Profit-Analyse Based Multi-level Association Rules Research

In this competitive era, most enterprises use computers to improve their performance. Data mining can be used to find useful information that is hidden in the database. The technique of data mining is quite a popular research field nowadays ,especially in the study of association rules. Base on association rules some information in the database can be discovered, for example customer purchasing patterns. How to find out association rules effectively has been a critical topic. A lot of methods have been published, but most algorithms are dealing with single-level association rules. Because of tremendous number of items involved and the data were not highly concentrated; the number of association rules mined might not be enough, that some important information in the database would be cloaked. In generating multi-lever association rules, data are transformed form the multi-level architecture to a corresponding format. Using popular single-level association rules such as Apriori algorithm, it is easy to generate all itemsets but database needs to be searched many times. Data mining is often to deal with large database, it is a waste to scan target database repeatedly. So that we use pincer search algorithm for searching long frequent itemsets instead of the Apriori for efficiency. From the literatures, we believe an effective algorithm in generating multi-level association rules including profit of sales and quantity of item purchased will meet the needs of decision makes in all marketing department.In our experiment,it is proved that with the data scales and levels growing and the data class increasing, multi-level pincer search algorithm shows more and more evident advantage in efficiency compared with Aprior algorithm.In this paper ,we add quantity and profit facts in analyzing association ruIes.Simply speaking,if some low profit item support the sale of another high profit item,we say it is a profitable association rule.Our new analysis method give more help than the traditional Maket Basket Analysis.