Application of the Algorithm of Subspace Ensemble Regression to Multivariate Calibration

Aiming to raising of the predictive ability and lowering of the computational complexity of the multivariate calibration model,the algorithm of subspace ensemble regression with multivariate linear regression,i.e.,the algorithm of SER-MLR,was proposed and applied to multivariate calibration.Good comprehensive performance of the proposed method was verified by its application to two substantial examples of NIRS analysis and by comparing with the algorithm of full-spectrum partial least square.The algorithm of SER-MLR was characterized not only by its easy in explanation,but also by the establishment of a simple,accurate and robust calibration model at low computational cost.Besides,it was also less sensitive to overfitting.