Optimizing Seat Utilization in Bus Transport Through AI-Powered Segment Management

Public transport seat utilization have been and still are a critical challenge for sustainable urban mobility. In this study, the authors introduce SPAR (Segmented Probabilistic Adaptive Reallocation), a new AI-based dynamic seat allocation framework that can help manage bus transport systems by the application of intelligent segment management. Through the proposed approach, real-time passenger boarding and alighting trends are modeled with a graph-based representation of bus routes, such that nodes symbolize stops and seats and edges capture the time-dependent occupancy relationships. A hybrid algorithm consisting of Graph Neural Networks (GNN), Reinforcement Learning (RL), and Entropic Optimal Transport (EOT) is applied to the dynamic seat reallocation based on the changing demand with the aim of minimizing idle capacity and at the same time maintaining a balanced occupancy over the route segments. To check the performance of SPAR, a synthetic dataset has been created which represents 5,000 bus trips with dynamic boarding events and 40 seating capacities. The experiment outputs show that SPAR maintains a Seat Utilization Ratio (SUR) of 0.94, a Dynamic Allocation Accuracy (DAA) of 0.93, and a 47 % decrease in Segment Utilization Variance (SUV) compared to the traditional static and heuristic allocation methods. Moreover, the model gets converged rapidly within 30 epochs with a stability index of 0.93. These results provide strong evidence that SPAR not only improves operational efficiency but also enhances flexibility in public transport systems, hence, a scalable framework is created for real-time intelligent fleet management and data-driven seat allocation in smart urban transit networks.