With the aim of optimizing train seat allocation strategies, this paper applies a series of machine learning techniques to re-identify the real willingness to purchase tickets of different seat classes from high-speed rail ticketing data. Based on the re-identified demand, this paper studies optimization approaches from two perspectives. First, considering the practical operation conditions of railway companies, a rule-based distance priority principle is evaluated as a benchmark. Then, an integrated optimization approach considering operational revenue and train capacity utilization is developed, an integer linear programming problem with the objective of maximizing the total train revenue is formulated and solved. Results show that the proposed optimization approach outperforms the state-of-the practice by around 10%.

CICTP