PREDICTION OF FREIGHT TURNOVER OF RAILWAY TRANSPORT USING REGRESSION MODELS WITH DETERMINISTIC AND STOCHASTIC EXPLANATORY VARIABLES

Author(s):  M.P. Bazilevskiy, candidate of Sciences, associate Professor, Irkutsk State Transport University, Belgorod, Russia, mik2178@yandex.ru

Issue:  Volume 46, № 1

Rubric:  System analysis and processing of knowledge

Annotation:  The paper is devoted to the problem of prediction using regression models with deterministic and stochastic variables. The latter, if there is only one explanatory variable, are called Deming regressions. A method for estimating such models and a well-known, not based on probabilistic nature, interval prediction method based on them are briefly considered. For the first time, the latter was tested using the example of interval forecasting of rail freight turnover. The intervals obtained were as reliable as the confidence intervals for the classical regression. A new method for obtaining point forecasts for the Deming regression is proposed, which involves solving the problem of choosing such a ratio of error variances of variables that minimizes the mean absolute error for the examining sample. The proposed method is applied for point forecasting of freight turnover of railway transport. The predictions found turned out to be significantly better than the predictions obtained using classical regression with deterministic variables.

Keywords:  regression model, Deming regression, trend, forecasting, mean absolute error, rail freight turnover

Full text (PDF):  Download

Downloads count:  235