AI-Powered Machine Learning Model for Geospatial Landslide Risk  Prediction using Satellite-Derived Environmental Data

G. Vamsi¹, K. Jhanshi²

¹Assistant Professor, ²MCA Final Semester, Master of Computer Applications,

Sanketika Vidya Parishad Engineering College, Visakhapatnam, Andhra Pradesh, India

ABSTRACT:Landslides are one of the most dangerous natural disasters that cause severe damage to infrastructure, environment, and human life. These disasters mainly occur in regions experiencing heavy rainfall, unstable slopes, and weak soil conditions. Accurate prediction of landslides is essential to reduce disaster risks and improve early warning systems. Traditional methods mainly depend on satellite imagery and deep learning models such as Convolutional Neural Networks (CNNs). Although these methods provide good accuracy, they require high computational resources and fail to capture temporal factors like rainfall patterns over time.To overcome these limitations, this project proposes an AI-based geospatial landslide risk prediction system that combines structured environmental data with machine learning and deep learning techniques. The system integrates historical landslide data with environmental parameters such as rainfall intensity, soil moisture, slope gradient, elevation, vegetation index, and land cover. XGBoost is used to identify important features, while GRU captures rainfall patterns over time. Additionally, Tiny Attention U-Net is used for image-based analysis when satellite data is available. The system provides an efficient, scalable, and accurate solution for landslide prediction and supports early warning mechanisms.Keywords: Geospatial Data, Disaster risk reduction, Landslide prediction, Machine learning, XGBoost, GRU, Early warning system, tiny attention U-net.