Comparison of the Population and Economic Variables using the Different Forecasting Models

Abstract

In this paper, the aim is to examine population and economic variables using the Exponential Smoothing, ARIMA, FTS, NNAR, and BSTS models. We have to forecast for the next ten years using these models to determine which is the best model based on the MAE, RMSE, and MAPE values, whose values are lower than the best prediction model. For population variables, the ARIMA model is best fitted as compared to other models, and the Total and Urban populations have well fitted in the ARIMA model as compared to other models and the Rural population has well fitted in the ARIMA and FTS models as compared to other models. For economic variables, the ARIMA model has the best performance as compared to other models in the Age dependency ratio variable, and the NNAR model has good performance as compared to other models in the GDP variable. Based on RMSE, MAE, and MAPE values, the ARIMA model performs better than other models like Holt's Exponential Smoothing, ANN, Fuzzy Time Series, and Bayesian Time Series Models. Except for the ARIMA model, the NNAR model has better fits than other models.