Generalized forecasting of cargo indicators for rail transportation

Abstract

To improve the efficiency of the rail transportation process it is important to forecast its indicators, including cargo loading. Given that the transportation of goods occurs under conditions of uncertainty, the task of forecasting is not trivial. This paper proposes a generalised forecasting method for the cargo indicators of rail transportation, which uses three models of local indicators of cargo with different weights and takes into account the scenarios of the transportation process. The following indicators were selected as local indicators: a) the value obtained by the three-factor model; b) the value obtained by the time-dependent factor model; c) selective expert judgement. The weight coefficients were determined by the analytic hierarchy process which uses expert judgement in the form of pairwise comparisons. This approach is a scientific novelty. Testing involved three scenarios of the transportation process on the Far Eastern Railway. Statistical data from 2000 to 2018 were used, and forecasting for 2019 was made. The first scenario only considered statistical data, the second scenario also used expert information based on the optimistic variant, and the third scenario used expert information based on the pessimistic variant. Knowledge of actual cargo loading allowed us to calculate the relative error values of the generalised forecasted cargo loading for each scenario. It can be noted that the method of generalised forecasting shows quite good results for all the three scenarios, although the forecasting models themselves give a significant error. The closest values were obtained with the third scenario.