KR20020045302A - Method of merchandise management in electronic commerce - Google Patents

Abstract

PURPOSE: A method for the product management on the EC(Electric Commerce) is provided to offer the credible sales prediction information by systematically analyzing the data accumulated in an EC company and to carry out the marketing and the product display through the sales prediction. CONSTITUTION: The product management method comprises the steps of deleting the unnecessary data in a web server log and converting a data format of the web server log into the data format necessary to the analysis, extracting the sales information by using the converted data and extracting a temporary sales product and quantity through the sales information, and extracting a use pattern of a buying customer by using the converted data and storing the pattern in a buying pattern DB. A search pattern of the searching customer is extracted by using the converted data and a latent buying customer and product is extracted by comparing the extracted pattern with the pattern stored in the buying pattern DB. A final sales prediction product and quantity is predicted by adding the latent sales product and quantity to the temporary sales product and quantity.

Description

Merchandise management in e-commerce {METHOD OF MERCHANDISE MANAGEMENT IN ELECTRONIC COMMERCE}

The present invention relates to a product management method in electronic commerce, and more particularly, to a product management method in electronic commerce using web mining.

Electronic commerce refers to all transactions in which commerce, that is, exchange of goods, services, information, and money, is performed by an electronic medium using Internet technology.

Stores on-line rather than on-line, such as the Internet, can only perform analysis using localized data such as user personal information and sales data through partial computerization, and only within a partially implemented information system. As the analysis takes place, there is limited information available. That is, only information when the customer generates sales, such as customer information and sales information, can be used. This limited information makes it difficult for accurate user pattern analysis, sales forecasting, and inventory management.

In addition, in the conventional offline store, since only a fractional prediction through sales pattern analysis is performed when a sales prediction is to be performed, there is a problem that cannot be predicted with respect to the demand of potential customers.

The present invention provides reliable sales forecasting information by systematically analyzing data accumulated in an e-commerce company between a company and a consumer, thereby performing appropriate inventory management to optimize inventory costs, marketing and products through sales forecasting. We want to provide a commodity trading system that performs the display.

1 is a view showing the flow of data analysis in the product management method according to an embodiment of the present invention,

2 is a diagram illustrating a process of forecasting sales items and quantities using sales information data;

3 is a diagram illustrating a process of predicting a potential sale product and quantity using search path data;

4 is a diagram illustrating a final sale product and quantity prediction process.

In order to achieve the above object, the present invention searches not only the user's personal information and sales data, but also the user's path analysis using web mining technology that can scientifically analyze the Web server log. Analyze usage patterns such as patterns and purchasing patterns.

The present invention includes a first step of preliminary processing of deleting unnecessary data from a web server log and converting the data format of the web server log into a data format for analysis; A second step of extracting sales information by using the data converted in the first step and extracting a provisional sale product and quantity through this; A third step of extracting a usage pattern of a purchase customer by using the data converted in the first step and storing it in a purchase pattern database; A fourth step of extracting a search pattern of a search customer by using the data converted in the first step and extracting a potential purchase customer and a product by comparing the extracted pattern with a pattern stored in the purchase pattern database; A method of managing a product in an e-commerce transaction comprising five steps of predicting a final sale expected product and quantity by comparing the potential sales product and quantity extracted through the second step with the potential sale product and quantity extracted through the fourth step. to provide.

In this case, the data format converted through the first step may be purchase information data, purchase path data, and search path data.

Here, the second step is to analyze the purchase information data generated in the first step by using clustering, association and sequencing procedures of the web mining technique, and to predict the potential sales goods and quantity using the third step. Analyzing the purchase path data generated in the first step using clustering, association, and sequencing procedures of the web mining technique, and storing the extracted pattern in the purchase pattern database. The search path data generated in the step is extracted through the sequencing of the web mining technique, and the extracted pattern is compared with the patterns stored in the purchase pattern database to predict potential purchase customers, products, and purchase quantities. The fifth stage is the provisional sales goods and quantity predicted through the second stage and the fourth stage. The combined potential sales side items and quantities that may be to predict the final sale items and quantities.

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

Through e-commerce between corporations and consumers, user search, purchase, and store usage information is already stored in web server logs. Therefore, if you can systematically analyze web server logs, you can analyze user search, purchase, and store usage patterns, and use them to implement functions necessary for product management such as reliable sales forecasting, inventory management, and product display. Can be. The present invention applies a web mining technology that can systematically analyze web server logs, extracts information necessary for product management that is not provided by the prior art, and builds a reliable product management system using the same.

Web mining, also known as data mining, is a technique for discovering patterns buried between large amounts of data and inferring rules to support decision making and predict the effects. Web mining technology penetrates deep into databases, discovering patterns in data, and inferring rules, which can be used to support decision making and predict changes in the environment.

In the present invention, clustering, association, and sequencing are used in a web mining technique. "Clustering" is a method of grouping objects with similar tendencies, "Association" is a method of extracting non-time series associations of static relationships without considering time between objects, and "Sequencing" is a dynamic relationship considering time. This method extracts the time series correlation of.

The present invention builds a merchandise management system that performs systematic analysis using the accumulated data by maximizing the advantages of electronic commerce between companies and consumers accumulating vast amounts of commerce usage information.

Figure 1 shows the flow of analysis constituting a product management system according to an embodiment of the present invention.

First, in the first step, unnecessary data is deleted from a web server log storing data generated in an e-commerce process, and a preprocessing process is performed to convert the data format of the web server log into a data format for analysis.

This pretreatment is performed prior to performing the analysis.

The preprocessing process consists of a data cleaning process and a data transformation process.

The data cleaning process is a process of extracting and deleting unnecessary log data included in a web server log. The data needed during the analysis is the information the customer uses the store. However, the web server log contains a large amount of unnecessary data such as unrelated graphic information and access failure information. Therefore, such unnecessary information must be deleted first.

After this data cleaning process, the data is called access history, which contains information about who accessed the company's web site and where they made their purchases. .

The data conversion process is a process of transforming the log data after the data cleaning process into the analysis data format. The actual data format of the web server log is difficult to analyze. Thus, prior to analyzing the usage history of the customer, the log data is converted into a format that the analysis program can easily handle. At this time, the converted data is the purchase information data including only the sales information of the product actually generated, the purchase path data using the store until the actual purchasers purchase, and the store use path of the customers who have not made a purchase yet. It may also be divided into data such as search path data.

In the second step, the purchase information data is analyzed among the conversion data generated during the preprocessing to predict the potential goods and quantity.

In the process of analyzing purchasing information data, we analyze the association between products and customers through clustering technique, analyze the non-time series static association between products and products through association method, and time series dynamic association between products and products through sequencing technique. Analyze these to make tentative assessments and predictions.

In other words, product grouping and classification are performed by classifying products into product groups having similar sales tendencies using characteristics of purchase customers through clustering techniques. In addition, through the association technique, the sales propensity of the product in the state without considering the pro- and post-relationship of the purchase is analyzed to extract the related products. And, through the sequencing technique, we analyze the sales tendency of the product considering the pro-relationship of the purchase during the analysis period to extract the continuous pattern of purchase.

By analyzing such purchase information data, it is possible to grasp the sales trend of the product during the analysis period, and to predict the prospective sales trend later. That is, the sales information is extracted by using the purchase information data, and the potential sales goods and quantity are extracted through this.

The detailed process of predicting the sale goods and the quantity using the web mining technique is shown in FIG. 2.

Through the sequencing process of the sales path data, the preceding and following products can be extracted, and the factors influencing the sales volume of the preceding products can be obtained. In other words, it can be extracted that "the sales volume of the next season B products is 50% of the sales of the A products this season". Therefore, the sales volume of the preceding product can be predicted using the sales volume of the preceding product and the sequencing factor, so the primary inventory product and quantity are determined using this.

In addition, through the association process, it is possible to obtain a correction factor representing the correlation of sales between products of the same period. In other words, it can be extracted that "the sales volume of the next season B products is 70% of the sales of the next season A products". Therefore, the quantity of the product selected as the primary inventory and the correlation factor can be used to determine the sales volume of the products that are related to each other. Therefore, the secondary inventory and the quantity of the product are determined. The ultimately predicted sale product and quantity is determined by summing up the primary inventory and quantity and the secondary inventory and quantity.

In a third step, the purchase path data is analyzed from the conversion data generated during the preprocessing, and the usage pattern of the purchase customer is extracted and stored in the purchase pattern database.

The analysis process of purchase path data is the same as the analysis process of purchase information data. In other words, we analyze the potential product-to-customer association through the clustering technique, the non-time series static association between the product-to-product through the association method, and the time series dynamic association between the product-to-product through the sequencing technique. The difference is that the analysis of the purchase information data is used to predict the sales volume of the next season based on the relationship between the products using the sales volume of the products already sold this season, and the analysis of the purchase channel data is unique to the customers who have purchased. By analyzing and storing store usage patterns, it is used as basic data to extract potential customers for next season's purchase. Therefore, we store in the database a correlation factor representing the static association between products of the same period from the association technique and a sequencing factor representing the dynamic association between the products from the sequencing technique.

By analyzing such sales path data, it is possible to extract a specific usage pattern using the store until the actual purchasers make a purchase. The extracted usage pattern is later stored in a purchase pattern database for potential customer extraction.

The next step is to analyze the search path data among the conversion data created during the preprocessing process, extract the usage pattern of the search customer, and compare the extracted pattern with the patterns stored in the purchase pattern database to identify potential purchase customers and products. Extract.

In the process of analyzing the search path data, the search pattern of the prospective customer is extracted through the sequencing process, the extracted pattern is compared with the usage pattern of the purchasing customer stored in the purchase pattern database, and the matching pattern is extracted. The customer who owns the extracted search pattern becomes a potential purchaser, and through this potential purchaser, predicts potential sales products and quantities.

This potential sale product and quantity prediction process is shown in FIG.

As such, by analyzing the search path data, a search pattern for a specific store that a search customer is searching may be extracted. The extracted search pattern is compared with the usage pattern of the purchasing customer stored in the purchase pattern database to extract a pattern matching each other. Customers with this extracted search pattern can be selected as potential purchase customers. The potential buyers and volumes are predicted through the selected potential buyers.

Next, in the fifth step, the final sales anticipated product and quantity are predicted using the potential sales product and quantity extracted in the second stage and the potential sales product and quantity extracted in the fourth stage. The final sales product and quantity forecast is determined by combining the results of the prospective sales product and quantity forecast using the sales data and the results of the potential sales product and quantity forecast using the search path. Final sale goods and quantity prediction process is shown in FIG.

In other words, the prospective inventory management products and quantity predicted by analyzing purchase data of the purchasing customer in the second stage and the prospective purchased products and quantities predicted by analyzing the navigation route data of the searching customer in the fourth stage are finally inventoryed. Determine the managed goods and quantity.

In this way, the present invention extracts the potential purchase customers by analyzing the sales pattern of the purchase user, as well as the search patterns and usage patterns of the customers who have not yet purchased using web mining, and thus the potential sales products and quantities Predict. The type and quantity of goods to be managed by inventory are determined based on the predicted sale goods and quantity, and the products having non-time series and time series relations between the products can be displayed in connection. In addition, by comparing the customer's search pattern with the purchase customer's search pattern, it is possible to extract potential customers and predict the time of purchase, thus enabling one-to-one marketing, which is a marketing policy specific to each customer.

As described above, the present invention can provide reliable sales forecasting information by systematically analyzing data accumulated in an e-commerce company between a company and a consumer, thereby performing proper inventory management, thereby optimizing inventory cost and selling. Predict marketing and product display.

Claims (5)

A first step of preliminary processing of deleting unnecessary data from the web server log and converting the data format of the web server log into a data format for analysis; A second step of extracting sales information by using the data converted in the first step and extracting a provisional sale product and quantity through this; A third step of extracting a usage pattern of a purchase customer by using the data converted in the first step and storing it in a purchase pattern database; A fourth step of extracting a search pattern of a search customer by using the data converted in the first step and extracting a potential purchase customer and a product by comparing the extracted pattern with a pattern stored in the purchase pattern database; A fifth step of predicting the final sales expected product and quantity by adding the potential sales product and quantity extracted through the second step and the potential sales product and quantity extracted through the fourth step Commodity management method in the e-commerce comprising a. In claim 1, And a data format converted through the first step is purchase information data, purchase path data, and search path data. In claim 2, In the second step, the merchandise management method in the electronic commerce which analyzes the purchase information data generated in the first step using clustering, association and sequencing procedures of the web mining method and predicts potential sales items and volumes using the same. . In claim 1, In the third step, the purchase path data generated in the first step is analyzed using clustering, association, and sequencing procedures of the web mining technique, and the extracted pattern is stored in the purchase pattern database. How to manage your goods. In claim 1, In the fourth step, the search path data generated in the first step is extracted through sequencing of the web mining technique, and the extracted pattern is compared with the patterns stored in the purchase pattern database. Product management method in e-commerce to predict purchase volume.