Arabica Coffee Sales Forecasting Using ARIMA Neural Network (Case Study: KOTEM Bondowoso): Author's Country: Indonesia

Muhammad Ariful Furqon (Universitas Jember)
Saudi Efendi (Universitas Jember)
Yanuar Nurdiansyah (Universitas Jember)

Buana Information Technology and Computer Sciences (BIT and CS)
Vol. 7 No. 1 (2026): Buana Information Technology and Computer Sciences (BIT and CS)

Kopi Tembakau (KOTEM) Bondowoso is a local coffee producer specializing in Arabica ground coffee, sourced directly from nearby farmer cooperatives. A major hurdle they face is accurately forecasting sales, a critical factor for optimizing production and inventory. To tackle this, a hybrid forecasting model blending ARIMA (for linear/seasonal trends) and Neural Networks (for non-linear patterns) was developed. The study analyzed KOTEM’s sales data from September 2019 to August 2022, preprocessed to address non-stationarity via differencing and normalization. Results revealed the hybrid model outperformed standalone ARIMA, achieving a 1.0% MAPE (vs. ARIMA’s 1.3%). It also better captured sales volatility and seasonal shifts, offering more dependable forecasts. ARIMA-NN could significantly enhance KOTEM production scheduling and stock management.