Analysis of the SUV Involved Pedestrian Crashes in Pennsylvania

This thesis addresses the pivotal challenge of predicting the SUV involved pedestrian crash severity and proposes improvements to existing methodologies, underscoring the substantial threat posed by such incidents. Utilizing a comprehensive dataset spanning five years from the state of Pennsylvania, USA, the study acknowledges and addresses the challenge of class imbalance through the application of the Synthetic Minority Oversampling Technique (SMOTE) for data augmentation. Methodologically, diverse artificial neural network (ANN) architectures are explored, with meticulous evaluation through K-fold cross-validation to ensure the robustness of the model. Descriptive statistics and correlation analyses are employed to investigate crash characteristics and inter-variable relationships. The outcomes underscore the efficacy of SMOTE in improving predictive accuracy. Beyond its primary predictive contributions, this research offers nuanced insights into factors impacting model efficacy. By addressing prevailing limitations and introducing an innovative approach to handling class imbalances, our research informs the development of interventions to enhance road safety. The findings carry crucial implications for policy and practice, with the ultimate goal of reducing pedestrian accidents and mitigating their severity.