The Research And Application Of Improved K-Means Algorithm In Data Mining

Data mining is interdisciplinary science of many fields, for analyzing mass data in order to provide people effective help.Cluster analysis is an important technology in data mining.Cluster algorithm can partition and classify data obiects inaccordancewith certain requirements and rules with-out priori knowledge.It makes a higher similarity among data in the same cluster and a lower similarity in different clusters,so cluster analysis has a vast range of prospects in appli- cation and research.This paper works over improving the k-means clustering algorithm. Based on the analysis of simulated annealing global optimizing technology and harmonic mean fuction,it studies simulated annealing k-means clustering algorithm and simulated annealing k- harmonic mean clustering algorithm. Simulated annealing technology and harmonic mean fuction are used to get a global optimal solution and decrease the dependence on the initial cluster centers respectively. In study of simulated annealing k-means clustering algorithm, it presents a DK-t0 select method to choose control parameter original value t0 . In study of simulated annealing k- harmonic mean clustering algorithm, it presents KH&K combination approach. Through the analysis of science department enter fraction line dataset of university entrance exam in 2009, it proves that DK-t0 selecting method is better than stochastic t0 selecting method in simulated annealing k-means clustering algorithm. According to clustering IRIS dataset, it compares performance of k-means algorithm, simulated annealing k-means clustering algorithm and simulated annealing k-harmonic mean clustering algorithm. Finally, simulated annealing k-harmonic mean clustering algorithm is used to analysis database of The School Paper System to realize application of clustering algorithm in data minning.