Abstract: In this paper, the model based on wavelet transform and support vector machine(SVM) was applied to forecast the annual maximum earthquake magnitude in China's mainland. Firstly, using wavelet transform the series of the annual maximum earthquake sequence in china's mainland was decomposed into several sub-sequences with different scales (frequency band). Secondly, the support vector machine was applied to forecast the decomposed sub-sequences, respectively. Finally, the reconstruction of forecasted sub-sequences was used as the final predicted result, to forecast the maximum earthquake magnitude in China's mainland next year. The comparison of SVM with ANN method was made. The forecasted results show that our method possesses higher prediction precision and excellent forecasting effect, and prove that the method to combine wavelet transform with SVM is feasible to forecast time series of strong earthquakes.

Key words: Support vector machines, Wavelet transform, Forecasting of strong earthquakes