Road accident analysis and prediction

Abstract

Traffic collisions occur at an alarming rate in many parts of the world. More than 151,000 people were killed on India's roads in 2019, for example. Finding out what factors contribute to traffic collisions is crucial in lowering financial losses. We explore two models using the US accident cluster dataset: logistic regression and random forest. Random Forest achieves higher accuracy ratings and Auc curve results. We use them for forecasting and incorporate them into our web app to display the likelihood of each accident hotspot.