Application of Classifier Fusion for Protein Fold Recognition

Protein data patterns which are discriminative can be used in many beneficial applications if they are defined correctly such as molecular medicine, agriculture, and microbial genome applications. Prediction of protein folding patterns by which the function of a protein whose structure is unknown can be determined, is much more complicated than that of protein structural classes. The classification rates achieved using different methods to solve this problem are not satisfactory and there is an urgent need to improve this classification rate. In this paper, a set of basic classifiers is used where each one is trained in different parameter systems all extracted from a common training dataset. Each individual classifier uses Probabilistic Neural Networks for classification in which the radial basis function parameter is optimized by Particle Swarm Optimization algorithm. Their outcomes are combined thru a weighted voting and Ordered Weighted Averaging (OWA) for final determination of classifying a query protein. The recognition rate achieved is 5-8% higher than the corresponding rates obtained by various existing Neural Networks.