PSSP with dynamic weighted kernel fusion based on SVM-PHGS

摘要: Since 1960s, researchers have proposed several prediction methods, for protein secondary structure (PSSP), whereas the accuracy of them is no more than 80%. In this case, there an urgent need to introduce a high method. One learning method called support vector machines (SVMs) has shown comparable or better results neural networks on bioinformatics applications. This research proposes based SVM which been improved by new parallel multi class (PMC) method, hierarchical grid search (PHGS), cross validation (CV) technique and weighted kernel fusion (WKF) The presented PHGS applied regularize parameters SVM's function important impact accuracy. Using suitable input data particular problem can improve remarkably. Also our Position Scoring Matrix (PSSM) profiles are used as information it. goals study calibrate different functions' result in order determine classes accurately. right choice issue creating supreme performance so we propose dynamic weight allocation non-linear analysis system. obtained classification accuracies 84.65% 83.94% RS126 CB513 datasets respectively they very promising with regard other methods literature problem. evaluating behavior comparison state arts independent dataset used. show that comprehensibility WKF SVM-PHGS much methods.