The Technique Of Confident Rule Mining In Data Mining

How to get useful data from kinds of information is a main research field of computer, Data Mining is just an important research content of this field. Data Mining concerns multiple subjects and is a newly-established frontier subject. It is being used extensively and its application future is bright.This paper studies and analyses the Data Mining technique systematically and deeply, especially for rule mining, and propounds basic concept of "confident rule" and mining methods. The main contents are listed as follows:Research some association rules mining techniques in existence. Because all these mining techniques are based on support to perform prune. In this pruning there is a problem is that some rules with high support have low confidence, while rules with high confidence are pruned because of low support. So it is necessary to research another technique to mining all rules with high confidence without taking support into account. The necessity to research this technique is that: rules with very high support are obvious and well-known, so mining these rules is useless some times;on the other hand, the rules with high confidence and low support are always ignored, so mining these rules is valuable and always can get some beat all results.In order to resolve the problem mentioned above, we propound the concept of "confident rules". We propound two different techniques as follows:The first one is based on MFC principle. This technique is a level-wise search process. It is based on the especial property of confidence, existential upward closure.The other one is based on the similarity between items. First find all pairs of item with high similarity, and then generate confident rules. Several methods to find pairs of item with high similarity mentioned in this paper are: MH, K-MH, M-LSH, and H-LSH. These four methods all have their own advantages and disadvantages, and we can choose different method according different need.