Research And Application Of Clustering Ensemble Algorithms Based On Weight Designing

With the development of information technology, especially the development of the Internet and the explosive growth of the amount of information, the importance of information is in a rapidly growing. As an important tool for Data Mining, clustering technology is receiving more and more attention. With the extensive application of data mining techniques, clustering study also face more new content and challenges. In the field of cluster analysis, stability, accuracy and validity of clustering are some of extensively studied problems.Clustering ensemble algorithms is a new research direction of cluster analysis. In this thesis, the knowledge of clustering has been analyzed systematically, and the basic principles and characteristics of clustering ensemble have been fully studied. Most of the current algorithms do not consider the qualities of the members, but the result of clustering ensemble would be affected when some members are poor quality or there is noise interference. To solve this problem, this paper presents a clustering ensemble algorithms based on weight design. The main idea of the paper is designing the weights of members by clustering evaluation and different degree analysis of members, and then gets a better result. This algorithm has improved influence of members with better quality on the results, effectively reduced the noise interference, and improved the clustering accuracy. Finally, after the analysis of current intrusion detection technology, this paper applied the improved algorithm to intrusion detection system.