KR20010111913A - Complex filtering apparatus and method for database marketing in electronic commerce - Google Patents

Abstract

The present invention relates to a complex filtering device and method for database marketing in an electronic commerce to recommend a group of products and information that match each consumer's preferences on the web, the present invention is an item of a website selected by the consumer HTML that shows the, and a merchant server that shows the product and service information in detail when clicking the item column shown in the HTML or provide a recommended item to the consumer, among the items provided by the merchant server according to the consumer's taste A front module that collects data to recommend items, a database that accumulates and analyzes data collected from the front module, and makes a database in accordance with consumer preferences, and collects data extracted from the front module and the database. filter And applying complex content-based filtering algorithms are configured into a hybrid filter that the probability of purchase in accordance with the preferences of the individual consumer provides the highest item.

Description

Complex filtering apparatus and method for database marketing in electronic commerce}

The present invention relates to electronic commerce, and more particularly, to a complex filtering device and method for database marketing in electronic commerce to recommend a group of items (product and service information) corresponding to each consumer's preferences on the web. will be.

In the modern society, as the information and communication technology develops, the field of information and communication using the Internet is rapidly developing. With the development of information and communication using the Internet, it was necessary to go directly to the store that sells the goods in order to purchase the goods, but it is also possible in each home or workplace by internet access through a computer or an internet TV.

As described above, the electronic commerce refers to a transaction in which all or part of the transactions of goods or services are purchased through a network through a computer or the like through electronic methods such as electronic document exchange. It has a form similar to that of commerce.

In electronic commerce, a virtual shop (CYBER MALL, hereinafter referred to as a shopping mall) that provides goods or services and a consumer who purchases goods or services in the shopping mall become subjects of transactions in the electronic commerce.

In the electronic commerce having the above-described configuration, the shopping mall operator posts information about a product on a shopping mall site or distributes the information to consumers by e-mail to help consumers purchase.

In addition, in order to make a purchase, a consumer may refer to an e-mail transmitted from a shopping mall operator or directly access a shopping mall site to search for information about a product. At this time, since the shopping mall site presents only the most popular products or information, the process of searching for the products that consumers want is complex when directly accessing them. It is difficult to connect to direct buying behavior because it cannot be satisfied, and the indiscriminate sending of e-mail has a problem that is rather unpleasant to customers.

The present invention has been made in order to solve the above problems according to the prior art, an object of the present invention, by accumulating and analyzing various personal information (age, gender, hobbies, personality, etc.) and past purchase information generated on the Web, consumers The present invention provides a complex filtering device and method for database marketing in an electronic commerce that recommends an item suitable for the user.

A feature of the complex filtering device for database marketing in the e-commerce according to the present invention for achieving the above object is, HTML showing the item column of the Web site selected by the consumer, and when clicking the item column shown in the HTML product and Merchant server that shows the service information in detail or provides a recommendation item to the consumer, Front module for collecting data for recommending items according to the taste of the consumer among the items provided by the merchant server, Collecting in the front module It is possible to purchase data according to the preferences of individual consumers by applying a combination of a database that scales and analyzes the collected data to a database according to consumer preferences, and the data output from the front module and the database by using a combination of collaborative filtering and content-based filtering algorithms.A hybrid filtering providing the most probable item, a connection manager managing an operation state of the hybrid filtering, a client managing the connection manager, and an item having the highest preference of the consumer data recorded in the database. It includes a content creator to output through the module and a mail agent for providing the item output from the content creator to the consumer via e-mail.

In addition, a feature of the complex filtering method for database marketing in the electronic commerce according to the present invention for achieving the above object is, in the electronic commerce transactions transactions of goods or services electronically with a merchant server to obtain information about the item Inputting personal information, and then comparing the inputted personal information to present the user with a sample of the degree of affinity for the item through collaborative filtering and content-based filtering of hybrid filtering. Analyzing and deriving the information derived therefrom, i.e., storing a sample item in a database, collecting data on the activities of the users (product search, shopping cart, purchase, cancellation, item preference, etc.) on the e-commerce site, Processing and storing the data in the database; Calculating primary item preferences from factors for explaining the relationship between item preferences and item components, calculating an error between a user and other users in the primary item preferences calculated in the step; Calculating the similarity between users using the Pearson correlation coefficient between the users using the error, and calculating the similarity with the average value of the errors that other users have about the item for the item resulting from the error. Calculating a predicted error value using a value, calculating a first predicted item preference and a predicted error value for an item for which the item preferences are not accumulated, calculating and storing a final preference for each item, and Database with the highest item preference for each user Extracting from the service provider and recommending the user to the user, and displaying the recommended goods and services through a merchant server or sending an e-mail through a content composer and a mail agent.

1 is a block diagram showing a complex filtering device for database marketing in an electronic commerce according to the present invention,

2 is a flowchart illustrating a complex filtering method for database marketing in an electronic commerce according to the present invention.

<Explanation of symbols for the main parts of the drawings>

10: HTML, 12: Merchant Server,

14: front module, 16: database,

18: collaborative filtering, 20: content based filtering,

22: Hybrid filtering.

Hereinafter, an embodiment of a complex filtering device and method for database marketing in an e-commerce according to the present invention will be described with reference to the accompanying drawings.

1 is a block diagram showing a complex filtering device for database marketing in an electronic commerce according to the present invention.

As shown in FIG. 1, a hypertext markup language (HTML) 10 showing an item (product and service information) column of a website selected by a consumer (for example, LG Home Shopping, Samgu Shopping, Daum, Daum Tech, etc.) And a merchant server 12 that displays the item in detail or provides a recommended item to a consumer when the item column shown in the HTML 10 is clicked, and an item provided by the merchant server 12. Front module 14 for collecting personal information (age, gender, hobbies, personality, etc.) and personal past purchase information in order to recommend product and service information according to the taste of consumers (degree of preference), and the front module ( 14) accumulate and analyze the data collected in the database and database the data according to the consumer's preference, and output the data output from the front module 14 and the database 16. Applying multiple algorithms of Collaborative Filtering (18) and Content-based Filtering (20) (Model of Equation 1 to explain the relationship between item preferences and elements of each user) After calculating how the factors affect the item preference, calculate the error using the formula of Equation 2 with respect to the data observed the actual item preference, and to the data that can not observe the item preference After predicting the error of missing values using the formula of Equation 3, the final secondary item preference is calculated using the formula of Equation 4.

: Primary item preference prediction value.

A component that determines item preference.

: Parameter.

: Error value between actual item preference and estimated item preference.

: Actual item preference.

: Estimated item preference.

: Estimated error result.

: Average value of user error.

: Weight (Pearson's correlation coefficient) indicating similarity of error.

: Constant that normalizes the sum of weights.

: Error on the item.

: Average of the error for the item.

: Secondary item preference.

: Primary item preference.

A hybrid filtering 22 that provides an item most likely to be purchased according to an individual consumer's preference, and an access manager managing an operation state of the hybrid filtering 22. 24, a content creator 28 for outputting, through the front module 14, the client 26 managed by the connection manager 24 and the item with the highest preference of the consumer, which is data recorded in the database 16. ) And a mail agent 30 that provides the consumer with an item output from the content creator 28 via e-mail.

The Collaborative Filtering 18 of the hybrid filtering 22 uses Pearson correlation, significance weightung, selecting neighborhoods, etc. based on collecting the preferences that consumers have for the item. Can predict a preference for an item that has not been checked.

The content-based filtering 20 of the hybrid filtering 22 uses an X variable (content) that determines a preference for an item, based on a traditional regression model, which is a traditional statistical method. Predict the preferences that consumers have for each item.

In particular, the hybrid filtering 22 combines and encompasses collaborative filtering 18 and content-based filtering 20, and the hybrid filtering in the hybrid filtering 22. Failure to implement Collaborative Filtering 18 results in the same results as Content-based Filtering 20.

In addition, if content-based filtering 20 is not implemented in the hybrid filtering 22, the same result as that of collaborative filtering 18 is obtained.

The operation of the complex filtering device for database marketing in the e-commerce configured as described above is as follows.

First of all, after accessing a web site desired by a consumer, personal information (age, gender, hobbies, personality, etc.) is input in an electronic commerce where a transaction of goods or services (hereinafter, items) is electronically traded.

The consumer then clicks on the product and service information (record, book, video, general commerce, movie, restaurant, travel destination, etc.) provided by the merchant server 12.

The personal information of the consumer is stored in the database 16 through the front module 14, and the collaborative filtering 18 and the content-based filtering of the hybrid filtering 22 are performed. The filtering (20) algorithm is applied in combination to add product and service information that is most likely to be purchased according to the preferences of individual consumers.

At this time, the product and service information with the highest probability of purchasing are stored in the database 16 or shown to the consumer through the front module 14 and the merchant server 12. In other words, a sample of the degree of preference (hereinafter, item preference) that the user has for the item is presented to the user.

The consumer then clicks, for example, "add to cart" to purchase the recommended goods and services, and delivers the product to the consumer in the shopping mall in accordance with the procedure and payment confirmation for the payment of the goods between the shopping mall and the consumer. E-commerce is complete.

A complex filtering method for database marketing in an electronic commerce according to the present invention corresponding to such an operation will be described with reference to FIG. 2.

First, a consumer accesses a web site implemented by a merchant server 12 or the like to obtain information on goods and services, and then inputs personal information (age, gender, hobby, personality, etc.) (S101).

Subsequently, the inputted personal information is compared so that the user can present a sample of the degree of affinity that the user has for the item through the collaborative filter filtering 18 and the content-based filtering 20 of the hybrid filtering 22. Analyze and store information derived therefrom, that is, sample item in database 16 (S103).

In addition, data about activities (e.g., product search, shopping cart, purchase, cancel purchase, item preference, etc.) performed by the user in order to obtain an item in the e-commerce are collected, processed, and stored in the database 16 (S107). .

Then, factors for explaining the relationship between the item preference of each user and the component of the item are extracted, and the primary item preference factor, which affects how much the factors affect the item preference, is calculated using the model of Equation 1 (S109). ).

Here, the calculation is performed through the user's sample item preference and purchase activity data. If there is no sample item preference and purchase activity data for the user, the calculation is performed by the following method.

First, using the average value of other users' item preferences,

Second, the regression analysis method is applied to the user's personal information and specific item factors.

Third, using the cluster analysis method for the user's personal information and specific item factors,

Subsequently, an error is calculated using the calculation formula of Equation 2 with respect to the data of the primary item preference and the actual item preference calculated in the above step (S109).

Then, the similarity between the users is calculated using the Pearson correlation coefficient between the users using the error (S111).

Also, for the item represented as the result of the error, a prediction error value is calculated using the calculation formula of Equation 3 from the average value of the errors that other users have with respect to the item and the calculated similarity value (S113).

Then, the final preference for each item is calculated and stored by using the calculation formula of Equation 4 as the sum of the first prediction item preference and the prediction error value for the item for which the item preference is not accumulated (S115).

Thereafter, an item (product and service) having the highest item preference for each user is extracted from the database and recommended to the user (S117).

Accordingly, the recommended goods and services are shown through the merchant server 12 or sent via e-mail through the content composer 28 and the mail agent 30 (S119).

As described above, the present invention can expect substantial increase in sales because it analyzes individual consumer information and consumption tendency and recommends product and service information having the highest probability of purchase.

In addition, by analyzing the consumption propensity or purchase type of consumers visiting the shopping mall site in real time to predict and provide high-preferred product information, the shopping time is shortened and the shopping mall can be conveniently used.

In particular, the present invention can maximize the consumer satisfaction by providing personalized information, products, and services that meet the needs of consumers by accurately understanding the characteristics of the current consumer, and based on this, secure a two-way communication channel with the consumer to It is effective in maximizing life value.

In addition, by analyzing the information of individual consumers in real time, there is an effect that can use a differentiated marketing strategy according to the characteristics of each consumer through pin-point communication according to the needs of the consumer.

In addition, it can be applied to information providing sites that require a recommendation system such as film, travel, as well as records, books, videos and general electronic commerce.

Claims (3)

HTML showing an item field of a web site selected by the consumer; A merchant server for displaying goods and service information in detail or providing a recommended item to a consumer when the item column shown in the HTML is clicked; A front module for collecting data in order to recommend an item according to a consumer's taste among items provided by the merchant server; A database that scales and analyzes the data collected by the front module to make a database in accordance with consumer preferences; Hybrid filtering for recommending items having the highest probability of purchasing data extracted from the front module and the database according to a preference of individual consumers by applying a combination of a collaborative filtering and a content-based filtering algorithm; A connection manager for managing an operation state of the hybrid filtering; A client managing the connection manager; A content creator for outputting the item having the highest consumer preference data recorded in the database through the front module; Complex filtering device for database marketing in electronic commerce, characterized in that it comprises a mail agent for providing the item output from the content creator to the consumer via electronic mail. Inputting personal information in an electronic commerce transaction in which goods or services are traded electronically with a merchant server to obtain information about the item; Then, through the collaborative filleting and content-based filtering of hybrid filtering, the inputted personal information is compared / analyzed and the information derived therefrom, so as to present the user with a sample of the degree of affinity for the item. Storing the sample item in a database: Collecting, processing, and storing data on activities of the users (product search, shopping cart, purchase, cancel purchase, item preference, etc.) performed in the e-commerce site; Calculating a primary item preference from factors for explaining the relationship between each user's item preference and item component; Calculating an error between the user and other users in the primary item preference calculated in the step; Calculating similarity among users using the Pearson correlation coefficient between users using the error; Calculating a prediction error value for the item represented as a result of the error using a mean value of the errors that other users have with the item and a calculated similarity value; Calculating a first predicted item preference and a predicted error value for an item for which the item preferences are not accumulated, and calculating and storing a final preference for each item; Recommending to the user by extracting an item having the highest item preference for each user from the database; And displaying the recommended products and services through a merchant server, or sending them through an e-mail through a content composer and a mail agent. The method of claim 2, In calculating the primary item preference factor, Calculating an average value of item preferences of other users when there is no sample item preference and purchase activity data for the user; Calculating the demographic data of the user and a specific item factor by applying a regression analysis method; Comprising the step of calculating the demographic data of the user and a specific item factor using a cluster analysis method comprising a complex filtering method for database marketing in e-commerce.