KR102520651B1 - Independent product recommendation service establishment system and method of providing product recommendation service using the same - Google Patents

Abstract

The present invention relates to a system and method for developing a product recommendation service. According to the present invention, the development of an e-commerce product recommendation service based on machine learning requires specialized knowledge and skills, but the prior art In the content, only the contents of the service method or system itself are presented, and the recommendation model generation engine that creates a recommendation model suitable for the user by resolving the problem that the development environment was not presented, and evaluating the performance of the generated recommendation model Development of an integrated and independent product recommendation service for the development of machine learning-based e-commerce product recommendation service It provides an environment to increase the accessibility of service development using machine learning, and product recommendation services can be easily developed without professional knowledge or technology. An independent product recommendation service construction system configured to minimize the burden on an e-commerce operating company and a product recommendation service providing method using the same are provided.

Description

Independent product recommendation service establishment system and method of providing product recommendation service using the same}

The present invention relates to a system and method for providing a platform for developing a product recommendation service that automatically suggests products that are determined to be necessary for consumers during e-commerce, and more particularly, to a method for providing only a product recommendation service. However, only the contents of the system itself are presented, and the problems of the prior art product recommendation service systems and methods, which have not been presented for the development environment in which such product recommendation services can be developed, and the need for specialized knowledge and skills, make general users In order to solve the problems of the existing machine learning-based e-commerce product recommendation service development work, which had limitations in development due to high development difficulty, machine learning algorithms were automatically applied to the development of e-commerce product recommendation models. It relates to an independent product recommendation service construction system configured to provide an integrated and independent product recommendation service development environment for developing a machine learning-based e-commerce product recommendation service and a product recommendation service providing method using the same.

In addition, the present invention, as described above, in order to provide an integrated and independent product recommendation service development environment for developing a machine learning-based e-commerce product recommendation service, a recommendation model that generates a recommendation model suitable for a user is generated. Increase accessibility of service development using machine learning, including a performance evaluation module that evaluates the performance of the engine and the generated recommendation model, and a distribution module that distributes the recommendation model verified through performance evaluation to the operating e-commerce server. Product recommendation service can be easily developed without professional knowledge or technology on artificial intelligence, and at the same time, e-commerce operation server and service development server are operated independently, making it easy to develop new services as well as existing services. It relates to an independent product recommendation service construction system configured to minimize the burden of an e-commerce operating company and a product recommendation service providing method using the same.

BACKGROUND OF THE INVENTION [0002] Recently, as electronic commerce through the Internet or mobile is actively conducted, various services are provided to increase the efficiency of product purchase and to satisfy various needs of consumers.

That is, as an example of a service for increasing the efficiency of product purchase and satisfying various needs of consumers in e-commerce as described above, for example, by analyzing consumption patterns such as consumer preferences and purchase histories, suitable products suitable for each user are analyzed. There is a product recommendation service that automatically recommends products.

Here, as an example of the prior art related to the e-commerce product recommendation service as described above, for example, as presented in Korean Patent Registration No. 10-2179397, "product recommendation service system and method, and apparatus applied thereto" "There is

More specifically, the above-mentioned Korean Patent Registration Publication No. 10-2179397 determines at least two sales ranking lists related to buyers with respect to products for sale based on sales-related information related to the products for sale, and at least two Based on at least one of the sales ranking list, a sales product suitable for the buyer is selected from among the sales products and the selected sales product is provided as a recommendation product to the buyer, thereby increasing the reliability of the determined recommended product and increasing the buyer's satisfaction with the product. It relates to a product recommendation service system and method configured to increase a seller's profit.

In addition, another example of the prior art related to the e-commerce product recommendation service as described above is, for example, "product recommendation method and server using the same" as presented in Korean Patent Publication No. 10-2019-011470. .

More specifically, Korean Patent Laid-Open Publication No. 10-2019-011470 discloses a step of receiving a selection input for a first product from a user among a plurality of registered products, a correlation model modeling the relationship between registered products. The step of extracting a plurality of candidate products related to the first product using and comparing the user's taste vector with the expression vector of the candidate products to set a recommended product for the first product, It relates to a product recommendation method capable of providing an accurate customized product recommendation service according to taste or preferred style and a server using the same.

As described above, in the past, in order to increase consumer satisfaction during e-commerce, a number of product recommendation services have been proposed that recommend products in a customized way by analyzing the taste or consumption pattern of each consumer, but as described above, the prior art The contents of this had the following problems.

That is, in order to provide a customized product recommendation service during e-commerce, a separate product recommendation model must be developed using an artificial intelligence algorithm, such as machine learning.

However, since artificial intelligence technology and development such as machine learning is a very difficult task, it is not easy to develop a separate product recommendation model using artificial intelligence algorithms at the level of companies or individuals operating e-commerce. , In general, most of the recommendation services are built by commissioning artificial intelligence experts or development companies.

In addition, as shown in the above-mentioned prior art documents, most of the contents of the prior art present only the service provision method or system itself for providing a customized product recommendation service during e-commerce, and depend on experts. However, there was no specific presentation on the development method that can actually implement this product recommendation service.

Therefore, as described above, only the service provision method or system itself for providing a customized product recommendation service is presented, but the development environment such as a method or platform for developing such a product recommendation service is not presented, so artificial intelligence algorithms In order to solve the problems of the product recommendation service and its development environment in the prior art, which had limitations in that it was not easy to develop a product recommendation model using, an independent development environment was provided to increase the accessibility of service development using machine learning and anyone It is desirable to provide a product recommendation service development platform with a new structure configured to easily develop product recommendation services, but a device or method that satisfies all such demands has not yet been presented.

Korean Registered Patent Publication No. 10-2179397 (2020.11.10.) Korean Patent Publication No. 10-2019-011470 (2019.10.10.)

The present invention is intended to solve the problems of the prior art as described above, and therefore, an object of the present invention is to provide only the method or system itself for providing a product recommendation service, and to develop such a product recommendation service. Conventional machine learning (Machine Learning), which has limitations in development due to the high development difficulty of ordinary users due to the problems of the prior art product recommendation service systems and methods, which have not been presented for development environments that have existing technology, and the need for specialized knowledge and skills Integrated and independent product for the development of machine learning-based e-commerce product recommendation service by automatically applying machine learning algorithm to the development of e-commerce product recommendation model It is intended to provide an independent product recommendation service construction system configured to provide a recommendation service development environment and a product recommendation service providing method using the same.

In addition, another object of the present invention is to generate a recommendation model suitable for a user in order to provide an integrated and independent product recommendation service development environment for developing a machine learning-based e-commerce product recommendation service as described above. Service development using machine learning by including a recommendation model generation engine, a performance evaluation module that evaluates the performance of the generated recommendation model, and a distribution module that distributes the recommendation model verified through performance evaluation to the operating e-commerce server. product recommendation service can be easily developed without professional knowledge or technology on artificial intelligence, and at the same time, the e-commerce operation server and the service development server are operated independently, thereby not only developing new services but also operating existing ones. It is an object of the present invention to provide an independent product recommendation service construction system configured to minimize the burden on e-commerce operators by easily applying the same to existing services, and a product recommendation service providing method using the same.

In order to achieve the above object, according to the present invention, in an independent product recommendation service construction system configured to provide a development platform for developing a product recommendation service that recommends products to users during e-commerce, An e-commerce server in which e-commerce is performed by providing a product recommendation service; and a recommendation model that operates independently of the e-commerce server, receives various information collected through e-commerce from the e-commerce server in order to implement a product recommendation service, and implements a product recommendation service based on the received information. and a product recommendation service development server configured to generate a product recommendation service development server configured to perform processing for providing product recommendation service by distributing the generated recommendation model to the e-commerce server. system is provided.

Here, the product recommendation service development server includes a recommendation model data module for collecting recommendation model data including transaction data, user data, and product data of the e-commerce server to develop the recommendation model; A recommendation model generation engine that derives an optimal learning algorithm for generating the recommendation model through association analysis based on the recommendation model data and generates the recommendation model through learning using the derived learning algorithm. module; a performance evaluation module for performing a process of evaluating performance of the recommendation model generated by the recommendation model generation engine module; a distribution module for distributing the recommendation model verified through the performance evaluation of the performance evaluation module to the e-commerce server; An OpenAPI module in which a process for providing a user interface for the process for generating the recommendation model is performed; and a web service module in which a process for providing a web service environment is performed for the process for generating the recommendation model.

In addition, the product recommendation service development server, through the recommendation model data module, the recommendation model data including transaction data, user data, and product data of the e-commerce server to apply to learning for generating the recommendation model. a learning data input step in which a process of inputting is performed as learning data; a data schema setting and pre-processing step in which a structure of the learning data is defined based on contents previously set by a user through the recommendation model data module and pre-processing is performed to ensure integrity of the learning data; a data analysis step in which a process of performing data correlation analysis for deriving a learning algorithm based on the structure of the set learning data is performed through the recommendation model generation engine module; a learning algorithm derivation step in which a process of deriving an optimal learning algorithm based on the analysis result of the data analysis step is performed through the recommendation model generation engine module; a learning step of performing a process of setting hyperparameters and performing learning based on contents previously set by a user for the derived learning algorithm through the recommendation model generation engine module; Through the performance evaluation module, performance evaluation is performed on the learning result performed in the learning step, and the learning step and performance evaluation process are repeated until the evaluation result reaches a preset standard to generate an optimized recommendation model. a performance evaluation step in which processing is performed; a distribution step of distributing, through the distribution module, the recommendation model optimized according to contents previously set by a user to the e-commerce server; and through the recommendation model generation engine module, re-learning is performed using data generated through a product recommendation service provided using the distributed recommendation model as learning data, and a process of reflecting the data to the recommendation model is performed. characterized in that it is configured to perform a process comprising the step.

Here, in the step of inputting learning data, big data for a product recommendation service is built by collecting various data related to users, products, and transactions obtained through e-commerce in the e-commerce server and storing them in the form of a database, It is characterized in that the processing of using the big data as the learning data is configured to be performed.

In addition, the data analysis step classifies dependencies in stages according to predetermined criteria according to relationships between fields in the same record or relationships between fields in different records based on an associated distance between each data, and associates the data. It is characterized in that it is configured to perform a process of identifying dependency, similarity, and time series of each data through analysis.

Moreover, in the learning algorithm derivation step, based on the data association analysis result of the data analysis step, an optimal algorithm among learning algorithms including regression, clustering, and recurrent neural network (RNN) It is characterized in that it is configured so that the processing of deriving is performed.

In addition, in the distributing step, when distributing the recommendation model to the e-commerce server, the recommendation model and the recommendation model data are located in the e-commerce server, so that a product recommendation service is performed in the e-commerce server. It is characterized in that the e-commerce server and the product recommendation service development server are independently operated so that the product recommendation service can be applied even when the e-commerce server is in operation.

In addition, in the distributing step, when distributing the recommendation model to the e-commerce server, the recommendation model and the recommendation model data are located in the product recommendation service development server, so that the e-commerce server is subordinate to the product recommendation service development server. The e-commerce server and the product recommendation service development server are independent by the dependent operation distribution method operated and the recommendation model and the recommendation model data being located in the e-commerce server so that the product recommendation service is performed within the e-commerce server. It is characterized in that a process of distributing the recommendation model is performed by selecting one of the independently operated distribution methods operated by .

Moreover, the independent product recommendation service building system includes a plurality of e-commerce servers, and for each e-commerce server, the product recommendation service development server builds a customized product recommendation service and distributes it to each e-commerce server. It is characterized in that it is configured to provide a differentiated product recommendation service with respect to each e-commerce server.

In addition, according to the present invention, in the product recommendation service providing method using the independent product recommendation service establishment system described above, the service implementation step of implementing the product recommendation service through the product recommendation service development server of the independent product recommendation service establishment system. ; and a service providing step of providing the product recommendation service by distributing the implemented product recommendation service to an e-commerce server.

As described above, according to the present invention, e-commerce operating a recommendation model generation engine that generates a recommendation model suitable for a user, a performance evaluation module that evaluates the performance of the generated recommendation model, and a recommendation model verified through performance evaluation. Provides an integrated and independent product recommendation service development environment for the development of machine learning-based e-commerce product recommendation services by automatically applying machine learning algorithms to the development of e-commerce product recommendation models, including distribution modules deployed on servers By providing an independent product recommendation service construction system and a method of providing product recommendation service using the system configured to be able to do so, the accessibility of service development using machine learning is increased, and product recommendation service is easily provided without professional knowledge or technology on artificial intelligence. At the same time, since the e-commerce operation server and the service development server are operated independently, it can be easily applied to existing services as well as new service development, thereby minimizing the burden on the e-commerce operating company.

In addition, according to the present invention, as described above, an independent product recommendation service construction system configured to provide an integrated and independent product recommendation service development environment for developing an e-commerce product recommendation service based on machine learning and product recommendation using the same As the service providing method is provided, only the method for providing product recommendation service or the contents of the system itself are presented, but the product recommendation service system and method in the prior art have not been suggested for the development environment in which such product recommendation service can be developed. At the same time, professional knowledge and skills are required, so general users can solve the problems of the existing machine learning-based e-commerce product recommendation service development work, which had limitations in development due to high development difficulty.

1 is a diagram schematically showing the basic concept of a system for building an independent product recommendation service according to an embodiment of the present invention.
2 is a diagram schematically illustrating a hierarchical structure of each module provided to develop a product recommendation service in a service development server of an independent product recommendation service construction system according to an embodiment of the present invention.
3 is a diagram schematically illustrating a process flow of a service development process for developing a product recommendation service in a service development server of an independent product recommendation service building system according to an embodiment of the present invention.
4 is a flowchart schematically illustrating a process performed to develop a recommendation model in a service development server of an independent product recommendation service construction system according to an embodiment of the present invention.
5 is a diagram schematically showing data association analysis processing for deriving a learning algorithm among processing processes performed to develop a recommendation model in a service development server of an independent product recommendation service establishment system according to an embodiment of the present invention.
6 is a diagram schematically illustrating a process performed to distribute a recommendation model in a service development server of an independent product recommendation service construction system according to an embodiment of the present invention.
7 is a view showing a comparison between a distribution method in the existing product recommendation service development process and a distribution method of the independent product recommendation service construction system according to an embodiment of the present invention. 7b shows a distribution method of the independent product recommendation service construction system according to an embodiment of the present invention.
8 is a diagram schematically showing the overall configuration of a system for building an independent product recommendation service according to an embodiment of the present invention.

Hereinafter, with reference to the accompanying drawings, specific embodiments of an independent product recommendation service construction system and a product recommendation service providing method using the same according to the present invention will be described.

Here, it should be noted that the contents described below are only one embodiment for carrying out the present invention, and the present invention is not limited to the contents of the embodiments described below.

In addition, in the following description of the embodiments of the present invention, for parts that are the same as or similar to the contents of the prior art or are determined to be easily understood and implemented at the level of those skilled in the art, the detailed descriptions are provided to simplify the description. It should be noted that .

That is, the present invention, as will be described later, only presents information on a method or system itself for providing a product recommendation service, and no development environment for developing such a product recommendation service has been suggested in the prior art. Solving the problems of the existing machine learning-based e-commerce product recommendation service development work, which had limitations in development due to the high development difficulty for general users due to the problems of service systems and methods and the need for specialized knowledge and skills In order to do so, an independent product configured to provide an integrated and independent product recommendation service development environment for the development of machine learning-based e-commerce product recommendation services by automatically applying machine learning algorithms to the development of e-commerce product recommendation models. It relates to a recommendation service building system and a product recommendation service providing method using the same.

In addition, the present invention, as will be described later, in order to provide an integrated and independent product recommendation service development environment for machine learning-based e-commerce product recommendation service development, a recommendation model that generates a recommendation model suitable for the user. Increase accessibility of service development using machine learning, including a performance evaluation module that evaluates the performance of the generation engine and the generated recommendation model, and a distribution module that distributes the recommendation model verified through performance evaluation to the operating e-commerce server. , It is possible to easily develop a product recommendation service without professional knowledge or technology on artificial intelligence, and at the same time, it is easy to develop new services as well as existing services by operating the e-commerce operation server and service development server independently. It relates to an independent product recommendation service construction system configured to minimize the burden on an e-commerce operating company and a product recommendation service providing method using the same.

Continuously, with reference to the drawings, details of an independent product recommendation service building system and a product recommendation service providing method using the system according to the present invention will be described.

First, referring to FIG. 1, FIG. 1 is a diagram schematically showing the overall configuration of an independent product recommendation service building system 10 according to an embodiment of the present invention.

As shown in FIG. 1, the system 10 for establishing an independent product recommendation service according to an embodiment of the present invention is broadly divided into an e-commerce server 11 in which e-commerce is performed by providing a product recommendation service to users, and e-commerce Various information for product recommendation service development is received from the server 11, the product recommendation service is developed and built based on the received information, and the processing of distributing the built service to the e-commerce server 11 is performed It may be configured to include a product recommendation service development server 12 .

More specifically, the above-described e-commerce server 11 may be configured as a general e-commerce system in which a product recommendation function and an e-commerce function are performed, and the product recommendation service development server 12 provides a product recommendation service as described later. A process of establishing a recommendation model through a development structure and development procedure and distributing it to the e-commerce server 11 may be configured to be performed.

Here, in the embodiment shown in FIG. 1, the present invention is shown by taking the case where the product recommendation service development server 12 builds and distributes the product recommendation service for one e-commerce server 11 as an example, but the present invention is not necessarily It is not limited to this case only, that is, the present invention, other than the case shown in FIG. 1, for example, product recommendation service development server 12 customized product recommendation service for a plurality of e-commerce servers 11, respectively. It should be noted that it can be configured in various forms as needed, such as being able to provide a differentiated product recommendation service for each e-commerce server 11 by building and distributing.

Continuing to refer to FIGS. 2 and 3, first, FIG. 2 is for developing a product recommendation service in the product recommendation service development server 12 of the independent product recommendation service establishment system 10 according to an embodiment of the present invention. It is a diagram schematically showing the hierarchical structure of each module provided, and FIG. 3 is a method for developing a product recommendation service in the product recommendation service development server 12 of the independent product recommendation service establishment system 10 according to an embodiment of the present invention. It is a diagram schematically showing the processing flow of the service development process for

As shown in FIG. 2, the hierarchical structure of each module constituting the product recommendation service development platform in the product recommendation service development server 12 includes recommendation model data (Recomm. Resource) and recommendation model generation engine (Recomm. Engine). It can be composed of four layers: recommendation model creation OpenAPI (Recomm. OpenAPI) and recommendation model creation web service (Recomm. Web Service).

In addition, in the hierarchical structure of each module shown in FIG. 2, as a core service module, it is composed of a performance evaluation module and a distribution module centering on the recommendation model generation engine, and through this structure, the recommendation model through the recommendation model generation engine. is created, recommending service is provided using recommendation model data, and each module provides Web Service as a development tool providing user interface by providing OpenAPI.

In addition, as shown in FIG. 3, the processing flow of the development platform configured as described above derives a learning algorithm, which is a function associated with the core module of the development platform, to generate a recommendation model, performance evaluation and distribution service Through this, recommendation models are developed and distributed.

Here, the customization service is a service for applying an optimization algorithm when generating a recommendation model, and derives an optimization learning algorithm through association analysis of learning data (transactions, users, products, etc.).

In more detail, referring to FIG. 4, FIG. 4 is based on the development platform configured as described above with reference to FIGS. 2 and 3 of the independent product recommendation service establishment system 10 according to an embodiment of the present invention. It is a flowchart schematically showing a process performed to develop a recommendation model in the product recommendation service development server 12.

As shown in FIG. 4, the procedure for developing a recommendation model is broadly divided into learning data input, data schema setting, data preprocessing, data analysis, learning algorithm derivation, hyperparameter setting, learning, performance evaluation, and recommendation model distribution setting. , it can be configured including the processing of each stage of re-learning.

First, the learning data input step converts various data to be applied to learning for recommendation model development, such as e-commerce transaction data, user data, and product data, for example, to include a product recommendation service, into CVS. It may be configured to perform input processing.

Here, the above learning data is, for example, big data for product recommendation service by collecting various data related to users, products and transactions obtained through e-commerce in the e-commerce server 11 and storing them in the form of a database. is built, and learning can be performed based on the contents of such big data.

Next, the data schema setting step is a step of defining the structure of input learning data, and the data preprocessing step is a step of setting preprocessing items and performing preprocessing as necessary to ensure the integrity of the learning data, and data analysis. This step is a step of performing correlation analysis between data based on the data structure set in the data schema setting step.

In this case, the data schema setting step and the data pre-processing step may be configured to perform corresponding processing according to criteria or items preset by the user.

More specifically, referring to FIG. 5, FIG. 5 is a process performed to develop a recommendation model in the product recommendation service development server 12 of the independent product recommendation service establishment system 10 according to an embodiment of the present invention. It is a diagram schematically showing the data association analysis process for deriving the learning algorithm.

As shown in FIG. 5, the data association analysis process for deriving the learning algorithm determines the dependence according to the relationship between fields in the same record or between fields in different records based on the associated distance between data. Data association analysis is performed by classifying into three categories: strong dependence, medium dependence, and weak dependence.

Therefore, through this data analysis, dependency, similarity, and time series can be identified, and through this, learning such as regression, clustering, and recurrent neural networks (RNNs) can be performed. derive an algorithm.

After that, in the hyperparameter setting step, hyperparameters are set for the algorithm automatically derived as described above, and optimized learning is performed in the learning step, and the learning process is repeated through performance evaluation in the performance evaluation step. Optimization Create a recommended model.

In this case, in the hyperparameter setting step, the hyperparameters may be set based on contents previously set by the user.

Subsequently, the distribution of the recommendation model optimized according to the contents set in the distribution setting step is performed, and the product recommendation service is provided in conjunction with the e-commerce service in operation through distribution, and the learning data generated through the service provision is used for relearning. used

Here, the learning algorithm derivation, performance evaluation, recommendation model generation and distribution, redistribution, and relearning functions among the development procedures of the recommendation model configured as described above can be used in a 3rd party recommendation model development tool. It is provided as an OpenAPI.

Subsequently, with reference to FIGS. 6 and 7 , a process of distributing the recommendation model generated as described above will be described.

That is, referring to FIG. 6, FIG. 6 schematically illustrates a process performed to distribute a recommendation model in the product recommendation service development server 12 of the independent product recommendation service establishment system 10 according to an embodiment of the present invention. It is a drawing that represents

As shown in FIG. 6, the processing process for distributing the recommendation model generated as described above with reference to FIGS. It can be provided in two ways of distribution.

In more detail, first, the dependent operation distribution method generates a recommendation model and recommendation data through learning as described above in the product recommendation service development server 12, and distributes them to the e-commerce server 11. As a method of applying a recommendation model, a recommendation model and recommendation model data are located in the product recommendation service development server 12 so that the e-commerce server 11 is subordinated to and operated by the product recommendation service development server 12 .

On the other hand, in the independent operation and distribution method, the product recommendation service development server 12 only generates and distributes a recommendation model, and the generated recommendation model and recommendation model data are located in the e-commerce server 11, and the e-commerce server (11) is a method in which the e-commerce server 11 and the product recommendation service development server 12 are operated independently by providing the product recommendation service.

Also, referring to FIG. 7, FIG. 7 is a diagram showing a comparison between a distribution method in an existing product recommendation service development process and a distribution method of the independent product recommendation service construction system 10 according to an embodiment of the present invention. 7a shows a distribution method in the existing product recommendation service development process, and FIG. 7b shows a distribution method of the independent product recommendation service establishment system 10 according to an embodiment of the present invention.

As shown in FIG. 7A, in the existing product recommendation service development platform, a recommendation model and recommendation model data are located in the development server, so that the e-commerce server is operated subordinately to the development server, or an independent product according to an embodiment of the present invention. In the recommendation service establishment system 10, as shown in FIG. 7B, the recommendation model and recommendation model data are distributed to the e-commerce server, so that the e-commerce server and the development server can be operated independently.

In addition, the independent product recommendation service construction system 10 according to an embodiment of the present invention, as shown in FIG. It should be noted that it can be configured in any way.

Therefore, as described above, the system 10 for establishing an independent product recommendation service according to an embodiment of the present invention can be implemented. That is, referring to FIG. 8, FIG. 8 shows the construction of an independent product recommendation service according to an embodiment of the present invention. It is a diagram schematically showing the overall configuration of the system 10.

As shown in FIG. 8, the product recommendation service development server 12 provides a development environment for product recommendation service development, that is, a product recommendation service development platform, through each internal module, and distributes the generated recommendation model. Depending on the method, it is possible to operate an e-commerce server and an independent product recommendation service development platform.

Here, the product recommendation service development platform, as described above with reference to FIGS. 2 and 3, provides OpenAPI based on a recommendation engine including a recommendation model generation engine, a performance evaluation module, and a distribution module as a development tool. It can be configured to provide Web Services.

In addition, as described above with reference to FIGS. 4 and 5, the product recommendation service development platform automatically derives a learning algorithm through data analysis with learning data as an input, and then performs learning based on this to make recommendations. It can be configured to create a model.

In addition, as described above with reference to FIG. 6 , the product recommendation service development platform may be selectively configured as a dependent operation distribution method or an independent operation distribution method according to the distribution method of the generated recommendation model.

That is, in the product recommendation service development server 12, when learning data including e-commerce transaction data, user data, and product data is input, pre-processing is performed based on the data schema and pre-processing items preset by the user, and data Through analysis, data correlation analysis is performed to automatically generate a learning algorithm, learning is performed based on hyperparameters preset by the user for the generated learning algorithm, and performance evaluation is performed on the learning result, followed by performance evaluation. A product recommendation service is provided by generating an optimal recommendation model by repeating learning according to the result, packaging the generated recommendation model, and distributing it to the operating e-commerce server 11. At this time, the product recommendation service development server ( 12) and the e-commerce server 11 are independently operated.

Therefore, through the configuration as described above, a developer can develop a service by using the product recommendation service development platform in order to add a product recommendation service to the e-commerce server 11 in operation, and thereby, for the learning algorithm. Even without professional knowledge or skills, a learning algorithm can be automatically generated through data analysis of the product recommendation service development platform, and product recommendation service development based on machine learning can be easily performed.

Furthermore, by configuring the distribution method of the product recommendation service development platform as an independent operation distribution method, the operating e-commerce server and the development server can be operated independently, thereby minimizing the burden on the e-commerce operating company.

Therefore, as described above, it is possible to easily implement the independent product recommendation service building system 10 and the product recommendation service providing method using the same according to the embodiment of the present invention, whereby, according to the present invention, recommendation suitable for the user Recommendation of e-commerce products, including a recommendation model generation engine that generates models, a performance evaluation module that evaluates the performance of the generated recommendation models, and a distribution module that distributes the recommendation models verified through performance evaluation to the operating e-commerce server. An independent product recommendation service construction system configured to provide an integrated and independent product recommendation service development environment for developing a machine learning-based e-commerce product recommendation service by automatically applying a machine learning algorithm to the development of a model, and using the same By providing a method of providing product recommendation service, it is possible to increase the accessibility of service development using machine learning and easily develop product recommendation service without professional knowledge or technology on artificial intelligence, and at the same time, e-commerce operation server and service development server is operated independently, so it can be easily applied not only to new service development but also to existing services, so the burden on e-commerce operators can be minimized.

In addition, according to the present invention, as described above, an independent product recommendation service construction system configured to provide an integrated and independent product recommendation service development environment for developing an e-commerce product recommendation service based on machine learning and product recommendation using the same As the service providing method is provided, only the method for providing product recommendation service or the contents of the system itself are presented, but the product recommendation service system and method in the prior art have not been suggested for the development environment in which such product recommendation service can be developed. At the same time, professional knowledge and skills are required, so general users can solve the problems of the existing machine learning-based e-commerce product recommendation service development work, which had limitations in development due to high development difficulty.

In the above, the details of the independent product recommendation service establishment system and product recommendation service providing method using the same according to the present invention have been described through the embodiments of the present invention as described above, but the present invention is described in the above embodiments. Therefore, it is natural that the present invention is capable of various modifications, changes, combinations, and substitutions according to design needs and other various factors by those skilled in the art to which the present invention belongs. I'd say it's work.

10. Independent product recommendation service establishment system
11. E-Commerce Server
12. Product recommendation service development server

Claims (10)

In an independent product recommendation service construction system configured to provide a development platform for developing a product recommendation service that recommends products to users during e-commerce,
An e-commerce server in which e-commerce is performed by providing a product recommendation service; and
It operates independently of the e-commerce server, receives various information collected through e-commerce from the e-commerce server in order to implement a product recommendation service, and creates a recommendation model for implementing a product recommendation service based on the received information. and a product recommendation service development server configured to perform processing for providing a product recommendation service through the e-commerce server by generating and distributing the generated recommendation model to the e-commerce server,
The product recommendation service development server,
a recommendation model data module for collecting recommendation model data including transaction data, user data, and product data of the e-commerce server to develop the recommendation model;
A recommendation model generation engine module that derives a learning algorithm for generating the recommendation model through association analysis based on the recommendation model data and generates the recommendation model through learning using the derived learning algorithm;
a performance evaluation module for performing a process of evaluating performance of the recommendation model generated by the recommendation model generation engine module;
a distribution module for distributing the recommendation model verified through the performance evaluation of the performance evaluation module to the e-commerce server;
An OpenAPI module in which a process for providing a user interface for the process for generating the recommendation model is performed; and
By including a web service module in which a process for providing a web service environment is performed for the process for generating the recommendation model,
An independent product recommendation service construction system characterized in that it is configured to build a customized product recommendation service for each e-commerce server by providing a machine learning-based development environment for developing an e-commerce product recommendation service.
delete According to claim 1,
The product recommendation service development server,
Through the recommendation model data module, processing of inputting the recommendation model data including transaction data, user data, and product data of the e-commerce server as learning data to apply to learning for generating the recommendation model is performed Learning data input step;
a data schema setting and pre-processing step in which a structure of the learning data is defined based on contents previously set by a user through the recommendation model data module and pre-processing is performed to ensure integrity of the learning data;
a data analysis step in which a process of performing data correlation analysis for deriving a learning algorithm based on the structure of the set learning data is performed through the recommendation model generation engine module;
a learning algorithm derivation step of deriving a learning algorithm representing the best result based on the analysis result of the data analysis step through the recommendation model generation engine module;
a learning step of performing a process of setting hyperparameters and performing learning based on contents previously set by a user for the derived learning algorithm through the recommendation model generation engine module;
Through the performance evaluation module, a process of generating a recommendation model by performing performance evaluation on the learning result performed in the learning step and repeating the learning step and the performance evaluation process until the evaluation result reaches a preset standard. A performance evaluation step in which is performed;
a distribution step in which a process of distributing the recommendation model generated according to contents previously set by a user to the e-commerce server is performed through the distribution module; and
A re-learning step in which re-learning is performed using data generated through a product recommendation service provided using the distributed recommendation model through the recommendation model generation engine module as learning data, and a process of reflecting the data to the recommendation model is performed. An independent product recommendation service establishment system, characterized in that configured to perform processing including.
According to claim 3,
In the step of inputting the learning data,
The e-commerce server collects various data related to users, products, and transactions obtained through e-commerce and stores them in the form of a database to build big data for a product recommendation service, and use the big data as the learning data. A system for establishing an independent product recommendation service, characterized in that the processing is configured to be performed.
According to claim 4,
The data analysis step,
Based on the associated distance between each data, the relationship between fields in the same record or the relationship between fields in different records is classified in stages according to predetermined criteria and data association analysis is performed. An independent product recommendation service construction system characterized in that it is configured to perform processing for determining dependency, similarity and time series.
According to claim 5,
In the step of deriving the learning algorithm,
Based on the results of the data association analysis in the data analysis step, a process of deriving an algorithm showing the best result among learning algorithms including regression, clustering, and recurrent neural network (RNN) is performed. An independent product recommendation service construction system, characterized in that configured to be.
According to claim 6,
The distribution step is
When distributing the recommendation model to the e-commerce server, the recommendation model and the recommendation model data are located in the e-commerce server so that a product recommendation service is performed in the e-commerce server. An independent product recommendation service construction system, characterized in that the commercial transaction server and the product recommendation service development server are independently operated so that the product recommendation service can be applied even when the e-commerce server is in operation.
According to claim 7,
The distribution step is
When the recommendation model is distributed to the e-commerce server, the recommendation model and the recommendation model data are located in the product recommendation service development server, so that the e-commerce server is operated subordinately to the product recommendation service development server. ,
Among the independent operation distribution methods in which the e-commerce server and the product recommendation service development server are operated independently by the recommendation model and the recommendation model data being located in the e-commerce server and the product recommendation service being performed in the e-commerce server. A system for constructing an independent product recommendation service, characterized in that a process of distributing the recommendation model by selecting one method is performed.
According to claim 8,
The independent product recommendation service building system,
Including a plurality of e-commerce servers, the product recommendation service development server for each e-commerce server is configured to build a customized product recommendation service and distribute it to each e-commerce server, so that each e-commerce server An independent product recommendation service construction system characterized in that it is configured to provide a differentiated product recommendation service for a commercial transaction server.
In the product recommendation service providing method using the independent product recommendation service building system according to any one of claims 1 and 3 to 9,
a service implementation step of implementing a product recommendation service through a product recommendation service development server of the independent product recommendation service establishment system; and
A product recommendation service providing method comprising a service providing step of providing a product recommendation service by distributing the implemented product recommendation service to an e-commerce server.

Images