Abstract: In order to improve the ability to recognize and interpret the abnormal response boundary in the transient electromagnetic inversion imaging results of boreholes in coal mines, and to complete geophysical interpretation such as an accurate description of hidden water disasters in front of the heading face and a fine analysis of hidden disaster-causing factors, this paper proposes the application of apply clustering method of unsupervised machine learning to analyze the borehole transient electromagnetic processing results. According to the numerical distribution characteristics of resistivity and the characteristics of the two clustering methods, the K-Means clustering algorithm is selected to aggregate and classify the resistivity imaging results. In the process of clustering calculation, the maximum distance principle is used to determine the initial centroid of the cluster, the Euclidean distance is selected as the distance calculation method, and the elbow method based on the sum of square errors is used to determine the number of clusters. In view of the hidden water disaster in the heading face, three-dimensional numerical simulation and underground field application examples are used to verify the practicability and effectiveness of this method. The results show that the method can automatically identify the optimal number of clusters and realize accurate clustering of resistivity. The clustering imaging results can improve the smooth transition problem of the original imaging model, highlight the boundary of abnormal response, clearly display the shape and position of abnormal response, and help to identify and classify hidden disaster-causing water bodies in the inversion results of borehole transient electromagnetic advanced detection.