Fuzzy Association Rules Extraction Based On Particle Swarm Optimization And Its Implementation In Parallelization

Association rules extraction is one of the most important parts in datamining, which is aid to find interesting correlations among data sets. Researchof association rules has spread to every field, with a variety of data typesappeared, such as quantitative, categories and so on. But using the generalrule extraction framework to extract association rules from these attributevalues tends to cause problems. Some researchs have introduced the conceptof fuzzy in order to solve such problems. In the extraction of fuzzyassociation rules, the sample sets are transformed into fuzzy sets firstly. Thenextracting the association rules. In spite of the data type is complicated, thehuge data quantity requires more strictly conditions, for example, I/Obottleneck, memory, and other limitations about hardware ressource. Soparallel association rules extraction has become one of the hot topics.The particle swarm optimization algorithm which is proposed byKennedy et al. in1995has developed nearly20years and become theimportant component of swarm intelligence theory. Since particle swarmoptimization algorithm has simple and clear in concept, fast and convenient inimplementation and other advantages, it has been widely researched, andapplied to the economic, social, biological and other fields.In conclusion, this paper describes and analyses the concept, definitionand research status of association rules and particle swarm optimization.Proposing the thinking of the particle swarm optimization as the searchingtool to extract the fuzzy association rules. This paper focuses on the followingresearch work content:Research and improve classical particle swarm optimization algorithm,and propose method based on variable search region adaptive particleswarm optimization algorithm to the shape error detection. First of all, thebasic concept parameters and performance of particle swarm optimization algorithm are studied and analyzed. Then, be aimed at the swarm in lowefficiency of search the multi-modal function, and easy to fall into localoptimal solution, using the index of inertia weight, symmetric accelerationfactor, the variable search region and other ways to improve the searchperformance of the population.Extraction model based on association rules, the defect of classicassociation rules mining algorithm, the advantage and disadvantage ofparticle swarm optimization algorithm are analyzed. Then put forward afuzzy association rules extraction mathod based on particle swarmoptimization algorithm. The method firstly transforms the data set intofuzzy set, and then the fuzzy association rules are extracted by particleswarm optimization algorithm from the fuzzy set.Improving the multi-mutation particle swarm optimization algorithm to aparallel searching method by using multi-swarm of multi-mutationparticle swarm, and appling to parallel extraction of fuzzy associationrules. This paper works on parallel computing environment, algorithmdesign and other aspects, preliminarily implements the parallel searchingassociation rules by using the swarms. The experiments prove that theimproved algorithm has advantages of parallel algorithm, and is betterthan serial extraction algorithm in large data sets.