Reseach On Association Rule Mining Based On Serial Hybrid Ant Colony Optimization Algorithm

Due to the continuous advancement of digital technology,various industries have accumulated a large amount of data.Mining knowledge from the massive data set and using it to guide production practice has attracted wide attention.Association rule mining is one of the main technologies of data mining.In view of the low efficiency of common data mining algorithms,the ant colony algorithm proposed for discrete problems is considered as a new association rule mining algorithm.At the same time,with regard to the slow convergence of ant colony algorithm and easy to fall into local optimal problems,the mining performance of ant colony algorithm is improved by serial mixing with other algorithms,and applied to the mining of association rules.The main work is as follows:1.The thesis studies the algorithm serial mixing strategy,the cuckoo search algorithm,bat optimization algorithm,water wave algorithm and fruit fly optimization algorithm based on the convergence and search ability and the ability to jump out of local optimal performance strength,chooses the appropriate population strategies alternately,choosing the suitable mixture of order and ant colony algorithms fusion,constitute the serial hybrid algorithm based on ant colony algorithm.2.The optimization experiments of 0-1 knapsack problem based on ant colony algorithm,cuckoo search algorithm,bat optimization algorithm,water wave algorithm and fruit fly optimization algorithm in multiple dimensions are implemented.The differences of convergence,search ability and leaping out of local optimal ability of the five algorithms in solving discrete problems are analyzed through experiments.Respectively according to the serial mixed strategy will four algorithm mixed with ant colony algorithm to constitute new serial hybrid algorithm,the experiments on multi-dimensional knapsack data sets,and compared with the single algorithm of ant colony algorithm,comparing the experimental results verify the serial hybrid algorithm in multiple dimensions relative to the single ant colony algorithm in optimizing efficiency and quality are improved.3.An association rule mining method based on serial hybrid ant colony algorithm is designed.Different from traditional association rule mining methods,this method does not need to generate frequent item sets,but directly mines association rules.In view of this characteristic,binary encoding is used for the data set in this thesis,and Michigan encoding method is adopted for the particles to make the operation simpler and processing more convenient.This method is used to conduct association rule mining experiments on multi-dimensional data sets.The experiments prove that this method can effectively improve the efficiency and quality of mining on multi-dimensional problems.In general,this thesis studies the application of serial hybrid algorithm based on ant colony algorithm in association rule mining.Experiments on multiple dimensions prove that the improved algorithm is better than the single ant colony algorithm in both mining efficiency and quality,and could be applied to a variety of association rule mining problems.