Prediction of SO_2 Concentration Based on Fuzzy Time Series and Support Vector Machine

摘要

The existing prediction methods for SO_2 concentration mainly have the disadvantages such as no unified sources and influences, sensitive to small sample, easy to fall into local optimum etc. In order to solve these problems, a method for SO_2 concentration prediction is proposed, based on fuzzy time series and support vector machine (SVM). The method takes the seasonal variation of SO_2 concentration as the basis, takes the four seasons as the time series, and takes the 24 hours as the graining window width, then extract the characteristic values of the original sample data through the Gaussian kernel function for SVM training, and optimize model parameters by k-fold cross validation method combined with the grid division. Finally, a SO_2 concentrations prediction model is established by using one-hour average SO_2 concentrations as sample data, and the calculation process is realized by using LIBSVM tool. The results show that the prediction method of SO_2 concentration based on fuzzy time series and SVM is not restricted by the machine rational theory, can solve small-sample learning problems, and has good nonlinear fitting effect.