پيش‌بيني تقاضاي مسافر در راه‌آهن مبتني بر كلان داده

One of the important modes of transportation systems which plays an important role in sustainable development is rail transportation.This transportation mode boasts a special contribution due to its advantages such as being environmentally friendly and its ability to transport large numbers of people, which distinguishes it from other modes of transportation such as road, air, and sea. In most developed countries around the world, rail transportation is often considered an important and efficient mode of transport. One of the critical factors that can play a significant role in policy-making and long-term planning to achieve goals is the accurate prediction of transportation sector movements, such as the number of passenger trips or freight shipments. This forecasting information allows managers and transportation experts to make informed decisions in the areas of goalsetting, policy-making, and long-term planning for the transportation sector. Given the high costs associated with the construction and development of rail lines and fleets, the critical importance of accurately predicting demand in this system becomes even more apparent. This research first identifies the key indicators that can effectively predict rail demand and then uses ARIMA methods and neural networks to forecast the demand. Further, the Dickey–Fuller statistical test was utilized to determine the optimal prediction model and assess the correlation between data from each origindestination pair. Mean absolute value of error (MAE), Mean squared error (MSE), and Coefficient of determination (R2score) are measures used to assess the performance of the prediction model. As found in the literature review, the longterm demand for rail travel is influenced by several factors and indicators. The main indicators influencing rail demand in the long term include: the population density around stations, the quality of service delivery, the price of transportation services, the price of alternative modes of passenger transportation (such as buses), the distance between the origin and destination, the duration of the trip, accessibility to the station, the length of main railway lines, and the travel demand time. The research models were implemented based on actual data from the Iranian railway network. For this purpose, monthly passenger demand data for the main stations in 13 origin-destination pairs was collected from the year 1380 until 1400 in solar calendar (from the year 2001 till 2021 in Christian calendar). The results indicated that time is the most significant factor in predicting passenger demand. This index determines the strength of the relationship between each origin and destination. Based on the results of the research, the neural network model consistently performed better and had lower error rate in predicting passenger demand compared to the time series method in all origins and destinations eva‎luated in the study.

5