Hybrid Clustering Algorithm Based On Hierarchy

With the development of science and technology, everyone has been drowned by a largeamount of data, and has no time to look at the data which has made people not to find thedesired information quickly.The face of huge amounts of data, we must find effective methods,which can classify and analysis data, gather data and flag exceptions data. Data Mining is putforward when we solve such problem of technology. Clustering analysis is a major technologymeans of the field of data mining, which divides the similar target into clusters, and helpspeople search and find useful information.At present, researchers have put forward lots of clustering algorithms, hierarchicalclustering algorithms is one of important means which are applied widely, and has been paidmuch attention.First, this article introduces data mining technology briefly. The clustering technologies areconducted in-depth research and analysis. Clustering algorithms should have characteristics.And principle and key techniques of these algorithms are introduced systematically, then tocompare the advantages and disadvantages of the various algorithms.Then, we propose one clustering algorithm which based modularity and backtracking BM-Chameleon because the Chameleon algorithm requires human to give the clustering relatedparameters as well as irreversibility of the merged cluster operation. The algorithm willautomatically find the clustering parameters of the most suitable for the data set, while achievesbacktracking operation in order to get the best clustering effect. To test the two algorithms withanalog data, the results show that the BM-Chameleon algorithm can improve the accuracy ofclustering results.Last, in order to solve the problem of time complexity when the introduction of themodularity and backtracking mechanism KBCM algorithm, we proposed a hybrid clusteringalgorithm which combines BM-Chameleon algorithm with k-means algorithm which is a kindof traditional clustering algorithm based on partition. Under the premise of ensuring the qualityof clustering it can greatly improve the time complexity of the algorithm. The experimentalresults show that the hybrid algorithm has high classification accuracy and time complexity.