Abstract <p>Permanent and temporary seismic networks designed for monitoring mine seismicity are aimed at assessing the risk of rock and rock–tectonic bursts. Operational processing of monitoring data, which includes both detection and location of seismic events, and determination of their main focal parameters (seismic energy and seismic moment), represent the most important task in assessing seismic hazard at a mine site. This article presents a neural network MLP machine learning model used to estimate the parameters of seismic events induced by massive explosions at the Korobkovskoe iron ore deposit of the Kursk Magnetic Anomaly. The temporary monitoring systems used at the mine are characterized by different sensor placement conditions in terms of microseismic noise levels. When training the model on data that takes into account different configurations of temporary monitoring networks, the model demonstrates high efficiency (<i>R</i>² = 0.81–0.93) and allows for reliable determination of the focal parameters and statistical characteristics of mine seismicity.</p>