Millions of people die every year from road accidents, which are becoming more commonplace worldwide. They cost society heavily in terms of money and the economy. Road accident prediction has primarily been researched as a classification problem in the literature, meaning that the goal is to forecast whether or not a traffic mishap will occur in future deprived of delving into the intricate interactions between the many components that contribute to these incidents. Fewer studies have examined a subcategory of models of collaborative machine learning modals than majority of research conducted to date on the significance of subsidizing aspects for road accidents and respective severity. Thus, using the road accident dataset as a basis, we assessed a number of ML model in this study for forecasting the traffic accidents’ severity. Additionally, we have examined the anticipated outcomes and utilized an explainable machine learning (XML) approach to assess the significance of contributing factors to traffic accidents. This work has taken into consideration various ensembles of ML models, like Random Forest (RF), Extra Tree, Extreme Gradient Boosting (XGBoost), to forecast road accidents with varying injury severity. Extra tree is the top classifier, according to the comparison results, with 81.06% accuracy, 81.06% recall,75.91% precision, and 77.34% of F1-Score models.