Research On Web Text Mining Based On Artificial Immune Algorithm

With the coming of information society in 1990s, the network has become increasingly popular which provides effective means and methods for accessing all kinds of information. So how to find the information from great deal of text data is a most urgent problem to be solved. Thus, research of text mining came into being. Text clustering is an important branch of the field of text mining, which has important and far-reaching significance.This article first solved the problem of K-Means algorithm which is sensitive to initial centers combining with improved artificial immune algorithm so as to make cluster centers more reasonable and obtain ideal clustering results. This is the first stage which aim to obtain optimized initial clustering centers. Utilizing aiNet Model which is proposed from De Castro in 2002 to cluster the Web texts based on the relatively optimized initial centers in the second stage.This article first elaborates the researching situation and existing problems of Web clustering text,and then introduces related process of the text clustering which is from pre-processing, representation of text model to the clustering algorithm in common use,meanwhile,evaluating and analyzing the advantages and disadvantages of these clustering algorithms.The paper analyzes the present problems of traditional artificial immune algorithm, putting forward the best strategies to improve the three genetic operators to accelerate the convergence speed of the algorithm and ensuring the optimality of generations.Eventually aiNet model which is utilized to cluster is improved to overcome the disadvantages of low accuracy for calculating affinity of high-dimention text clustering.The results of the experiment indicateds that the improved algorithm has relatively better dynamic adaption and ameliorate the quality of clustering.