Research Of Positive And Negative Association Rules Mining Based On Intelligence Algorithms

Facing the challenges that people were flooded with data, but feel knowledge hunger, data mining and knowledge discovery were proposed and now develop vigorously. Association rules mining is one of the important forms of data mining. At first it discovered relationships among attributes from large databases. And later researchers improved and extended the problem prototype. Now association rules tend to reach perfection not only in principle but also in application. In reality, people may be interesting in such rules "when buy this thing, rarely buy that thing". This relation is called negative association rule. It has the same importance as positive association rule. Researchers put more and more attention to negative association rules. When mining negative association rules, if we use conventional criteria-"support-confidence", it may get redundant or false rules. So it is highly important to make effective measure criteria.Many data mining problems can be treated as searching problems. Datasets can be treated as searching space. Intelligence algorithms can be treated as searching strategy. As intelligence evolutionary algorithms, cultural algorithm and immune clone algorithm have global optimization ability. They have evolution learning mechanism such as unsupervised learning, resistance of noise and memory. So they provide a new method to solve problems. Because of the advantages of intelligence algorithms, association rules mining uses cultural algorithm and immune clone algorithm to seach frequent items, and then extracts association rules.In this paper, it mainly researches as follows:(1)It reviews the basic concepts and mining algorithms of positive and negative association rules and studies the shortcomings of the existing measure criteria when negative association rules mining. And it proposes a new measure criterion-effective criterion. It can mine positive and negative association rules and removes valid rules at the same time using this criterion. The experiments show the validity of the algorithm.(2)It puts forward an algorithm for mining positive and negative association rules based on cultural algorithm and immune clone algorithm. This algorithm combines cultural algorithm and immune clone algorithm and introduces in the new measure criterion. It could extract positive and negative association rules at the same time. The results of experiments show that the algorithm in this paper has a fast convergence speed. It can mine association rules effectively.(3)It researches the design ideas and main functions of the association rules mining system's prototype. This system can mine positive and negative association rules at the same time using the algorithms in this paper. It realized partly functions of the association rules mining system's prototype using mining algorithms proposed in this paper. It establishes scientific system architecture and combines mining algorithms with other modules in the system using the reusability and embedded ability of the mining algorithms.