The ionospheric Vertical Total Electron Content (VTEC) with the characteristics of non-stability and seasonal variation, and the Empirical Orthogonal Function (EOF) also has advantages of simplifying and eliminating redundant information of unstable time series. In this work, the EOF and ARIMA models are combined, named as EOF- ARIMA model, to predict the Vertical TEC (VTEC) in China. The EOF decomposition was performed on the GIM data (the Global Ionospheric Maps data provided by the IGS Center) in the first 10 days of different seasons, and then the ARIMA model was used to predict the principal components; finally, the VTEC values could be obtained after reconstructed for the next 5 days. The prediction results of the EOF-ARIMA model were compared with the commonly used ARIMA model. The results show that the average relative accuracy and the average standard deviation are 83.3% and 3.51TECu for EOF-ARIMA model, respectively, which has a 3.3% and 0.16 TECu improvement for those of the ARIMA model, respectively. The predicted results of EOF-ARIMA model has no obvious seasonal variations, while the ARIMA model shows smaller values in autumn than other seasons. Besides, the predicted performance did not affect by the equator abnormal for EOF-ARIMA model. Thus, the EOF-ARIMA model shows higher accuracy and stability when used to predict VTEC at short-term for China area.