Coal industry is one of the largest parts of the world fuel and energy complex, and 36% of the world coal reserves are extracted by underground mining. The process of undermining leads to Earth surface deformations, and it can cause damage and destruction of buildings and structures. Therefore, it is necessary to monitor undermined objects with certain frequency and predict deformations of the territory. The article describes the solution of the prediction problem with the help of neural network technology. The creation of a neural network includes several steps, described in the article in detail: choice of network architecture, preparation and normalization of input data, development of a mathematical model for network calculation, training and testing of the network. Training and testing were done based on the materials of the gas pipeline undermining project. The developed tool allows predicting the Z coordinate for profile line benchmarks for any day of displacement based on the data of instrumental observations.
