Road traffic accidents present a premier global public safety and economic crisis, responsible for millions of annual fatalities and severe infrastructure disruptions [1]. Conventional safety interventions heavily rely on reactive, post-incident reporting rather than proactive, predictive systems. This paper introduces an integrated machine learning framework engineered to forecast road accident impact levels and delineate high-risk geographic clusters, colloquially termed "black spots." Leveraging a highly detailed, pre-processed manual record dataset from India (2017–2022) featuring 32 attributes across 12,316 distinct records alongside regional data arrays [6], we evaluate the predictive efficacy of four major supervised classification algorithms: Logistic Regression, Decision Trees, Naïve Bayes, and Random Forest Classifiers. Distinct from clinical injury models, the severity metric optimized wherein primarily measures localized operational degradation and impact on regional traffic flow stability [9]. Empirical results demonstrate that the Random Forest Classifier achieves superior performance, outperforming competitive architectures in terms of global classification accuracy ($92.0\%$), robustness to feature collinearity, and multi-class F1-scores [7]. Furthermore, a spatial clustering protocol is applied to isolate accident black spots [5]. The outputs of this framework provide a foundational system capable of translating multi-modal data streams into real-time hazard warnings for intelligent transportation networks and target policy interventions for urban planners.