Research And Application On The Exception Rule Mining Algorithm

This paper researches on association analysis of data mining, discussingmining algorithms of positive association rules and negative association rules andexception rules. Based on the study mining algorithm of positive and negativeassociation rules, this paper studies the Eclat algorithm and then try to improve it.Simultaneously, it proposes the definition of exception rule, its sub-problemsegments, evaluation parameters, and mining algorithm. Data set through the chessExperiment on chess data set shows that EclatP is superior. Experiment on miningexception rules illustrates that algorithm is effective, and it can obtain valuableinformation. This study is a meaningful and necessary work.This article studies on the following aspects, and have achieved good results.(1)Systematic study on the theoretical basis and classification of associationrules mining. The way for classification is different from the previous thatexception rules introduced association analysis.(2)Study on the positive association rules and negative association rulesalgorithm, we use the new classification method, which not only achieve a morecomprehensive purpose, but also make people at a glance.(3)This paper focus on Eclat algorithm of mining association rules, carefulstudy on the theory and ideas of the algorithm, analyze its advantages anddisadvantages. This paper study EclatP, and EclatM based on the defect of Eclat,and test on the two improved algorithms. The results showed: the numbers andcontents of frequent item sets of the different mining algorithm are same, there aresome differences in the efficiency of the algorithm. EclatM is affected by thelength of item sets, EclatP is not better than Eclat in the sparse data sets, but inintensive data sets mine more quickly and better.(4)Research exception rules and define exceptions rules, partition the problemand study theory. EclatP mines association rules, deriving comparison associationrules, obtaining candidate exceptions rules. Set the interestingness threshold withseveral common definitions and this paper definition for interestingness and teston the medical data set. Combined with the practical needs, mining exceptionsrules related with awareness and so on.The results indicate that several common definitions have different strengths and weaknesses, the definition given in thispaper vary widely and can better distinguish and select candidate exception rules.Exception rules mining by my algorithm can basically meet the user needs andprovide users with useful information.The purpose of this study is to satisfy the practical application. In thecircumstances which an increasingly large database and data is not obvious, thisexception rules of the application on medical data is a valuable study and providesome useful information to the medical workers, promote the further application ofmathematical theory also.