The size of losses in the transmission network is one of the indicators of the economy of its operation. Each system operator pays special attention to forecasting and managing electricity losses. We need loss forecasts to ensure timely and economic coverage of these losses. Loss forecasts are a complex problem due to many different and challenging to predict parameters that affect the number of losses. The Transmission System Operator has to provide energy to cover losses, so there is a particular interest in the problem. Despite the growing interest, loss forecasting is becoming a more complex problem due to the increasing penetration of renewable energy sources, higher flows in the transmission network, and market deregulation, which introduces additional uncertainties and unpredictability. Due to many parameters and the complexity of the problem, the approach to loss forecasting often turns to machine learning techniques. Machine learning techniques give good results precisely in problems with a large number of parameters that need to be taken into account simultaneously. The latest steps in the development of machine learning algorithms show exceptional results in terms of performance and the range of issues where such algorithms are applied. A particular category of problems where such algorithms show their full potential is the problem of establishing a functional connection between many influential parameters and a set of output variables. The problem of planning technical losses falls into the category of such problems. Namely, previous analyzes of loss planning have shown that losses are affected by various parameters such as transmission line load, corona, creeping currents, meteorological factors, consumption, working day, network flows related to large cross-border energy transfers due to high hydrology and the like. When the parameters' historical data take into account, the set of input parameters becomes significant and challenging to solve with standard techniques. This problem belongs to the category of problems that can be solved by machine learning. The subject of this paper is to investigate the possibilities of short-term forecasting of electricity losses in the transmission network by machine learning methods for transmission system operators and for the needs of planning and managing the power system. I will analyze the processing of collected data suitable for machine learning models, and based on these data, the developed models for short-term loss forecasting will be described through a software tool for automated short-term loss forecasting. The second part of the paper will analyze the economic effects obtained by the new method of predicting short-term losses and their acquisition.