Analisis Prediktif Ketahanan Pangan Berbasis Data Spasial dengan Metode Random Forest Dan Cellular Automata di Provinsi Nusa Tenggara Timur

Abstract

Food security remains a key concern in sustainable development, especially in regions like East Nusa Tenggara (NTT) that are prone to drought and land conversion. This study aims to explore future food security in NTT by applying spatial data and predictive models to forecast conditions in 2030. Two main approaches were used: the Cellular Automata–Artificial Neural Network (CA–ANN) model to simulate land cover changes, and the Random Forest Regressor to predict rice productivity using environmental variables such as NDVI, land surface temperature, rainfall, elevation, and slope. The CA–ANN model showed strong spatial accuracy at 87.6%, with results indicating a decrease in cropland in several areas. The Random Forest model performed well with an R² of 0.90 and RMSE of 1.74, highlighting elevation and temperature as key drivers of productivity. By 2030, projections suggest a rice deficit of 221,000 tons, equivalent to more than 790 billion kilocalories. These findings underscore the urgency for local governments to adopt data-driven approaches when planning for sustainable food security in the years ahead.