97 Applying ARIMA and Artificial Neural Networks Models to Predict Military Spending in China

This study took the initiative to forecast China's military spending based on a autoregressive integrated moving average (ARIMA) model and artificial neural networks (ANNs) models for predicting the short term (1 year), the medium term (3 years), the medium-long term (5 years), and the long term (10 years). The mean absolute percentage error (MAPE) index is applied to measure prediction accuracy. The results indicate that the single-variable ARIMA model shows more stability and higher accuracy across the four time periods, while ANNs models show only 1œgood1 accuracy. As to multiple-variable models, the prediction accuracy of different variables has advantages in the different time periods. Model 3 including variables of China's military spending, GDP, and inflation rate of the previous period shows the most accurate prediction to the next period of military spending both in short term and medium-long term, while Model 4 including China's military spending, GDP, inflation rate, and Taiwan's military spending of the previous period has the highest accuracy for medium term prediction. Meanwhile, Model 2 including China's military spending and GDP of the previous period indicates the most accuracy for the long term prediction. Overall, on the average of the four different time periods of ANNs models, Model 2 including China's military spending and GDP of the previous period proves most accurate prediction than that of others. This concludes the contributions of this study.