This study presents a machine learning-based seismic hazard assessment for Himachal Pradesh, India, focusing on site-specific effects within the Western Himalaya. The region’s complex geological structure, influenced by Precambrian formations and the tectonic collision between the Indian and Asian landmasses, necessitates advanced methodologies for seismic analysis. A seismic study area with a radial distance of 350 km centered at latitude 32.10º and longitude 77.56º was selected, covering 1150 earthquake events from 1900 to 2023. Following a comprehensive data processing approach, including declustering and catalogue completeness analysis, a main catalogue of 734 earthquake events was developed, with magnitude distributions categorized into four ranges: 3.0–3.99 Mw(126 events), 4.0–4.99 Mw(497 events), 5.0–5.99 Mw(96 events), and above 6.0 Mw(15 events). Support Vector Machines (SVMs) were employed to model seismic hazard based on the earthquake magnitude distribution across these time periods. The SVM model, optimized with an RBF kernel, achieved a significant accuracy of 85%, with hyperparameter tuning revealing an optimal regularization parameter С = 10 and a kernel coefficient ϒ = 0.1. The confusion matrix analysis further validated the model’s effectiveness in accurately classifying seismic events. The integration of machine learning with site-specific seismic data provides valuable insights for earthquake risk mitigation, infrastructure development, and disaster management in the Himalayan region, enhancing the precision of seismic hazard assessments compared to traditional methods.