The binary multi-SVM voting system for protein subcellular localization prediction

Abstract

Support Vector Machine (SVM) as a learning system has been widely employed for pattern recognition and data classification tasks such as biological data classification. Choosing appropriate parameters are essential for SVM to achieve a high global performance. In this paper, we propose a new binary multi-SVM voting system without difficult parameter selection for protein subcellular localization prediction. The sufficient experimental results demonstrate that the multi-SVM voting system can achieve higher average prediction accuracies for the protein subcellular localization prediction than the traditional single-SVM system. © Springer-Verlag Berlin Heidelberg 2005.