Optimizing Discriminant Model for Improved Classification of Protein

Abstract

Classifiers based on discriminant model achieved the highest accuracy compared to other protein classification methods in remote homology detection, but all of the classifiers were troubled by imbalance training in modeling. This paper presented a protein classification based on optimization of discriminant model to further improve the classifier performance by setting different penalty coefficients for the positive and negative samples to balance the training set weights. Comparative experiments show that the method based on optimized discriminant model obtained higher accuracy, and the method can improve the performance of all classifiers based on discriminant model by optimization of the parameters.