Road Crash Injury Severity Prediction Using Ordered Logit and Gradient Boosting

Predicting how badly an occupant of a road crash is injured helps engineers judge vehicle-safety measures and helps trauma systems plan. This project predicts the police-reported injury severity of every car, truck and motorcycle occupant on the four ordered KABCO classes - O no apparent injury, C possible, B suspected minor, KA suspected serious or fatal - from 27 person, occupant-protection, driver-behaviour, vehicle, crash, road and environment factors. It uses the public NHTSA Crash Report Sampling System 2020-2024 (593,598 occupants), compares the classical ordered and multinomial logit models with Random Forest and LightGBM gradient boosting on a strict year split (test year 2024), reports survey-weighted odds ratios of the safety factors and explains the boosting model with SHAP. A web application assesses single occupants, scores whole months or uploaded files, raises triage alerts for likely serious injuries and runs on five databases. A GNU Octave cross-check and a complete MATLAB version are included.

Category: Automobile Engineering

Automobile Engineering Project AU02 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.