Date of Publication : August 2026

Abstract: Motor Vehicle Transfer Fee (BBNKB) is one of the main sources of local revenue (PAD) in East Java Province. Since the enactment of Law No. 1 of 2022, BBNKB applies only to the first transfer of vehicle ownership, making revenue highly dependent on new vehicle sales. This condition creates uncertainty and fluctuations that require accurate forecasting to support fiscal planning. This study evaluates the application of a hybrid ARIMA–Artificial Neural Network (ANN) model for forecasting BBNKB revenue using weekly data from 2010–2024 obtained from the East Surabaya Vehicle Tax Office (Samsat). The research involves stationarity analysis using the Augmented Dickey–Fuller (ADF) test, ARIMA modeling based on the Akaike Information Criterion (AIC), and residual testing using the BDS test to detect nonlinear components. Results show that ARIMA residuals contain nonlinear patterns, so further an ANN model was employed to capture these components. The hybrid model combines ARIMA forecasts with ANN residual corrections. However, performance evaluation reveals slightly higher RMSE, MAE, and MAPE than the single ARIMA model, indicating limited accuracy improvement but providing insight into the linear and nonlinear fiscal data behavior.

Keyword : Fiscal Revenue Forecasting, Hybrid ARIMA–Artificial Neural Network (ANN), Motor Vehicle Transfer Fee (BBNKB), Nonlinear Time Series Analysis.

Reference :