FORECASTING THE AMOUNT OF INCOME FROM THE SALE OF SERVICES OF TRANSPORT COMPANIES

Abstract

The article examines the problem of increasing the reliability of the value of net income from the sale of services of enterprises in the transport industry. The paper analyzes the nature of the existing dependence of the level of net income on the implementation of a group of factors. The factors considered are: operating expenses, operating income, other expenses and other income. It turned out that operating expenses and income have the greatest impact on the dependent variable. They were used as factors in the subsequent stages of model building. The proposed regression models are built using the least squares method. To build some of the models, the primary data were standardized and transformed using the principal components method. To build some models, we used the transformation of primary data using the decimal logarithm. The significance of the internal parameters of the models obtained by the least squares method was tested by the value of their p-values. As an alternative to the models built using the least squares method, models were built using machine learning methods. Multivariate regression models were calculated using the following methods: k-nearest neighbors and binary decision tree. In the process of modeling, the best models in each group were selected: a linear least-squares model based on the initial data, a linear least-squares model based on the principal components, and a two-factor model based on a binary decision tree. The qualitative characteristics of the obtained regression models (coefficient of determination, Fisher's criterion, average approximation error) were compared. As a result, the best model was obtained, based on the value of the average approximation error, i.e., a multivariate regression model built using the decision tree method. The multivariate regression model based on a binary decision tree has a very small approximation error and exceeds the previously obtained models in all the parameters studied. Based on the results of the study, the advantages of the obtained regression model based on machine learning (binary decision tree) for predicting the amount of net income from the sale of services of transport enterprises are determined.