شكوفه وفائي

Predicting delays is one of the most important and practical issues in the passenger transport industry, especially passenger rail transport. Taking into account the factors that affect the delays leads to a more accurate prediction of the delays. Factors affecting delays include operational factors such as train arrival and departure times at each station and non-operational factors such as weather conditions in which data are recorded and stored. In addition, online forecasting of delays than offline forecasting can, on the one hand, predict delays more accurately and, on the other hand, help passengers and train dealers make better and more accurate decisions. The present research has been conducted in order to provide online models for predicting delays using data related to operational and non-operational factors. In this dissertation, the Tehran-Mashhad train named Fadak 11 train with train number 348 from Raja company has been studied. This train route has 50 stations, of which 5 main stations are Tehran, Garmsar, Semnan, Shahroud and Mashhad, from which 50 stations were selected to predict delays. For this purpose, data related to meteorological characteristics of these 5 stations, infrastructure characteristics of the train, planned operations and real-time operations of the train in the 5 main stations mentioned above during three years, from 1397 to 1399 have been collected. For each of these 5 stations, two forecast times are considered before leaving each station and after leaving each station, which is a total of 8 forecast times. Then, modeling for online forecasting of entry and exit to each of these stations has been done based on the two forecasting times. In order to model the prediction, artificial neural network, random forest regression and gradient amplification regression, and random gradient amplification regression were used. After eva‎luating the methods for predicting the entry and exit in each station for two cases before leaving each station and after leaving each station, the method that gave the best result based on the experimental data was selected and then the delays in each station. Arrival and departure times are calculated based on the difference between the predicted and planned values. Implementation of entry and exit prediction models has been done using DM-CRISP data mining methodology. The results of the forthcoming research can be used to identify the stations or factors that led to the most delays, as well as for other trains and areas.

داده‌كاوي , حمل‌ونقل ريلي , پيش‌بيني آنلاين , عوامل عملياتي و غيرعملياتي , رگرسيون

data mining , rail transport , online predicting , operational and non-operational factors , regression

Shekoofe Vafaei

Masood Yaghini