A Learning Method of Support Vector Machine Based on Particle Filters

Support vector machine (SVM) is a novel and popular technique for pattern classification and regression estimation. In the training process of SVM it is of great importance to determine a few tuning parameters to ensure the good performance of SVM. However, the widely used optimization methods such as Particle Swarm Optimization and Genetic Algorithm have the disadvantages of low convergent speed and limited overall searching ability. To solve this problem, this paper proposes an alternative approach whereby particle filters are used to estimate the key parameters in the training process of SVM. The SVM model built in this way is used to estimate process variables in the production of Bisphenol A. Simulations show the effectiveness of this method.