Online least squares support vector machine regression based on rectangular window with forgetting factor algorithm

Considering of the problem that the online training for the standard least squares support vector machine (LS-SVM) is difficult, an new learning algorithm of online least squares support vector machine regression (OLS-SVMR) based rectangular window with forgetting factor (RWFF) algorithm is proposed, by combining the RWFF algorithm with support vector machine, the present and past window data are considered simultaneously. The proposed algorithm has less computation cost and high accuracy. The proposed method is proved, and then it is applied to forecast a chaotic time series. The effectiveness of the algorithm is demonstrated by the simulation results.