Research On Clustering Ensemble Algorithms And Their Applications

With development of information technology, it's critical to extract relationships and connotative information from a large amount of data. So Data Mining was proposed to resolve this problem and comprises statistics, database, artificial intelligence, machine learning and so on. Clustering analysis is an important study field in data mining. It has been applied in many applications of data classification and plays a key role in assessing relationships among patterns of data.In this thesis, Clustering Ensemble is studied systematically in order to improve stability, accuracy and validity of clustering, which are some of extensively studied problems in clustering analysis. Inspired by the work in sensor fusion and classifier combination, a clustering combination approach had been proposed to measure the similarity between patterns. First of all, we introduce algorithm EA (Data Clustering Using Evidence Accumulation). Then in order to overcome the lack of traditional cluster algorithms and EA, the algorithm FNCE (Clustering Ensemble based on the Fuzzy KNN Algorithm) was proposed. We combine the results of multiple fuzzy KNN partitions and make certain instability results with less impact on the entire results of clustering; thereby it avoids a local optimum and improves the accuracy of clustering. Additionally, without specified number of clusters in advance, it can be automatically determined in the process of clustering.Intrusion detection is an important component of computer network security, while clustering analysis is a common unsupervised anomaly detection method. So after discussing some related topics, an intrusion detection model FNIDM (Intrusion Detection System Based on the FNCE) was designed and was approved that higher Detection Rate and lower False Positive Rate are got in network attacks according to the experiments.