Research on water quality predictive model based on robust splines partial least squares

Considering the correlation, nonlinearity and outliers of water quality data, a Robust Splines Partial Least Squares (RSPLS) algorithm is proposed and applied to study the predictive model of water quality. The nonlinear relationship of the measured data was transformed into quasi-linear by splines. Then robust method, which distributed different weights to the measured data, was implemented into the iterative operation of the transformed data to establish a robust nonlinear data predictive model of water quality. The method can overcome the multi-collinear of the variables and the outliers. The simulation of the algorithm shows that the water quality predictive model can predict the water quality accurately.