Concentration Prediction of 4-CBA Based on Local Weighted LS-SVM

In this paper, a new fuzzy adaptive local modeling method based on local learning and weighted least squares support vector machine (LS-SVM) is proposed by building fuzzy membership model for the training data. Just as LSSVM, local LS-SVM is also sensitive to outliers or noises. A proper fuzzy membership model based on support vector data description (SVDD) is proposed to deal with the problem. Fuzzy membership value to each input sample is confirmed according to it´s distance to the center of smallest nclosing hyper sphere determined by SVDD. The proposed local weighted LS-SVM is applied to predict the concentration of 4-Carboxybenzaldchydc (4-CBA) in purified terephthalic acid (PTA) oxidation process. Results indicate that the proposed method actually reduces the effect of outliers and its accuracy is improved.