Implementing FlowFlexDP for Advancing Passenger Demand Prediction using Cellular Footprints

Accurate prediction of citywide passenger demand is essential for optimizing driver scheduling in online car-hailing services. However, existing approaches often overlook deep dependencies or restrict the analysis to grid region partitioning. Inspired by recent advances in mobile traffic analysis, We introducing FlowFlexDP, an innovative demand forecasting model that merges regional crowd movement information and utilizes adaptable region segmentation techniques. FlowFlexDP offers a more comprehensive understanding of passenger demand dynamics. Analysis of a large-scale cellular data set from a major Chinese city shows that the passenger traffic data is highly correlated with how people move around. FlowFlexDP capitalizes on historical performance and crowd flow data using a Graph Convolutional Neural Network (GCNN) that allows for effective predictions over cities with different sizes and shapes. This new model represents a significant advancement in demand prediction, promising enhanced efficiency and passenger experience in urban transportation systems. Experimental evaluation demonstrates FlowFlexDP's superior performance in accurately capturing passenger demand patterns, highlighting its potential for optimizing driver scheduling and improving overall transportation efficiency.