COMPARING NEURAL NETWORK AUTOREGRESSIVE METHOD FOR IMPORT DUTY REVENUE FORECASTING

Abstract

Neural Network is one of the interesting methods in data analytics used to forecast time-series data in the previous years. In one of the initiative strategies, the Directorate General of Customs and Excise (DGCE) compares this method with Holt-Winters exponential smoothing to get more accurate forecasting of import duty revenue. This paper compares Holt-Winters and Neural Network(NN) to get better forecasting for import duty revenue using data from Customs and Excise Information System Automation (CEISA) billing system. As a result, NN gives a better result with a lower Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Square Error (RMSE). Therefore, neural networks should be used to forecast and monitor the realization of import duty revenue so DGCE can identify the change in economic indicators that can affect import duty revenue earlier and produce the right policy to respond to it.