Application Research Based On Ant Colony Algorithm In Mining Maximal Frequent Itemsets Problem

With the development of information technology, especially the emerging of the network technology, database computer technologies played an important role in every work of life, amassed data keeps growing .In the face of a mass of data which are bald and flavorless, it is not practical to interpret and digest the data artificially. So it is urgent to develop one technique for analyzing large volumes of data. In the back groud of this, data mining technology is emerging.As an important pattern in data mining, association rules mining which are used to discovering the relationships among the attributes to find out valuable dependencies among the fields, attracts many researchers. The efficiency of mining frequent itemsets is the key problem in association rules generating. Maximal frequent itemsets contains all of the frequent itemsets, and its amount is little then the latter. This paper makes a research on mining maximal frequent itemset because it could reduce the number of the candidate itemsets and cut down the cost.Ant colony algorithm is a bionic optimization algorithm, Because of its simple mechanism, plus-feedback, strong lustiness and excellent distributed computational mechanism, it has been successfully applied in many complicated optimization problem such as TSP, QAP, VRP and so on. When the number of items in database is large, there would generate combinatorial problem. Ant colony algorithm was proposed to improve this problem by using heuristic information and plus-feedback. This idea and method provide a new approach to solve mining maximal frequent itemsets problem.This paper analyses the maximal frequent itemsets problem firstly, and summarizes its mining algorithms and characteristics. It was abstracted as a subset problem. Through learning the solutions of TSP with ant colony algorithm and the characteristics, this paper solves the problems in ant colony algorithm: heuristic information selection, valid solution construction and pheromone updating. In the solution construction, there is an end condition estimate for each ant, and adopts back-off technique. In order to testify the efficiency and feasibility, this paper compared the simulation results of ant colony algorithm and Apriori algorithm on mushroom database. It is more efficient in the lower minimum support condition than Apriori algorithm. At last, this paper puts this algorithm in analyzing disaster weathers to discover the inherent relations, furthermore testified that it is feasible to use ant colony algorithm mining maximal frequent itemsets problem.