The Target Attribute-based Association Rule Mining

The discovery of association rules is a very important aspect in data mining. Many algorithms for association rules have been proposed. One of them, them most famous is Apriori algorithm presented by R Agrawal. Apriori Algorithm belongs to indirect mining algorithm, and some improved algorithms of Apriori and other algorithms remain with indirect mining too. So, many people take the discovery of association rules as indirect data mining.However, in practice, different commodities have different characters. It's not suitable to set up a single support threshold. So we should give each commodity a support threshold, but this can't be fulfilled by indirect mining. In addition, multiplayer association rules' mining is not suitable for indirect mining too. Thus, we propose the way of association rules' mining based on target attribute.In this article, an algorithm of association rules' mining based on target attribute is presented, which is improved from Apriori algorithm. Although this algorithm will bring some repeated computation, it has the characteristic of parallel that every target attribute's mining is independent. When parallel computation is adopted, the mining efficiency is greatly improved, and this is proved by experiment.On condition that the way based on target attribute is adopted, many methods used in classification problem such as genetic algorithm, decision tree, and neural network now can be used for association rules' mining. In this article, the applications of genetic algorithm and decision tree to association rules' mining have been discussed limitedly. In this aspect, there is still much study work needs to do.