Application of machine learning model optimized by improved sparrow search algorithm in water quality index time series prediction

Water quality index is an important indicator to evaluate the water quality of rivers. Machine learning models have been widely used in the task of water quality index prediction, but the problem of model parameter optimization still has not been effectively solved, which seriously affects the prediction accuracy and the applicability of the model. In recent years, a variety of intelligent optimization algorithms have been applied to solve model parameter optimization problems. For example, Sparrow Search Algorithm (SSA), Gray Wolf Optimization (GWO), Genetic Algorithm (GA), etc. However, the existing optimization algorithm has limited optimization capability and still needs further improvement so as to enhance the optimization capability. In this paper, we improve the standard sparrow search algorithm. The improved sparrow search algorithm (IMSSA) uses Tent chaotic sequences to initialize the sparrow population, thus increasing the population diversity; we add adaptive inertia weights and random inertia weights to the SSA, while incorporating the simulated annealing algorithm for optimization, which improves the search performance, convergence accuracy and the ability to jump out of local optimal solutions. We compare the IMSSA with other advanced optimization algorithms on several common test functions, and the results show that the algorithm outperforms other algorithms in terms of convergence accuracy and merit-seeking ability. Meanwhile, we used the IMSSA to optimize the parameters of support vector regression (SVR) and random forest regression (RFR) models to obtain two water quality index prediction models, IMSSA-SVR and IMSSA-RFR, and applied the models to river dissolved oxygen and permanganate index prediction. The experiments show that our model effectively improves the prediction accuracy of river water quality index and has strong practicality.