Integrating Topographic, Hydrological, and Geological Parameters for Landslide Prediction: A Machine Learning Approach in Himachal Pradesh

Landslides occur frequently in the Himalayan region and are highly destructive, especially in Himachal Pradesh, results in loss of human life and infrastructure. Modern tools are needed for assessing this hazard. The state of Himachal Pradesh is more prone to landslides due to steep slopes, fragile geology, and high rainfall intensity. This paper presents a ML framework for landslide susceptibility mapping based on optimised feature selection techniques. Feature reduction was applied to remove redundancy and emphasise the most influential conditioning factors: elevation, slope, rainfall intensity, and land cover. Unlike traditional approaches that produce binary predictions, the proposed framework allows for probabilistic forecasts. Based on certain thresholds in probability, the following categories of risks were defined: high-risk (>85%), moderate–high risk (75%–85%), moderate-risk (50%–75%), and low-risk (<50%). Further evaluation of the model's reliability was done by applying standard metrics: accuracy, precision, recall, F1-score, and AUC. Obtained susceptibility maps define the study area into distinguishable zones of hazards, which may be very useful in delineating areas of priority for disaster preparedness and mitigation efforts. All the above-mentioned findings prove that data-driven probabilistic modelling improves the assessment of landslide risk and provides a scientific backdrop to help policymakers and planners mitigate the vulnerability of Himachal Pradesh.