Combine Multi-predictor ofGas Concentration Prediction Based on Wavelet Transforms

Abstract

A method of combine multi-predictoris proposed based on wavelet transform to improve the prediction precision of coal mine gas concentration time series. Firstly, the proposed model employ Mallat algorithm to decompose and reconstruct the gas concentration time series to isolate the low-frequency and high-frequency information. Then, ARMA model is built forthe prediction of high-frequency information and rectifies deviations of the predicted values by Markov bias correction method while the SVM model is used to fitthe prediction of the low-frequency information. At last, these predicted values are superimposedto obtain the predicted values of the original sequence. This method makes an effective separation of the feature information of gas concentration time series and takes full advantage of the features of different prediction models to achieve complementary advantages. The comparison experiment with the single-predictor models(BP, SVM) and single-predictor models based on wavelet decomposition(W-BP, W-SVM)show that the proposed method improves the overall predictionprecision.The results show that method has high precision and strong practicability.