A Data-Driven Approach To Spatial-Temporal Prediction Of Floods And Landslides Using Machine Learning

Abstract

This paper proposes an approach grounded in machine learning techniques to predict landslides and food insecurity in susceptible areas. By integrating various datasets, including historical weather patterns, soil composition, land use, topographical data, and socio-economic indicators, the model aims to provide early warnings and actionable insights. The landslide prediction model employs techniques such as Gradient Boosting, Neural Networks and Random Forest to analyze the likelihood of landslide occurrences, while the food prediction model leverages time-series analysis and regression techniques to forecast food production and potential shortages. The models are validated using real-world data from high-risk regions, and the findings indicate a notable enhancement in predictive performance over conventional techniques. The method may contribute to assist policymakers and disaster management agencies in proactive planning, thereby reducing the impact of these events on affected communities.