The Application Research Of Association Rules Mining In Supermarket Marketing Analysis

Today's society has entered the era of information explosion, large amounts of information has become an important feature of the information society. However, how to make better use of the information, how to discovery knowledge and bring out value of massive data is an important issue facing mankind. The Apriori algorithm for association rules discovery was first proposed and applied in the domain of supermarket commerce. And then a large number of detailed information from transaction databases has been often treated by new approach of data mining technology. But so far China's supermarket retail enterprises have not used the accumulated information of customer sales data to engage in effective research and analysis, can not use the data mining method to provide valuable advise. This paper collected three months' transactions information from one store of a large supermarket chain in Harbin. Therefore, the rest of the paper applies association rules analysis for marketing decision-making in the supermarket.Firstly, this paper describes data mining and association rules algorithm related theories, as well as the basic knowledge of marketing and consumer behavior in the management of supermarket.Secondly based on the theory studies, this paper proposes the use of frequent items weighted assessment of profits to determine the value of association rules. Therefore three models are developed to"dig"out implicit, yet meaningful, patterns with profit information. These three models focus on cross-selling, shelves display analysis and competitive products analysis for effective decision-making in supermarket. Furthermore this study collected three months' transactions information from one store of a large supermarket chain in Harbin. Through the implementation of data mining processes in SEMMA model, all the calculations are realized by SAS Enterprise Miner module and SAS programming.Finally, as a result this study proposes several specific and feasible marketing recommendations for the supermarket involved. The research shows that although some patterns don't have a relatively high degree of confidence-support rate, but the profit generated is more interesting than those with high confidence-support. Thus, this method proves a certain rationality and superiority in using a weighted assessment of the profit for association rules analysis, and has strong promotional value in the actual business operations. This study has also provided the relevant supporting evidences for the feasibility of data mining projects in the nationwide supermarket chain.