New approach for forecasting rice and corn production in the Philippines through machine learning models

Abstract

This study presents a comprehensive analysis of machine learning models for forecasting rice and corn production in the Philippines, focusing on determining the most effective model for this purpose. Given the crucial role of these crops in the nation's economy and food security, accurate forecasting is essential. We compared four different models: Random Forest (RF), Echo State Network (ESN), Neural Network Autoregressive (NNAR), and Autoregressive Support Vector Machine (ARSVM), using historical production data from 1987 to the first quarter of 2023. The Random Forest model, configured with 500 trees and one variable at each split, emerged as the most accurate, demonstrating the lowest error rates across Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) metrics for both crops. This superior performance can be attributed to RF's ability to capture complex, non-linear relationships in the data. In contrast, ESN, NNAR, and ARSVM showed varying levels of accuracy, with ARSVM recording the highest error rates. This variation underlines the importance of model selection based on the specific characteristics of agricultural data. The findings of the study have significant implications for agricultural forecasting and planning in the Philippines. They highlight the potential of using advanced machine learning techniques to improve crop production predictions, thereby aiding in better resource allocation and policy formulation for sustainable agricultural development.