The Study Of The Algorithm Of Association Rule Mining Applied In Retail Data

With the explosive growth of transaction data in retail and the rapid development of information technology, the application of Business Intelligence in retail is becoming more and more important in Information System and E-commerce. The market basket analysis is the one of most effective ways of association rule data mining technical applied in retail marketing. Its aim is to find out the possibility of purchasing homogeneous goods additionally by customer, which is convenient for suggesting the arrangement of the goods to improve the efficiency of the sales promotion.At the beginning, the association rule mining algorithm can only deals with Boolean attributes, and the mining results are also Boolean association rules. The typical algorithms include Apriori and FP-growth. However there are many kinds of data types in the database of retail, e.g.:numeric and enumerated types etc. Therefore, the traditional Boolean association rule mining algorithm which named as Apriori is no longer applicable. So it is necessary to study the technology of Quantitative Association Rule. The key step of applying the Quantitative Association Rule into retail business is how to discrete numerical data in preprocess. This paper will introduce a method of how to successfully apply the Mining Algorithm of Fuzzy Clustering Quantitative Association Rules into the retail data.According to the research hotpot of Quantitative Association Rule, the U.S. retail DMEF (Direct Marketing Educational Foundation) data base on the application background of market basket, investigation arm to find out a Quantitative Association Rule algorithm that suitable for the retail data mining in this thesis. The main contributions are as the following:First of all, this paper analyzes the representative data of DMEF, found out that there are some inherent characteristics of the data which are data redundancy, vacancies value random distributed, non-uniform distributed. And in the same time, this chapter described their implementations in business intelligence for each method according to their application occasions. (Detail in Chapter Three)Secondly, in the fourth chapter analyzes the representative data of DMEF, found out that there are random, non-uniform distribution and etc. characteristics for the data. In the same time, according to the task of Association Rule mining, this paper described the necessary preprocessing steps of analyzing the DMEF data, analysis the advantage and disadvantage. (Detail in Chapter Four)Finally, this paper combines fuzzy C-means algorithm and the classical Boolean association rule mining algorithm named as Apriori, introduces an improved association rule mining algorithm based on fuzzy cluster, and designs the framework and procedures to apply into the data of retail business to address the shortcomings of existing methods. This new algorithm has the following advantages. For one thing, it can reflect the characteristics of data distribution. For another, it can soften boundary of the property field, rationalize the discrete interval. What's more, it can help to solve the "too small confidence" and "too little support" problems. It can mine more valuable association rules. This algorithm can guide the strategic decision of the retail business. (Detail in Chapter Five)...