METHODICAL APPROACHES TO THE ASSESSMENT OF THE VOLUME OF SERVICES PROVIDED BY ENTERPRISES IN THE TRANSPORT INDUSTRY

Abstract

The paper investigates the task of increasing the reliability of estimating the volume of sales of services of enterprises in the transport industry. The article analyzes the dependence of the volume of sales on the following factors: the initial cost of fixed assets, the cost of new fixed assets, the average number of employees, and capital investments. Taking into account the presence of multicollinearity between the factors, it was decided to build single-factor regression linear and nonlinear models using the least squares method. The significance of the internal parameters of the models obtained by the least squares method was tested by the value of their p-values. After checking the significance of the internal parameters of the models, the regression models were compared by the following indicators: coefficient of determination, Fisher's criterion, and average approximation error. As a result of the comparison, among the single-factor regression models built using the least squares method, the best model was identified - a stepwise model by the capital investment factor. To build a multifactor model using the least squares method, the principal components method was applied to the values of the factors. The resulting model is slightly worse than the step model in terms of the average approximation error. In order to increase the accuracy of estimating the volume of sales of services of transport industry enterprises, machine learning methods were used to build multivariate models: k-nearest neighbors, binary decision tree. Before applying these methods, the primary data were standardized. When comparing the models obtained by the k-nearest neighbors method, it was found that these models are inferior to the one-factor stepwise model. The regression models based on the decision tree method proved to be better than all the models studied earlier. The best quality model for estimating the volume of sales of services of transport industry enterprises, in terms of the average approximation error, among the binary decision tree models was the model with a tree depth of 3. In this model, the average approximation error is 0.614 %.