Decremental learning based on sample-weighted Support Vector Regression

In this paper, a new modeling method-decremental learning based on sample-weighted SVR(DSWSVR) is proposed, which introduces the decremental learning strategy into sample selection based on support vector regression (SVR). DSWSVR differs from SVR in that it builds a new sample set, where some sample in the original sample set are weighted differently to account for its representative to improve the prediction ability of the algorithm. Simulation results show that the proposed algorithm can improve the performance of the SVR modeling.