Research And Application Of Intelligent Algorithm In Association Rules Mining

Association rule mining which mainly study the correlation of data attributes inactual application. The mined rules have very high value in business decisions,personalized recommendation, product design. The algorithms which are used inassociation rules mining are Apriori algorithm and Genetic algorithm of intelligentalgorithm recently. The classic Apriori algorithm has some disadvantages that it scansthe transaction database many times and generates many frequent itemsets.Thegenetic algorithm has some disadvantages that it easily plunged into localconvergence and its’ convergence speed is slow.This paper aims the problems which appear in the traditional field of associationrule mining, combining with the latest domestic and foreign paper. This topic mainresearch work includes the following contents:Firstly, the function and process, methods, application and the research hot spot,etc in data mining are introduced. Then steps of association rule mining areexplained.Through a real instance and the flowchart, the classic Apriori algorithm’soperation process and the deficiencies are introduced detailly.On this basis, someoptimization algorithm and its’ mind in the field of association rule mining aresummarized.Secondly, the basic mind of genetic algorithm is introduced.The basic processand essential elements are explained, its parameter setting method are illustrated. Thefeasibility and the specific steps which are based on the genetic algorithm forassociation rules mining are expounded detailly.Thirdly, the forefront theory of artificial intelligence-Imperialist competitivealgorithm (Imperialist competitive algorithm, the ICA) thought, application situationat home and abroad and operation process are studied. The papers at home and abroadand experiments show that the algorithm can quickly converge to the optimal solution,and it is not easily to fall into local optimum. Then a scheme is put forward for miningassociation rules based on imperialist competitive algorithm, The specific miningscheme is introduced are explained.On the basis of this,UCI public data sets are usedto verify ICA algorithm. Experimental results show that compared with GA algorithm,the ICA algorithm on the mining association rules could achieve global optimal easily,the minied rules’ precision is higher. Finally, based on a large number of domestic and foreign papers in the field ofassociation rule and immune genetic algorithm (IGA) theory are analysed and studied.There is a shortcoming that mining time is long of a latest intelligence algorithm onthe mining association rules mining are found by us.To solve these problems, athree-step coding based improved IGA was proposed.The3steps encoding was usedto encode continuous association rule mining in order to educes the segmentationpoint of mining result’s influence, the option based on vector distance was used toreduce the running time of the algorithm,the adaptive crossover and mutation factorwas used to to decrease the disturbance of artificial setting parameters on the miningresults.The experimental results show that, The experimental simulation results showthat the improved algorithm on continuous attributes association rule mining withoutreducing accuracy with the advantage of mining time is short.