The Algorithm Research Of Association Rules Mining Based On Computational Intelligence

In recent years, many people in information industry pay more attention to the data miningtechnique, which is the necessary result of the conflicting movement between therapid-increasing data and the lack of information day by day. Researching the data miningtechnique systematically, deeply and detailedly is an objective requirement of the globalinformation development. One important area of many researching areas in data mining is theassociation rules (AR), which has great value in application, especially in businessdecision-making. This thesis is focus on the correlative study of AR mining.According to the traditional AR mining algorithms, such as Apriori algorithm and itsimproved algorithms are all focus on the mining in accurate concepts, and can not mine theinaccurate or fuzzy concepts. At the same time we can find that the main computation is thegeneration of frequent 2-itemsets in searching frequent itemsets by experiment, the generatingprocess of frequent 2-itemsets is the bottleneck of the Apriori algorithm mining, aimed at thiscircumstance, a kind of new AR mining algorithm based on fuzzy sets is presented in the paper.In this algorithm we introduce the fuzzy theory and the concept of semantics AR, carry throughthe inaccurate semantics conversion to the number property of database soundly, then improvethe efficiency of algorithm by improving on the prune part being scanning by itemsets' size, itavoids the increasing current according to exponential of the set length being scanning.Because the core issue of Apriori algorithm is finding maximal itemsets, the process offinding maximal itemsets is a global searching process, while genetic algorithms(GA) is a kindof global optimizing algorithm, it avoids the local optimization. So GA being used for the findingand distilling rules can find the useful rules. According to this, this paper brings forward the ARmining algorithms based on GA, it mainly mine quantitative AR, the algorithm mainly includesthe coding method designing of AR, the constitution of fitness function and the amelioration ofgenetic operator and so on.According to the two kinds of AR mining algorithms based on computational intelligencewhich are put forward and designed in the paper, the author takes the iatrical database andstudent database as the mining prototype separately to carry through the pick-up of AR, andanalyze the experimental result, which validate the validity of algorithms, and also explain thebroad application foreground of AR mining.