Research On Retail Data Mining Based On Rough Set Theory

With the rapid progress of information technology, a large amount of historical data has been accumulated in retail enterprise. Many commerce enterprises have realized that it is important for development of the enterprise how to use this information. Since it is essential for decision-makers to possess commerce information, including customers,opponents,financial states of company, gaining reliable, exact and timely information in order to make right decision is enterprisers' urgent requirement. Thus this paper will illustrate the research on application of data mining techniques in the field of retail business.This paper first make a introduce about the purpose and currently application of data mining in the field of retail business, it mainly probes into the characteristics and difficult of data mining in current inside retail business; then discuss concept of Rough Set theory and its application in data mining; on the base of above studies, this paper presents a novel architecture of retail data mining based on rough set theory, and designs some data mining algorithms during implement of this architecture; at the end of this paper, an instance is used to verify results and algorithms' validity.The architecture of our paper's retail data mining system is composed of three levels: data preparation level, data preprocess level, mining evaluation level. It can effectively extract useful knowledge, feed back and revise by evaluation system and establish business knowledge base in order to provide decision information with high quality and decision support for decision maker. Rough Set can achieve major mission in the data mining process.In data preparation level, the main task is data collection, data defining and data select. Data collection deal with data by method of decision oriented data processing and refining to data warehouse, data defining is mainly to define following data: Commodity sales class, commodity stock class, commodity client class .etc. Data select is mainly to clean noise data and irrelevant data. In data preprocess level, this paper presents an improved discretized agorithm, according to studies a discretization algorithm of continue attributes in rough set based on information entropy, and add a user control parameter. Avoid the fault of excessively break-point. data filtering fill the lost data by the algorithm of expand initially quantification matrix. After data preprocess, we can get a complete information system which rule extract algorithm can...