Research Of Classification Based On Negative Association Rules

Association rules describe the co-occurring relationships among data item in the large transaction database. They have been extensively studied in the literature for their usefulness in many real world areas .So association rule mining is always a major topic in data mining research community. There have been discovered many efficient algorithms and their variants of association rules, such as: Apriori, FPgrowth, etc. Associative classification is a well-known technique which uses association rules to predict the class label for new data object.This model has been recently reported to achieve higher accuracy than traditional classificationapproaches like C4.5.Only limits in view of positive association rules used in classification.Classifier strategy is the method that based on NARs,and which is based on the classification of associationrules, in order to demonstrate more effective directconnection among realistic events. The union between the classification method and PARs can carry on the classification effectively and accurately, and also can analysis inner links among each kind of hided factor more comprehensively. The researcher based on the classification of NARs will certainly to be the hot spot in the future study. This paper introduces main methods and the present situation of classification based on association rules, expatiates the present situation and main technologies of negative association rules, introduces the problem of exploding frequent itemsets and points out the development tendency of classification based on negative association rules. The existing association rules mining algorithms are most based on frequent itemsets, the researchs about infrequent itemsets are very little. However when we study the negative association rules, infrequent itemsets become more and more important because they contain a lot important negative association rules. The hill climbing search method was used to improve the CB_PNARCC algorithm to get the highest classification accuracy of the set of support and confidence thresholds. This strategy can avoid the unreasonable set of threshold, and enhance the classification accuracy.