Predicting 20 years of Agricultural Stability and Security in the Philippines: A Comparative Ma-chine Learning and Time Series Forecasting Ap-proach for Food Security Planning

Abstract

Food security remains a persistent national concern in the Philippines, driven by rapid population growth, climate variability, geopolitical disruptions, and fluctuating agricultural productivity. To support early interventions and strategic planning, it is essential to develop reliable forecasting tools capable of predicting food shortages and surpluses. This study presents a comprehensive time series forecasting model employing multi-decadal historical data (1961–2024) sourced from government agencies and international bodies, covering key indicators related to food production, supply, and consumption. The research focuses on the top 15 major agricultural crops in the Philippines, which collectively represent a significant portion of the national food system and economy. Multiple forecasting algorithms were implemented to identify optimal predictive models for each crop. These include Gradient Boosting Regressor, Decision Tree Regressor, K-Nearest Neighbors, Exponential Smoothing, Grand Means Forecaster, Naive Forecaster, Theta Forecaster, Auto ARIMA, and Prophet. Model performance was rigorously evaluated using statistical metrics such as RMSE, MAE, MAPE, and R² to determine accuracy, consistency, and analytical robustness. Results show that several models—particularly Gradient Boosting, Prophet, Theta, and Auto ARIMA, successfully captured seasonal patterns, long-term trends, and production shocks. Their forecasting outputs demonstrate strong predictive capability for future crop production and availability. The findings reinforce the importance of data-driven approaches in agricultural planning, especially for countries highly vulnerable to climate and market instability. The proposed forecasting framework serves as a decision-support tool for policymakers, local government units, agricultural agencies, and food industry stakeholders. By accurately predicting potential shortages and surpluses, the model can assist in resource allocation, importation strategies, crop diversification, farm support programs, and long-term food security planning. Ultimately, this research contributes to sustainable agricultural development, improved food resilience, and the broader national goal of ensuring stable and affordable food for all Filipinos.