Most of the classification problems frequently encounter a multi class predicament and offers a good scope for research. This paper has a comprehensive approach to the available multi-class technique using Artificial Neural Networks and then introduces a new algorithm to overcome the demerits of the former. In addition, a new algorithm combining ANN and chameleon clustering is suggested and validated. An SVM model for the above is also proposed and sufficiently tested with a typical example i.e. Image Segmentation. Also, the permutation effects prevailing in Half -against-Half multi class algorithm of SVM is efficiently tackled by developing an algorithm using "circular shift strategy" and employing the same. The use of clustering methods with SVM to improve its efficiency is also discussed. All the above mentioned models are extensively analyzed and the results are presented. It is found that the proposed method is an effective alternative for existing methods and offers consistent performance. Â© 2009 IEEE.
Dr. Ramanathan R., Rohini, P. A., Dharshana, G., and Soman, K. P., “Investigation and development of methods to solve multi-class classification problems”, in ARTCom 2009 - International Conference on Advances in Recent Technologies in Communication and Computing, Kottayam, Kerala, 2009.