Predictive Analysis of Long-Term Trends in Road Accidents and Casualties in India Using Machine Learning: A Focus on Total Fatalities, Killed, and Injured

Abstract

The increasing rates of road crashes and associated fatalities in India have led to immediate concern and action. Despite many efforts, the traditional approaches fail to capture the complicated patterns that exist in road safety dynamics. This paper suggests an innovative method involving machine learning based on historical data analysis and prognosticating models of long-term tendency for road accidents and casualties. Based on an extensive dataset, meticulous data pre-processing, and feature selection methods, this research utilizes feedforward neural networks (FNN) and Long Short-Term Memory (LSTM) networks. Furthermore, an ensemble model that captures the strengths of both approaches is formulated. Performance evaluation through Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and R-squared metrics indicate that the ensemble model has better performance with MSE = 0.00172280, RMSE = 0.0415067 and an impressive R2 value of 0.999426. Therefore, the findings of this study provide policy makers with useful insight and could be used for development of effective strategies aimed at enhancing road safety measures in India, having further potential for refinement based on future implementation and utilization of advanced machine learning algorithms tackling such important public health issue into the country’s transportation sector.