<p>Rockburst is a common geological hazard encountered in deep-buried tunnel and mining engineering, and its occurrence is closely related to rock mass fracturing activities. Microseismic monitoring technology captures signals generated by rock fracture, providing important data support for analyzing underground rock mass behavior. However, in practical engineering applications, microseismic signals are often contaminated by blasting, mechanical, and environmental noise, and manual identification suffers from low efficiency and poor stability. Therefore, accurately distinguishing effective microseismic signals from various types of noise is a key issue in microseismic data analysis. This study focuses on the task of microseismic signal classification and proposes a novel deep learning model, GhostRegNet-CBAM. The model integrates the Ghost module and the Convolutional Block Attention Module (CBAM) to enhance feature extraction capability and classification performance. A dataset is constructed based on measured data from the Baihetan Hydropower Station, and classification experiments are conducted on microseismic signals, blasting noise, mechanical noise, and environmental noise. Experimental results show that the proposed model outperforms the original RegNet model in terms of recognition accuracy and classification performance and demonstrates stronger robustness in complex noise environments. This study provides an effective method for the automatic identification of microseismic signals under complex geological conditions and can provide methodological support for subsequent research on the identification and classification of rockburst microseismic signals.</p>