The binary forecasting of dynamic indicators based on machine learning methods

The problem of binary forecasting of dynamic indicators based on machine learning methods in relation to the problem of cargo transportation by railway transport is considered. The probabilistic neural network and logistic regression were chosen as the methods. The binary forecasting consists on evaluating predictive values of the indicator which is based on the belonging probabilities to one of two intervals. The forecasting is called binary or interval as on this process is calculated interval for the indicator value where it will be, not the predicted value of the indicator. The software is developed using the Python programming language with open source libraries. The software and algorithm test were done on the examples of real values of railway transportation process and shown its high accuracy of binary forecasting both on the probabilistic neural network and logistic regression methods.