The CUGTEM-GK II transient electromagnetic instrument is used for real-time geophysical exploration of mine roadways,generating 3D model imaging results,and providing forecasting and warning data for water inflow in underground rock layers,displacement of roadway roof and walls,etc.Machine learning algorithms integrated deep convolutional neural networks,multilayer neural networks,data binary output,fuzzy decision matrix warning,etc.,to provide rough warning results with wide coverage and precise warning results with narrower coverage.Through simulation analysis,both rough warning results and precise warning results show certain data sensitivity,with the precise warning results able to provide high sensitivity warning results 24 h before encountering problematic rock layers during roadway excavation.