Skip to Main Content
A predictive model of water-quality, which based on wavelet transform and support vector machine, is proposed. This model uses wavelet transform to get water time sequence variations in different scale, and optimizes three parameters of Regression Support Vector Machine with improved Particle Swarm Optimization algorithm, to improve the accuracy of prediction model. This model is used to take one-step and two-step prediction for the dissolved oxygen density, which got from Wang Jiang Jing auto-monitoring station. the maximum MAPE is 4.54% in 10 samples, and then we make a comparsion between results of this model and the BP neural network. Results show that this model is good performance, higher precision, simple operation, and has better quality prediction at prediction effect than the model based on BP neural network, it provides a valid way for water-quality.