Abstract: The water quality system is an open, complex, and nonlinear dynamic system with time-varying complexity. Although some achievements have been made in the research of water quality prediction methods, there are still some difficulties such as prediction accuracy and computational complexity. Therefore, this paper proposes a water quality prediction algorithm based on least squares support vector regression. Support vector machine (SVM) is a kind of commonly used machine learning classification model, nonlinear data are mapped from low-dimensional space to high-dimensional space through the kernel function,
linear classification and regression are realized in the high dimensional space, the least squares support vector regression (LS-SVR) uses all samples to participate in regression fitting, which makes the regression loss function be no longer only related to a small number of support vector samples, but all samples participate in learning to correct error and improve the prediction precision. At the same time, by this algorithm, the standard SVR solving problem is transformed from inequality constraint conditions and convex quadratic programming problem into solving linear equations, which increases operation speed and solves the water quality prediction problem with nonlinear complex characteristics.

Key words: Support Vector Regression (SVR), Least Square Support Vector Regression (LS-SVR), water quality prediction