Abstract: Fuzzy ant-based clustering algorithm has been efficiently employed to serve real-world applications. Although, ant-based clustering algorithm can relieve the fast convergence during the search, limitation of such an algorithm to overcome the problems of local optimal traps along with divergence of the search are still non-trivial. Striking the balance between exploitation and exploration of the search is one of the significant keys to overcome such problems thus leads to achieve the global optimal solution. Nevertheless, arbitrarily defined parameters are usually used to control the cycle of exploitation and exploration mechanisms thus may lead to a biased and overly optimistic learning process. This study proposes an improved version of the fuzzy ant-based clustering. The objective is to apply a nonparametric method of balancing exploitation and exploration search during ant-based clustering, aiming to accomplish the global optimal solution. The criteria of performance evaluation rely on F-measures, FCM objective degree, Xie-Beni validity index and runtime as well. The experimental results, based on both real-world and artificial data sets indicate the high performance of the proposed method over the comparatively effective clustering algorithms.
Phichete Julrode and Siriporn Supratid, 2013. Improved Fuzzy Ant-Based Clustering: A Nonparametric Balance Between Exploitation and Exploration. Research Journal of Applied Sciences, 8: 425-434.