Integration Of Research And Application Of The Algorithm Based On The Improvement Of The K-means Clustering

As one of the most important technology of data mining's exploratory, cluster ensemble is becoming more and more hotter, which is used in all kinds of huge information fields such as telecommunication, banking and finance. With the rapid development of communication technology, major operators have entered a new era of 3G business operations. How to cater the customers' need and how to improve services of different customers are the key to success for different major operators. Using cluster analysis technology can master the information of all users and help the operators to enhance the quality of services.This paper studied a lot of academic literatures about cluster analysis and cluster ensemble domestic and foreign recently years. For the issue that a single cluster algorithm can only fit for special data sets, which can't be used widely, this paper proposed an improved cluster ensemble algorithm based in k-means.First, the algorithm defined a Difference Comparison Function, which was used to select cluster members that have smaller average of diversity measure. Second, this paper proposed a new Weighted Function to weight the clustering members. Finally, the Co-association matrix was used to ensemble. Experiments on several real data sets showed that the proposed method could effectively handle the different quality cluster members and get more accurate, more scalability and more robust ensemble results.The paper also used the improved algorithm in one customer communications operator behavior analysis.By studying the users' consumption behaviors; the using frequency of products, the paper could know the custom of users and get some useful information for marketing. The real dataset showed the effectiveness of the new algorithm.