Automatic elastic net clustering algorithm

• Published in: Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics pp. 2768 - 2773
• Main Authors: Chun-Wei Tsai, Tsung-Hsien Lin, Ming-Chao Chiang
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2014
• Subjects: Accuracy; Algorithm design and analysis; Approximation algorithms; Classification algorithms; cluster validity index; clustering algorithm; Clustering algorithms; Elastic net algorithm; Indexes; Traveling salesman problems
• ISSN: 1062-922X
• DOI: 10.1109/SMC.2014.6974347

Clustering has always been playing a vital role in many different disciplines because it is an important tool for analyzing a set of unknown input patterns. However, some important issues related to clustering, such as automatically determining the number of clusters and partitioning non-linearly separable data, are never fully solved even though many researchers work on this subject for a long time. As such, a novel method based on the so called elastic net clustering algorithm is presented in this paper to deal with exactly the two issues: partitioning non-linearly separable data and automatically determining the number of clusters. To evaluate the performance of the proposed algorithm, several well-known datasets are used. The experimental results show that not only can the proposed algorithm find the appropriate number of clusters, but it can also provide a higher accuracy rate than all the other methods compared in this study for most datasets.