Research On Association Rules In Retail Based On An Integrated Measure

The association analysis is a chief method of merchant behavior, and most of association analyses base on the architecture of support-confidence, and some researchers use interest as an important standard of analysis process. However, these methods may lead to conflict patterns, and unsatisfactory generalization. So far, there has no standard or approved method to judge interest of rules. On the other hand, if using association analysis to merchant field, we should focus on how to apply these techniques in true process environment. The patterns of data can work to strengthen enterprise to a certain extent only when involved in its decision process, and be interesting. Unfortunately, tradesmen often complain that they don't know how to optimize Cross-Selling. All of above prevent association analysis from taking effect in decision process of retail business.The research bases on retail, focuses on how to distil useful and interesting rules from huge volume of data. Combining relational production, selecting the most important and universal algorithm apriori, founded on support-confidence measure architecture, the thesis introduces a series of synthesis structure, including interest, correlation, negative association analysis, computing for positive and negative rules simultaneously. At last , the author designs a new Synthesis Measure-Based Association Rules Mining Algorithm, then builds a mining model on this algorithm, realizes this model and verifies it.Chapter 2 introduces some basic definitions and background about the research of this field, illustrates the origin, definition and present situation of synthetic measure, and introduces the present research situation of this field inland and overseas. By comparison of works in this reseach field, the author finds out the capmarative blind spot of this field, and gradually turns her focus to synthetic measure. The author has done a lot of reaearches, which include how to improve the efficiency of algorithm in working at huge volume of data, the future research on mining model to mine rapidly renewed data, the way for the uses to interact in mining, the combination of the uses' knowledge in this field, how to deal with the numerical field in the association analysis, and the visibility of the result,etc.Chapter 3 begins with the definitions and the detailed condition of related algorithms and synthetic measure, and the association analysis based on the architecture of support-confidence, then it shifts the focus from the the efficiency enhancement to the enhancement of practicability, reliability and redundance elimination. The author forms the primary research structure of this thesis by introducing a series of algorithms and restrictions like the original association analysis, interest.Chapter 4 is the most important part of this thesis, talking about the author's productions on this research, and Synthesis Measure-Based Association Rules Mining Algorithm. Beginning with the specific characteristcs of retail business, this author selects algorithm and restrictions step by step by thinking reasonably and logically, like the selection of interest, the actual requirement of algorithm, the elimination of redundance, and the enhancement of reliability, etc. The chapter ends with the introduction of Synthesis Measure-Based Association Rules Mining Algorithm. The author analyzes and validates SMBA, and uses it in the mining engine of CRM model, aiming at improving the decision support . Data mining results from decision support which most retail businessmen encounter. The development of bar-code technique enables retail businessmen to assemble and deposit huge volume of sales data, which are called basket data. A piece of record is commonly made up of the transaction data and the item in the trade, that is the article. More commonly, the data record also includes the customer-id, especially when the buyer is the customers with credit card or voucher card of the shop. These all enable retail businessmen to relate a purchasing behavior with a certain customer.Chapter 5 introduces the architecture and the working theory of SMBA. It also makes a series of evaluations combing SMBA with project. By analyzing the result of true data, this chapter offers the final rule under various circumstances, effectiveling proving the the reliability and practicability of SMBA model, which founds the basis for integrating it with the CRM model in enterprise.The result of the experiment proves that this algorithm is practical and can solve problems such as inconsistency and lower reliability of rules.