Predicting Food Crisis Risk in East Nusa Tenggara Using XGBoost

Abstract

The food crisis is one of the strategic issues that directly impacts social, economic, and public health stability, especially in highly vulnerable regions such as East Nusa Tenggara (NTT). The arid geographical conditions, high poverty rates, and climate variability make this region vulnerable to disruptions in food availability and access. This study aims to develop a food crisis risk prediction model in NTT using the Extreme Gradient Boosting (XGBoost) algorithm, utilizing climate and socio-economic data from 2020 to 2024. The variables used include rainfall, air temperature, soil surface temperature, vegetation index (NDVI), soil moisture, number of rainy days, rice prices, and spatial information such as district/city and month of observation. The test results show that the XGBoost model is capable of producing 93.75% accuracy with the best performance in the Safe and Emergency classes. These findings indicate that XGBoost is effective as an early prediction tool for mapping food insecurity in NTT.