Abstract—Wavelet neural networks (WNNs) are a variant of artificial neural networks (ANNs), which are powerful mathematical modeling techniques that are used to model and study a variety of complex real-life problems. During the unsupervised learning stage, the translation vectors of the hidden nodes need to be determined. The conventional k-means and fuzzy c-means clustering algorithms have been used for this purpose. Nevertheless, these methods are prone and sensitive to initial cluster centers that have been randomly chosen. In this paper, a new hybrid clustering algorithm is presented. The type-2 fuzzy c-means clustering algorithm is hybridized with the metaheuristic harmony search algorithm. The new algorithm is then used to determine the translation vectors of WNNs. By incorporating the evolutionary harmony search algorithm in the clustering algorithm, the hybrid algorithm is able to escape from local minima while searching for the global minimum. To validate the effectiveness of the proposed algorithm, a real world problem of epileptic seizure classification problem is studied. Simulation results showed that the hybridized algorithm outperformed the stand-alone clustering algorithms.

Index Terms—Wavelet neural networks, k-means clustering, fuzzy c-means clustering, harmony search algorithm, epileptic seizure classification.

Zarita Zainuddin and Kee Huong Lai are with the School of Mathematical Sciences, Universiti Sains Malaysia, 11800 USM, Penang, Malaysia (e-mail: zarita@cs.usm.my, laikeehuong1986@yahoo.com).
Pauline Ong is with Faculty of Mechanical and Manufacturing Engineering, Universiti Tun Hussein Onn Malaysia, 86400 Parit Raja, Batu Pahat, Johor, Malaysia (e-mail: ongp@uthm.edu.my).

[PDF]

Cite: Zarita Zainuddin, Kee Huong Lai, and Pauline Ong, "A Hybrid Algorithm for the Initialization of Wavelet Neural Networks: Application in Epileptic Seizure Classification," International Journal of Applied Physics and Mathematics vol. 3, no. 5, pp. 352-358, 2013.