Research on a new clustering algorithm in data mining

Abstract

Data mining is one of the leading fields in the combination area of database and decision supporting, and clustering is a significant task for data mining, in which clustering algorithm is the core technology. The new clustering method based on genetic algorithm and gradient descent method (G-G clustering algorithm) is proposed in this paper. Genetic algorithm has the advantages of global searching and strong robustness, and will not getting stuck at local optimal values. Unfortunately, it can only reach the near-optimal value after many generations of selection, crossover and mutation. Therefore, gradient descent method is utilized at the end of genetic algorithm based clustering method to get global optimal values. Clustering results of two groups of experimental data show that the new clustering method is one with global optimal, and the results is evidently better than k-means clustering method.