Research On Association Rules Based On Improved Multi - Population Genetic Algorithm

Along with the advancement of information technology,the data produced by all walks of life is sharply swelling. Using the method of the traditional database cannot meet the people found the thirst of knowledge. Hence, data mining technology arises at the historic moment,The emergence of data mining technology to help us solve the problem of some complicated issues. In data mining association rules is one of the areas where research deeper. The purpose of the association rules mining is to find the connection between the item and item in huge amounts of data. Now analysis of association rules is expanded to many aspects from the shopping basket.Genetic algorithm(GA) shows its unique advantages which is as the excellent global search algorithm in data mining. The main idea of this paper is in view of the traditional genetic algorithm easily fall into local optimum and precocious shortcoming, to propose some improvement multiple population genetic algorithm. The algorithm is based on the energy transfer in natural way, and generate the unequal scale of the three initial population. According to the three initial population’s different characteristic respectively take different evolutionary strategy, than achieve the global convergence rapidly.This thesis based on the theory and then finally applied to the actual,through the forest fire data which was provided by the UCI(University of California Irvine) to mining association rules. The experiment shows that this algorithm can find out the useful rules, does not affected the efficiency by premise,and according to the rules to analyze the main meteorological factors affecting the local fire.The work done by this thesis main includes:(1)Summarizes present research situation of data mining and association rules(2)Introduces the traditional genetic algorithm and put forward the improvement ways of thinking.(3)Aiming at the shortcomings of the traditional genetic algorithm, to propose genetic algorithm that based on the multiple population, and through comparing the traditional genetic algorithm to verify the feasibility and effectiveness of the algorithm.(4)The algorithm was applied to the actual, used to guide the occurrence of forest fire forecast.