Analysis and Dynamic Prediction of Bus Dwell Time Under Rainfall Conditions

Exploring the degree to which bus stop times are affected by rainfall is necessary for a reasonable formulation of bus-scheduling management schemes under rainy conditions. Although numerous mathematical models have been proposed, the predictive accuracy of existing models is insufficient for the precise formulation of bus policies. This study considered linear bus stops in Shenyang as research targets, and based on field survey data, we analysed the bus dwell time and its influencing factors under varying degrees of rainfall. The Pearson correlation analysis method and SPSS software were used to reveal the degree of influence of parameters, such as the number of passengers boarding and alighting buses, rainfall level, number of berthing spaces, load rate and presence of signalised intersections, on the bus stop time under rainfall conditions. Support vector machine, k-nearest neighbour and backpropagation (BP) prediction models were established, and the BP neural network model, having the best prediction effect, was optimised using a genetic algorithm (GA). The constructed GA-BP prediction model was more realistic than the BP prediction model and can be used to predict bus dwell times under rainfall conditions. The study findings will facilitate bus punctuality and improve customer appeal for bus services.