Post-crash analysis and injury severity prediction in vehicle-pedestrian collisions using logistic regression and AI-based predictive analytics

Omefe, Samuel (2025) Post-crash analysis and injury severity prediction in vehicle-pedestrian collisions using logistic regression and AI-based predictive analytics. International Journal of Science and Research Archive, 16 (1). pp. 912-923. ISSN 2582-8185

Post-crash analysis and road safety management is critical for prediction of injury severity after a vehicle-pedestrian collision. Efficient prediction of injury severity after the occurrence of such events is necessary to guideline emergence medical care enhancement, transportation planning, as well as evidence-based safety courses of action. Traditional statistical methods, such as logistic regression, have been utilized vastly because of their transparent and interpretable results. However, recent developments in Artificial Intelligence (AI) have proposed machine learning (ML) and deep learning methods, which demonstrate better predictive capabilities, particularly on processing complicated, high-dimensional, and nonlinear crash data. This reviewed study aims to reveal how AI-based predictive analytics can improve injury severity prediction in vehicle-pedestrian crashes. It contrasts the advantages and weaknesses of both logistic regression and AI techniques in terms of their methodological aspects and examines the suitability of both approaches in different data conditions and realities of decision-making. An integrative review of recent empirical research and technical developments was performed, covering the application of ML algorithms such as decision trees, random forests, support vector machines, and gradient boosting and also deep neural networks (DNNs) and convolutional neural networks (CNNs). Model performance measures, explanation frameworks like SHAP and LIME, and real-world application scenarios in real time traffic systems were all assessed. An AI model will almost always be more accurate and flexible in its predictions than a logistic regression. The relative research proves that ensemble and deep learning models are especially useful in discovering nonlinear relationships and dealing with the class imbalance. However, logistic regression is still useful in its interpretability and less data demands. Logistic regression remains valuable due to its ease of understanding and interpretability, AI-based models offer better performance in terms of predictive performance of injury severity, their practical use requires trade-offs between accuracy and explainability and ethical concerns which create possibilities to make more nuanced and correct predictions. A hybrid modeling method based on logistic regression and explainable AI and real-time data integration is a promising direction to go in terms of both predictive accuracy and policy explainability and to improve pedestrian safety and post-crash interventions in the complex urban transportation setting.