Study on Chinese Text Clustering Algorithm Based on K-mean and Evaluation Method on Effect of Clustering for Software-intensive System

Software fault continues growing in importance to the factors of equipment failure. Analysis and utilization in amount of text data is important content for research of software failure and reliability of software-intensive system. Algorithm based on k-mean is proposed, in which the calculated cluster-head (CH) out of real text node sample set was replaced with one text node nearest to it in the real text set. Distance between two text nodes is computed by Euclidean distance. Then evaluation method of clustering effect based on Entropy is presented for text clustering algorithm. At last, experiments show that the improved algorithm can solve the problem of empty cluster. Empty cluster means that the calculated CH may be visual text node out of real text node sample set in the progress of k-mean algorithm. Furthermore, the evaluation method can obtain the best clustering result and optimal number of cluster-heads (CHs).