For providing the government with effective monitoring of the trends of the economic variables in the future and good reference for developing a reasonable policy, in this paper, we establish a time series model on China's Foreign Direct Investment (FDI) by using wavelet analysis and intervention analysis and time series analysis and predict the trend of FDI in the next several years. This model eliminates the interference of noise for predicting by using wavelet analysis, and describes the autocorrelation and time-varying volatility of the financial time series by using ARIMA- GARCH-M model. The simulation results show that this model explains the dynamic structure of China's FDI trends well.