Impact Factors and Nonlinear Effects of Ride-hailing Charging Behavior Based on Order Data
This paper investigates the impact factors and nonlinear effects of charging behavior for electric ride-hailing vehicles in the built environment which is an important part of public charging station planning and operation.Based on the ride-hailing order data from Nanjing city,this paper develops an algorithm to identify charging behavior for electric ride-hailing vehicles and proposes ten built environment indicators from five dimensions:density,diversity,design,destination accessibility,and proximity to public transportation.The key impact factors on the charging behavior of electric ride-hailing vehicles are identified using the XGBoost model and Shapley additive explanation(SHAP)value algorithm,and the potential nonlinear relationships between these factors and charging behavior are further analyzed.Additionally,the model's fitting performance is compared with Random Forest(RF)and LightGBM to validate the effectiveness of the XGBoost model in regression fitting.The results show that the XGBoost model has better performance compared with traditional models,with smaller prediction error fluctuation and higher R2(0.446)than the traditional models.The number of restaurants,distance from the city center,and the number of leisure and entertainment facilities are found to have the most significant impacts on charging behavior.Moreover,all built environment factors show nonlinear effects on the charging behavior,with the distance to the city center showing apositive impact at the beginning and then becomes negative,while other variables exhibit a negative impact at the beginning and then becomes positive.
urban trafficnonlinear effectsmachine learningride-hailing charging behaviorpublic charging station planning
万方数据
维普
