Study of Integrating AdaBoost and Weight Support Vector Regression Model
Guosheng Hu, Feng-feng Zhu, Yingchun Zhang, Jin-Lian Yu · 2009
The performance and regression precision of weak learners (accuracies should be greater than 0.5) for pattern recognition and forecasting can be upgraded using AdaBoost algorithm. Support vector machine (SVM) is a state-of-the-art learning machines and have been widely used in pattern recognition area since 90's of 20th contrary, however the performance of SVM is not stable and easily influenced due to the choice of parameters and kernel function. Hence in this paper, an integrated AdaBoost algorithm and weight support vector regression (WSVR) model are proposed. The proposed regression model has higher forecasting accuracy, and eliminates uncertain. The influence effects of training samples are not same, the later the training sample, more important to SVM classification ability. So we endow a different weight to each training sample. The proposed WSVR in this paper is an ideal learning machine for utilizing weighted samples, it is different from FSVM. The experimental results illustrate the proposed algorithm have better WME and WPE than single SVM algorithm and BPNN network.