Automatic subspace clustering of high dimensional data for data mining applications
SIGMOD '98 Proceedings of the 1998 ACM SIGMOD international conference on Management of data
An evolutionary technique based on K-means algorithm for optimal clustering in RN
Information Sciences—Applications: An International Journal
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
| Hi-index | 0.00 |
As data mining having become increasingly important, clustering algorithms with lots of applications have attracted a significant amount of research attention in recent decades. There are many different clustering techniques having been proposed. Some conventional partitioning-based clustering methods, such as K-means, may fail if a set of incorrect parameters is chosen, or breakdown when the objects consist of non-spherical patterns. Although density-based approaches, e.g. DBSCAN and IDBSCAN, could deliver better results, they may increase time cost when using large data bases. In this investigation, a new clustering algorithm termed DenVOICE is provided to circumvent the problems stated above. As a hybrid technique that combines density-partitioning clustering concept, the proposed algorithm is capable of resulting in precise pattern recognition while decreasing time cost. Experiments illustrate that the new algorithm can recognize arbitrary patterns, and efficiently eliminate the problem of long computational time when employing large data bases. It also indicates that the proposed approach produces much smaller errors than K-means, DBSCAN and IDBSCAN techniques in most the cases examined herein.