Abstract:
In order to obtain the best parameters of least square support vector machine(LS-SVM), a novel least square support vector machine algorithm integrating with cultural differential evolution (CDE-LSSVM) is proposed. In CDE-LSSVM, CDE algorithm is used to optimize the parameters of kernel width and the factor of punishment so as to obtain the model with better forecasting performance. Further, by considering that quantitative structureactivity relationships (QSAR) model is of high nonlinearity and has relativity between independent variables, CDE-LSSVM is applied to develop HIV-1 protease inhibitors QSAR model. In order to illustrate the performance of CDE-LSSVM model, LS-SVM, back-propagation neural networks and radial basis function neural network are employed respectively to develop the QSAR models. The simulation results show that CDE-LSSVM model is of better performance of fitting and forecasting.