Applying Tree-Based Ensembles and Shapley Additive Explanations to Study the Factors Impacting Pedestrian Walking Speed Over Elevated Walkways

Abstract

The present study attempts to compare seven tree-based models (including ensembles) to suggest the best modeling approach for identifying the dominating factors (microscopic, macroscopic, and geometric) and predicting the walking speeds over elevated walkways. The results show that Gradient Boosted Trees and Light Gradient Boosted Trees outperformed other approaches in predicting walking speed. The variable importance of final models estimated using SHapley Additive exPlanations (SHAP) reveals that walking speed was dependent on the flow, density, and length of the facility. Moreover, other features such as gender, age, height of facility, and width of the facility also play a significant role in determining the pedestrian walking speeds. The identification of important variables not only provides better insight into factors that affect walking speed over elevated facilities but also provides a valuable source of information to researchers, planners, and policymakers for better designing, operation, and management of elevated pedestrian infrastructures.