Association Rule Mining Algorithm

Data mining technology has been a hot task in the field of database and artificial intelligence in recent years, and it has attracted extensive attention in science and technology industry. Association is mainly used to search for relationship among items in a data set, and it is one of the major issues in data mining. Association rules can be extensively applied in every field, and it not only can test the knowledge modes existing in the industry, but also can find some new rules hidden. One of the important ways to finish the task of data mining is to find effectively, apprehend and apply association rules.Firstly, the dissertation introduces data mining technology synoptically. Then after presenting the basic algorithm for mining association rule—Apriori, the dissertation analyses its performance, and induces and analyses some typical algorithms for mining association rules in detail relatively.Next, the dissertation introduces an algorithm for mining association rules based association graph, analyses its performance, points out its defects, then puts forward two improved algorithms—one based on complete sub-graph and another based on ordered tree. Concerning the defects of setting the minimum supports occurring in the existing algorithms, the dissertation brings forward a method of setting minimum supports of the items; based on the algorithm for mining association rules with multiple minimum supports using minimum constraints, the dissertation brings forward an improved algorithm based tree; based on the algorithm for mining association rules with multiple minimum supports using maximum constraints, the dissertation brings forward an improved algorithm based complete sub-graph. In the end, concerning the defects of the algorithm for mining association rules when dealing with batch purchase, the dissertation puts forward an imagination of an algorithm for mining batch association rules.In the improved algorithm based complete sub-graph, on the basis of the connection between complete sub-graph and large itemsets, the algorithm regards the degree of the nodes in complete sub-graph as a criterion for avoiding the compare among the items which needn't been compared. At the same time, by setting a different order for each node, the algorithm avoids some repeated compare for the same large itemsets. So during looking for k-large itemsets (k≥3), the time the new algorithm needs is much less than 1/(k-1) of the time the former needs. As a result, the algorithm occupies less memory and quickens the speed of mining, so it improves the efficiency of mining.In the improved algorithm based on ordered tree, after numbering the items, the ordered frequent 2-itemsets can be obtained, and the ordered frequent tree can be constructed based on the ordered frequent 2-itemsets. By setting a different order for each node, the algorithm avoids some repeated compare for the same large itemsets. By analyzing an example, the algorithm quickens the speed of mining, so it improves the efficiency of mining.