EFFICIENT ROAD ACCIDENT SEVERITY CLASSIFICATION USING RF-RFE AND DEEP LEARNING FRAMEWORK

Abstract

The purpose of this research is to enhance the precision and dependability of traffic accident severity predictions by introducing a robust framework for classifying road accident severity that incorporates Random Forest–Recursive Feature Elimination (RF-RFE) and a deep learning model. This work uses RF-RFE for optimal feature selection to reduce dimensionality and eliminate unnecessary variables from large accident datasets, with the goal of identifying the most important factors. The selected criteria are then fed into a deep learning system that can detect complex nonlinear relationships between variables including road conditions, weather, vehicle specs, and driver behaviors. Through the integration of deep neural network classification and machine learning-driven feature optimization, the suggested technique improves prediction accuracy, streamlines computations, and enables real-time decision-making. In terms of data categorization and resilience, the results show that the hybrid RF-RFE and deep learning methodology outperforms standard models. For this reason, it is a useful tool for traffic management authorities and lawmakers to use in their pursuit of safer roads and less severe accidents.