Analysis of Food Security Index Predictions in Indonesia Using Machine Learning Approach

Abstract

Food is one of the basic human needs that should always be available. To fulfill the role of in a region, the concept of food security is established to measure sufficiency, availability and quality of food. Food security for a country is expressed using Food Security Index (FSI). FSI score for a country reflects its ability for survival. It is therefore very important to measure the score and be able to predict future scores to enable control and improvement. To realize the improvement of Indonesia's food security, a model is needed to predict the Food Security Index in Indonesia. This paper explores the model using data from the Indonesian Food Security and Vulnerability Atlas (FSVA) at the Regency and City levels in 2018-2024 period with a total of 3,598 records. We evaluated Multiple Linear Regression, Least Absolute Shrinkage and Selection Operator, Random Forest, eXtreme Gradient Boosting, Support Vector Regression, and Ensemble Machine Learning models for predicting the FSI score. The models are evaluated using r-squared (R2), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). The results shows that the XGBoost method is the best method for predicting the Food Security Index in Indonesia with an R2 value of R² of 0.978, RMSE of 0.024, and MAE of 0.016. In addition, the XGBoost method predicts the average national Food Security Index score in 2025 and 2026 to be 75.14 respectively.