relative optimal products related to the best related products for each selected related rule, at each of the plurality of selection levels designated for determining the recommended products, so that the optimal association for each related rule selected at each selection level Along with the product, a related product that has been determined as a relative optimal product can be determined as a recommended product.

In addition, the service device 200 performs a function of determining a recommended product.

More specifically, the service device 200 provides the terminal device 100 by determining a related product selected as the best related product for each related rule at each of a plurality of selection levels or determined as a relative optimal related product as a recommended product. Is done.

At this time, the service device 200 determines the related products selected as the best related products for each related rule at each selection level or determined as the relative best related products as recommended products, and the number of recommended products accumulated and updated up to the current selection level. If is less than the set number, recommended products are additionally determined through a lower selection step.

On the other hand, when the number of recommended products cumulatively updated up to the current selection level exceeds the set number, the service device 200 randomly selects only some of the recommended products determined in the current selection step. Make sure that the number does not exceed the set number.

Hereinafter, the configuration of the service apparatus 200 according to an embodiment of the present invention will be described in more detail with reference to FIG. 3.

That is, the service device 200 according to an embodiment of the present invention includes a check unit 210 for identifying a related product having a related relationship with a reference product, and a calculation unit 220 for calculating a degree of association with the related product. ), a selection unit 230 for selecting an optimal associated product for each association rule, a determination unit 240 for determining a relative optimal associated product, and a determination unit 250 for determining a recommended product.

In addition, in addition to the above-described configuration, the service device 200 according to an embodiment of the present invention includes an update unit 260 for accumulating and updating the number of recommended products, and a calculation unit for determining a recommendation degree for recommended products ( 270), and a providing unit 280 for providing the determined recommended product to the terminal device 100 may be further included.

Here, the above-described verification unit 210, calculation unit 220, selection unit 230, determination unit 240, determination unit 250, update unit 260, book unit 270, providing unit 280 Each component of the service device 200 including) may be implemented as a software module executed by a processor, or may be implemented by a combination thereof.

The verification unit 210 performs a function of checking related products having a related relationship with the reference product.

More specifically, when a recommendation product related to a reference product is requested from the terminal device 100, the confirmation unit 210 is distributed and stored in a distributed system corresponding to a plurality of data nodes (Data Node #1 ~ Data Node #N). A number of associations mapped to the identification information (ID) of the reference product and stored in order to calculate the degree of association between a reference product and a related product that has a correlation relationship with the reference product and the relationship between the reference product and the related product. Collect the rules.

At this time, when the distributed system is configured with, for example, Hadoop Distributed File System (HDFS), the verification unit 210 applies MapReduce as a framework for collecting information from the distributed system, By using the identification information (ID) of the reference product as the key (KEY), it is possible to collect all of a plurality of related rules mapped and stored therein from the distributed system.

In addition, the verification unit 210 selects at least two association rules for determining a recommended product from among a plurality of collected association rules, so as to select related products that have a correlation with the reference product in relation to the selected at least two association rules. Will confirm.

The calculation unit 220 calculates a degree of association for the identified related product.

More specifically, when the calculation unit 220 confirms the related products having a correlation with the reference product in relation to at least two selected association rules, for each of the related products, each of the at least two association rules The degree of association with the reference product based on is calculated.

For example, if at least two association rules previously selected are the confidence association rules and the lift association rules, the degree of association in the confidence association rules through [Equation 3] below. Reliability is calculated, and the degree of improvement corresponding to the degree of association in the lift association rule can be calculated through [Equation 4] below.

[Equation 3]

ㆍRatio of transactions including products A and B to transactions including product A

[Equation 4]

ㆍReliability from product A to B, equivalent to the reliability of product B

The selection unit 230 performs a function of selecting an optimal associated product for each association rule.

More specifically, when the degree of relevance to the related products in each of the selected at least two relevance rules is calculated, the selection unit 230 determines the relevance of the at least two relational rules in a specific relational rule based on the calculated relevance. By selecting the best related products for each related rule, which is the related product with the greatest degree of correlation, the related products selected as the best related products for each related rule can be determined as recommended products.

For example, when the reliability (Confidence) association rule and the lift (Lift) association rule are viewed as a two-dimensional plane with X and Y axes, respectively, as shown in Fig. 4, the related products are It can be sorted according to the size of the calculated correlation (reliability, improvement).

Here, in the case of'Product #13', it is the related product with the highest reliability calculated in the Confidence association rule, and in the case of'Product #14', the improvement rate is the largest in the Lift related rule. Since it is a related product,'Product #1' and'Product #14' can be selected as the best related products according to the related rules.

On the other hand, in the case of'Product #1', the reliability is the same as that of'Product #13', whereas it is a related product with a small improvement, and in the case of'Product #2', the improvement is the same as that of'Product #14'. However, since it is a related product whose reliability is calculated to be small, it does not correspond to the best related product by association rule.

The determination unit 240 performs a function of determining a relative optimum associated product.

More specifically, when the best related products for each related rule are selected, the determination unit 240 determines the different related products than the best related products in a related rule different from a specific related rule that calculates the greatest degree of correlation of the best related products for each related rule. By determining whether there is a relative optimal associated product, which is a related product with a large degree of association calculated in the association rule, a related product determined as a relative optimal associated product can be determined as a recommended product.

Likewise, referring to FIG. 4, in the case of'Product #12', the degree of improvement is smaller than that of'Product #14' selected as the best associated product for each association rule in the case of'Product #12', but the degree of reliability is high. It is a product, and in the case of'Product #15', the reliability is lower than that of'Product #13', which was selected as the best related product by the related rule in the Confidence related rule, but because it is a related product with a large degree of improvement,'Product # 12' and'Product #15' can be identified as relatively optimal associated products.

Here, the determination unit 240 determines the relative optimal products related to the best related products for each selected related rule, at each of the plurality of selection levels designated for determining the recommended products, so that the optimal association for each related rule selected at each selection level Along with the product, a related product that has been determined as a relative optimal product can be determined as a recommended product.

At this time, in the case of a related product that is selected as the best related product for each related rule at each of the plurality of screening levels or is determined as a relative best related product, in the case of the neighboring sub-selection, the best related product for each related rule and the target for determining the relative best related product. Of course it is excluded.

For example, as shown in FIG. 4, the service device 200 includes an optimum associated product for each association rule at each of a plurality of selection levels (first selection level to fifth selection level), and a correlation corresponding to a relative optimum associated product. In the product, that is, the'first screening level','Product #14','Product #12','Product #15'; In the'second screening level','Product #2','Product #17','Product #8','Product #10','Product #11','Product #1'; 'Product #5' in the'third screening level'; In the '4th screening level','Product #3','Product #6','Product #9','Product #4'; In the '5th screening level', by checking'Product #7' and'Product #16', the related products selected as the best related products according to the relevant rules at each screening level, or the related products identified as the relative best related products are recommended as recommended products. Can be determined.

On the other hand, the operation of selecting the optimal associated product for each association rule and determining the relative optimal associated product may be described in the following form.

For example, P association rules C1(x), C2(x),… , When C_p(x) is present, multi-purpose optimization is performed to minimize (maximize) C_n(x), and for this purpose, a concept of dominate may be introduced.

That is, assuming that there are two solutions x1 and x2, and the case of maximization (the larger the value is, the better), it is defined that x1 dominates x2 in the following cases. A case of maximization (the larger the value, the better) is assumed.

If C_n(x1) >= C_n(x2) (n=1,...,P) for all objective functions C_n(x), and the following holds for more than one objective function: C_n(x1)> C_n(x2 )

In other words, when x1 is substituted, if the values of all objective functions are greater than or equal to the values of the objective function when x2 is substituted, and the value of the objective function is greater for at least one objective function, then x1 is said to dominate x2. It can be said, and it can be expressed as [Equation 5] below.

[Equation 5]

In this regard, in the case of the multi-purpose optimization set that corresponds to the result of performing the multi-purpose optimization, it is a set of non-dominate solutions that are not dominated by any other solution, and this It corresponds to a set of related products.

Referring to FIG. 4, since'Product #14' has a high degree of reliability and improvement with respect to'Product #17', that is,'Product #14' dominates'Product #17'. Product #14' is also dominating'Product #5'.

In addition, when comparing'Product #14' and'Product #12','Product #14' has a greater improvement than'Product #12', but its reliability is small. Compared to this, the reliability is large, but the improvement is small.

In other words, both products are not dominated from the viewpoint of the second-order multi-purpose optimization corresponding to the lift association rule and the confidence association rule.

In this way,'Product #14' and'Product #12' are not dominated by any related products, and this is a non-dominated set. , It may correspond to a set of related products that determine the relative optimal related products.

The determination unit 250 performs a function of determining a recommended product.

More specifically, the determination unit 250 determines a related product selected as the best related product for each related rule at each of a plurality of selection levels or determined as a relative best related product as a recommended product, and thus the terminal in the providing unit 280 To be provided to the device 100.

In other words, when the related products with the greatest degree of relevance are the same in each of the relevance rules, the determination unit 250 determines only the relevant products with the highest degree of relevance, that is, the best associated products for each association rule, as recommended products. .

On the other hand, the determination unit 250 determines that when the related products with the greatest degree of correlation in each of the related rules are not the same, that is, the related products selected as the best related products are not all related rules, but only in some related rules. In the case of having a degree, the best associated product for each association rule, and the relative optimal association, which is the associated product with the greatest degree of association, in a different association rule from some association rules with the greatest degree of association. Products are decided together as recommended products.

At this time, the update unit 240 accumulates and updates the number of recommended products determined at each of the plurality of selection levels.

As such, the reason for accumulating and updating the number of recommended products determined in each of the plurality of selection levels is to check whether the accumulated and updated number of recommended products exceeds the number of recommended products that can be provided to the terminal device 100.

That is, the determination unit 250 determines the related products selected as the best related products for each related rule at each selection level or determined as the relative best related products as recommended products, and the number of recommended products accumulated and updated up to the current selection level. If is less than the set number, recommended products are additionally determined through a lower selection step.

On the other hand, when the number of recommended products cumulatively updated up to the current selection level exceeds the set number, the determination unit 250 randomly selects only some of the recommended products determined in the current selection step. Make sure that the number does not exceed the set number.

Meanwhile, the determination unit 250 may change the exposure position (sequence) of the recommended products determined in each selection step in the terminal device 100 between the recommended products determined in the same selection step.

That is, in the past, as shown in Fig. 5 (a), it is impossible to change the designated exposure positions (order) between recommended products, whereas in an embodiment of the present invention, as shown in Fig. 5 (b), the same It is possible to freely change the exposure position (order) in the terminal device 100 between the recommended products determined in the selection step.

For example, as shown in Fig. 6, in some cases, within the same selection step, the exposure positions (order) between recommended products are randomly or changed by applying a specified rule, so that the same or different users It is possible to experience a slightly modified recommendation result in the terminal device 100 in possession.

The designation unit 270 performs a function of setting a degree of recommendation for the determined recommended product.

More specifically, when the recommended products are determined in each selection step, the designation unit 270 sets a recommendation degree of the same size for all of the recommended products selected at the same selection level.

At this time, the designation unit 270 sets a recommendation degree of the same size for all of the recommended products determined at the same selection level, and makes the recommendation degree for the recommended products determined at the higher selection level more than the recommended products determined at the neighboring lower selection level. It is set to be large.

To this end, in order to set the recommendation level higher than the lower level as the selection level is higher, the calculation unit 270 calculates the result of subtracting the order of the current selection level from the order of the lowest selection level as shown in [Equation 6] below. A value divided by the number of currently applied association rules can be determined as a recommendation level.

[Equation 6]

Recommendation = (order of lowest screening level-order of current screening level) / number of applied related rules

For example, referring to FIG. 4, in the case of recommended products ('product #14','product #12', and'product #15') determined at the'first screening level', all of the recommendation levels are '2'. It is set as'[(5-1)/2], and the'Product #3','Product #6','Product #9','Product #4' In this case, all the recommendations may be set as '0.5' [(5-4)/2].

On the other hand, the setting unit 270 may set the recommendation level smaller than the lower level as the selection level is higher, and for this purpose, a value obtained by dividing the order of the current selection level by the number of association rules as shown in [Equation 7] below. Can be set as a recommendation level.

[Equation 7]

Recommendation = Current selection level order / number of applied rules

For example, referring to FIG. 4, in the case of recommended products ('Product #14','Product #12', and'Product #15') determined at the'first screening level', all of the recommended products are '0.5 '(1/2), and in the case of'recommended products determined at the 4th screening level ('product #3','product #6','product #9','product #4'), All recommendations can be set as '2' [4/2].

In this way, determining the degree of recommendation for the recommended product determined at each selection level calculates the degree of association in the merged association rule in which at least two similar association rules that have correlations with each other are merged in the association rules applied to the recommended product. It is to do.

That is, assuming that each of the plurality of association rules applied to determining the recommended product is an incommensurable association rule as shown in FIG. 7, it is also possible to determine the recommended product from the parallel relationship between the association rules. When the number of association rules increases, excessive resources are required for determining recommended products, and further, assuming that they are incommensurable association rules can also serve as another cause of deterioration of recommendation quality.

Accordingly, as shown in FIG. 8, a hierarchical relationship between the association rules may be applied so that the merged association rules combined with the similar association rules having correlations with each other can be applied.

That is, referring to FIG. 8, the degree of recommendation for the recommended product determined from'similar association rule #3' and'similar association rule #4' corresponding to the lowest layer is'merger association rule #B' located in the middle layer. It is calculated as the degree of association for each of the related products in'Similar Association Rule #2.'

Similarly, it is determined from'Similar Association Rule #1','Merge Association Rule #B (Similar Association Rule #2)','Similar Association Rule #3' and other'Similar Association Rule #P', which are the middle class. The degree of recommendation for the recommended product may be calculated as a degree of association for each of the related products in'Merge Association Rule #A' at the top level.

On the other hand, in the case of the correlation between the similar correlation rules, after calculating using the actual purchase log according to the formula or empirical determination method, each association rule It can be determined through a method of grouping association rules with high correlation by obtaining correlation, or a method of grouping association rules based on jacquard similarity.

Here, in the case of a method of grouping the association rules based on the correlation between the association rules, for example, assuming a two-dimensional plane representing each product on the X axis and each association rule on the Y axis, corresponding to each association rule By performing a correlation test on the Y-axis, association rules showing similar trends can be grouped based on this result value.

In addition, in the case of grouping related rules based on jacquard similarity, for example, a list of related products that have a related relationship with product A by association rule #1 is called A_R(C_1), and similarly related rule# If the list of related products that have a relationship with product A selected according to 2 is called A_R(C_2), the jacquard similarity for each product can be compared through [Equation 8] below.

[Equation 8]

Here, the jacquard similarity is obtained by obtaining the number of intersections corresponding to the total number of unions, and may have a value from 0 to 1.

By using the list of related products generated based on different criteria, each jacquard similarity is calculated for all products and the average is averaged as shown in [Equation 9] below, so that the similarity between all related rules can be found.

[Equation 9]

ㆍSimilarity between the i-th criterion and the j-th criterion

As a result, similarity can be calculated between each pair of association rules for all the P criteria, and the criteria with high similarity can be grouped into one group.

As described above, according to the product recommendation service system according to an embodiment of the present invention, the degree of relevance of the related products calculated from a plurality of different association rules is solved by solving the limitation of randomly selecting one association rule. You can determine the recommended product based on it. In addition, when one association rule is applied, a threshold value that must be arbitrarily designated by a user through analysis or statistical calculation is not required, so that a reduction in recommendation quality due to an incorrect threshold value can be prevented. In addition, when there are many applicable association rules, recommendation quality can be greatly improved by applying a hierarchical structure that can be applied by merging similar association rules with each other.

Hereinafter, a product recommendation service method according to an embodiment of the present invention will be described with reference to FIGS. 9 and 10. Here, for convenience of description, the configurations illustrated in FIGS. 1 to 8 described above will be described with reference to the corresponding reference numerals.

First, an operation flow in the product recommendation service system according to an embodiment of the present invention will be described with reference to FIG. 9.

First, the terminal device 100 accesses the service device 200 through the Internet or a mobile web and requests a recommended product by clicking, for example, a detailed page related to a reference product, thereby causing the service device 200 to By checking the related products related to the product, a related product having a high correlation with the reference product can be determined as a recommended product (S110).

Then, when a recommendation product related to a reference product is requested from the terminal device 100, the service device 200 is distributed and stored in a distributed system corresponding to a plurality of data nodes (Data Node#1 to Data Node#N). A plurality of association rules mapped and stored in identification information (ID) of the reference product in order to calculate the degree of association between a product and a related product that has a correlation with the product and the relationship between the reference product and the related product, which is quantified To collect (S120).

At this time, the service device 200 selects at least two association rules for determining a recommended product from among the collected plurality of association rules, so that the related products having a correlation with the reference product in relation to the selected at least two association rules are selected. Will confirm.

Then, when the service device 200 confirms the related products having a correlation with the reference product in relation to at least two selected association rules, for each of the related products, the at least two association rules are The degree of association with the reference product is calculated (S130).

Next, the service device 200 calculates the association degree for the related products in each of the selected at least two association rules, the association in a specific association rule among the at least two association rules based on the calculated association degree. By selecting the best related products for each related rule, which is the largest calculated related product, the related products selected as the best related products for each related rule can be determined as recommended products (S140).

In addition, when the optimal associated product for each association rule is selected, the service device 200 may be configured with a different association rule than the optimal associated product in a different association rule than the optimum associated product. By determining whether there is a relative optimal related product, which is a related product with a large degree of relevance, the related product determined as the relative optimal related product can be determined as a recommended product (S150).

At this time, the service device 200 determines the relative optimal products related to the best related products for each selected related rule, at each of the plurality of selection levels designated for determining the recommended products, so that the optimal association for each related rule selected at each selection level Along with the product, a related product that has been determined as a relative optimal product can be determined as a recommended product.

Thereafter, the service device 200 provides the terminal device 100 by determining a related product that has been selected as an optimal associated product for each association rule at each of a plurality of selection levels or determined as a relative optimal associated product as a recommended product. (S160-S170).

At this time, the service device 200 determines the related products selected as the best related products for each related rule at each selection level or determined as the relative best related products as recommended products, and the number of recommended products accumulated and updated up to the current selection level. If is less than the set number, recommended products are additionally determined through a lower selection step.

On the other hand, when the number of recommended products cumulatively updated up to the current selection level exceeds the set number, the service device 200 randomly selects only some of the recommended products determined in the current selection step. Make sure that the number does not exceed the set number.

Hereinafter, the operation of the service device 200 according to an embodiment of the present invention will be described in detail with reference to FIG. 10.

First, when a recommendation product related to a reference product is requested from the terminal device 100, the verification unit 210 is a reference product distributed and stored in a distributed system corresponding to a plurality of data nodes (Data Node #1 ~ Data Node #N). In order to calculate the degree of correlation of each of the related products having a correlation with the reference product and the correlation between the reference product and the reference product, a plurality of association rules mapped and stored in the identification information (ID) of the reference product are used. Collect (S210).

At this time, the verification unit 210 selects at least two association rules for determining a recommended product from among the collected plurality of association rules, so as to select related products that have a correlation with the reference product in relation to the selected at least two association rules. Will confirm.

Then, when the calculation unit 220 confirms the related products having a correlation with the reference product in relation to at least two selected association rules, for each of the related products, each of the at least two association rules The degree of association with the reference product is calculated (S220).

Then, the selection unit 230 calculates the degree of association for the related products in each of the selected at least two association rules, based on the calculated degree of association, the association in a specific association rule among the at least two association rules. By selecting the best related products for each related rule, which is the largest calculated related product, the related products selected as the best related products for each related rule can be determined as recommended products (S230).

Then, when the best related product for each related rule is selected, the determination unit 240 determines that the best related product is selected from the other related rule than the best related product in a related rule different from the specific related rule in which the greatest degree of correlation is calculated. By determining whether there is a relative optimal associated product, which is a related product with a large degree of association, the associated product determined as the relative optimal associated product can be determined as a recommended product (S240-S250).

At this time, the determination unit 240 determines the relative optimal products related to the best related products for each selected related rule, at each of the plurality of selection levels designated for determining the recommended products, so that the optimal association for each related rule selected at each selection level Along with the product, a related product that has been determined as a relative optimal product can be determined as a recommended product.

Here, in the case of a related product that is selected as the best related product for each related rule at each of the plurality of selection levels or is determined as a relative best related product, in the case of a neighboring sub-selection, the best related product for each related rule and the target for determining the relative best related product. Of course it is excluded.

Then, the determination unit 250 determines a related product selected as the best related product for each related rule at each of the plurality of selection levels or determined as a relative best related product as a recommended product, and thus the terminal device in the providing unit 280 It should be provided to (100) (S260-S300).

At this time, the determination unit 250 determines the related products selected as the best related products for each related rule at each selection level or determined as the relative best related products as recommended products, and the number of recommended products accumulated and updated up to the current selection level. If is less than the set number, recommended products are additionally determined through a lower selection step.

On the other hand, when the number of recommended products cumulatively updated up to the current selection level exceeds the set number, the determination unit 250 randomly selects only some of the recommended products determined in the current selection step. Make sure that the number does not exceed the set number.

As described above, according to the product recommendation service method according to an embodiment of the present invention, the degree of relevance of the related products calculated from a plurality of different association rules is solved by solving the limitation of randomly selecting one association rule. You can determine the recommended product based on it. In addition, when one association rule is applied, a threshold value that must be arbitrarily designated by a user through analysis or statistical calculation is not required, so that a reduction in recommendation quality due to an incorrect threshold value can be prevented. In addition, when there are many applicable association rules, recommendation quality can be greatly improved by applying a hierarchical structure that can be applied by merging similar association rules with each other.

Meanwhile, the steps of the method or algorithm described in connection with the embodiments presented herein may be directly implemented in hardware or implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like alone or in combination. The program instructions recorded on the 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 recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -A hardware device specially configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of the program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The above-described hardware device may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

Although the present invention has been described in detail with reference to preferred embodiments so far, the present invention is not limited to the above-described embodiments, and the technical field to which the present invention belongs without departing from the gist of the present invention claimed in the following claims. Anyone of ordinary skill in the art will say that the technical idea of the present invention extends to the range in which various modifications or modifications are possible.

According to the product recommendation service system and method of the present invention, and the apparatus applied thereto, in the case of determining a recommended product by applying a plurality of association rules, an association having a degree of association that can satisfy all of a plurality of association rules In the sense that a product can be determined as a recommended product, as it exceeds the limitations of the existing technology, the possibility of marketing or sales of the applied device is sufficient as well as the degree to be practically obvious. It is an invention that can be used in the future.

100: terminal device
200: service device
210: confirmation unit 220: calculation unit
230: selection unit 240: identification unit
250: decision unit 260: update unit
270: government administration 280: provision

Claims (15)

A calculation unit that calculates a degree of association for each association rule for each associated product that has an association relationship between the first association rule and the second association rule different from the first association rule; And
Among the first association rules and the second association rules, a specific association rule has the highest degree of relevance, and for other related products with the same degree of association in the specific association rule, the other association rules have a higher association than the other associated products. The relative products, which are related products, which are related products with a higher degree of correlation than the best-related products, in the order of the neighboring order in which the degree of relevance decreases in the specific association rule. A service apparatus comprising a determination unit for determining an optimal associated product as a recommended product. delete delete delete delete delete delete A calculation step of calculating a degree of association for each association rule for each associated product that has an association relationship between a first association rule and a second association rule different from the first association rule; And
Among the first association rules and the second association rules, a specific association rule has the highest degree of relevance, and for other related products with the same degree of association in the specific association rule, the other association rules have a higher association than the other associated products. The relative products, which are related products, which are related products with a higher degree of correlation than the best-related products, in the order of the neighboring order in which the degree of relevance decreases in the specific association rule. And a determining step of determining an optimal associated product as a recommended product. delete delete delete delete delete delete delete

Images