Perbandingan Metode Seasonal ARIMA dan Extreme Learning Machine dalam Prediksi Produksi Padi di Sulawesi Selatan

Abstract

South Sulawesi is one of the provinces that significantly contributes to national rice production. Therefore, accurate forecasting of rice production is crucial for food security planning and agricultural policy-making. This study aims to compare the performance of the Seasonal Autoregressive Integrated Moving Average (SARIMA) and Extreme Learning Machine (ELM) methods in predicting rice production in South Sulawesi. SARIMA is a statistical forecasting method effective for data with seasonal patterns, while ELM is a machine learning approach capable of handling complex relationships among variables with high computational speed. Rice production data from the Central Statistics Agency (Badan Pusat Statistik) were used to evaluate the accuracy of both methods. The evaluation was conducted using forecasting error metrics such as Mean Absolute Percentage Error (MAPE). The results show that the SARIMA(1,1,0)(1,1,0)12 model outperformed ELM in predicting rice production in South Sulawesi. This is indicated by a lower MAPE value of 19.937%, compared to 21.632% for the ELM method.

Keywords

• Rice Production
• SARIMA
• Extreme Learning Machine
• Forecasting
• MAPE