Data-Driven Strategies for Improving Railway Ticket Demand Forecasting Accuracy

The accurate prediction of railway ticket demand is vital for effective operational planning and resource management in the transportation sector. This study investigates various time series analysis techniques, including ARIMA, SARIMAX, and neural networks such as Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN), to forecast railway ticket demand. Utilizing an extensive dataset of ticket sales spanning several years, we trained and validated these models, evaluating their performance through key metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). Demand patterns were represented using Origin-Destination (OD) matrices, where the CNN model was employed to predict the entire OD matrix, while the other models focused on individual OD pairs. The findings reveal that the CNN model outperforms ARIMA, SARIMAX, and LSTM in terms of prediction accuracy, offering a more reliable approach for forecasting demand in railway networks. This study underscores the importance of data-driven strategies in enhancing the precision of demand forecasting, thereby contributing to more informed decision-making and optimized railway operations.