The data of pest occurrence are complicated and unpredictable time series. The linear or nonlinear features of pest time series can not be captured based on single prediction model. A new hybrid forecasting model based on autoregressive integrating moving average (ARIMA) and support vector machine (SVM) is proposed in this paper. ARIMA model was used to predict the linear component while SVM model was used for the nonlinear residual component of pest time series, and then the hybrid forecasting results were obtained. The prediction performances of the method were tested on Dendrolimus punctatus occurrence area. The results show that the hybrid model, which combines the respective advantages of both linear and nonlinear models, has better accuracy than any single model. Hybrid model is a good and effective method for pest forecasting.