PREDICTING THE ESTIMATED TIME OF CARGO DISPATCH FROM A MARSHALING YARD
| Autor: | Andrii Prokhorchenko, Ievgen Medvediev, Artem Panchenko, Dmytro Gurin, Sergii Panchenko, Oleksandr Dekarchuk |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2020 |
| Předmět: |
Physics
marshaling yard 020209 energy Applied Mathematics Mechanical Engineering railroad 0211 other engineering and technologies Energy Engineering and Power Technology 02 engineering and technology Industrial and Manufacturing Engineering Computer Science Applications machine learning Control and Systems Engineering Management of Technology and Innovation 021105 building & construction lcsh:Technology (General) 0202 electrical engineering electronic engineering information engineering lcsh:T1-995 lcsh:Industry lcsh:HD2321-4730.9 Electrical and Electronic Engineering expected departure time cargo dispatch Humanities |
| Zdroj: | Eastern-European Journal of Enterprise Technologies, Vol 4, Iss 3 (106), Pp 6-15 (2020) |
| Popis: | A method has been proposed to predict the expected departure time for a cargo dispatch at the marshaling yard in a railroad system without complying with a freight trains departure schedule. The impact of various factors on the time over which a wagon dispatch stays within a marshaling system has been studied using a correlation analysis. The macro parameters of a transportation process that affect most the time over which a wagon dispatch stays within a marshaling system have been determined. To improve the input data informativeness, it has been proposed to use a data partitioning method that makes it possible to properly take into consideration the impact of different factors on the duration of downtime of dispatches at a station. A method has been developed to forecast the expected cargo dispatch time at a marshaling yard, which is based on the random forest machine learning method; the prediction accuracy has been tested. A mathematical forecasting model is represented in the form of solving the problem of multiclassification employing the processing of data with a large number of attributes and classes. A classification method with a trainer has been used. The random forest optimization was performed by selecting hyperparameters for the mathematical prediction model based on a random search. The undertaken experimental study involved data on the operation of an out-of-class marshaling yard in the railroad network of Ukraine. The forecasting accuracy of classification for dispatching from the wagon flow "transit without processing" is 86% of the correct answers; for dispatching from the wagon flow "transit with processing" is 54%. The approach applied to predict the expected time of a cargo dispatch makes it possible to considerably improve the accuracy of obtained forecasts taking into consideration the actual operational situation at a marshaling yard. That would provide for a reasonable approach to the development of an automated system to predict the duration of operations involving cargo dispatches in a railroad system |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|