Authors : Athraa Jasim Mohammed, Yuhanis Yusof and Husniza Husni
Abstract: Hierarchical text clustering plays a significant role in systematically browsing, summarizing and organizing documents into structure manner. However, the Bisect K-means which is a well-known hierarchical clustering algorithm is only able to generate local optimal solutions due to the employment of K-means as part of its process. In this study, we propose to replace the K-means with firefly algorithm, hence producing a Bisect FA for hierarchical clustering. At each level of the proposed Bisect FA, firefly algorithm works to produce the best clusters. For evaluation purposes, we performed experiments on 20 newsgroups dataset that is commonly used in text clustering studies. The results demonstrate that Bisect FA obtains more accurate and compact clustering than Bisect K-means, K-means and C-firefly algorithms. Such a result indicates that the proposed Bisect FA is a competitive algorithm for unsupervised learning.
Athraa Jasim Mohammed, Yuhanis Yusof and Husniza Husni, 2016. Integrated Bisect K-Means and Firefly Algorithm for Hierarchical Text Clustering. Journal of Engineering and Applied Sciences, 11: 522-527.