Abstract: The identification of microseismic events is the foundation of microseismic monitoring in coal mines. Most existing microseismic monitoring technologies are developed based on the variation law of single physical quantities, which can easily lead to misjudgment when processing coal mine microseismic data containing a large amount of noise and interference signals. In order to solve the above problem, the loss function of canonical variate analysis (CVA) is optimized and improved using ridge regression algorithm to achieve sparse modeling and enhance the model's generalization capability. The ridge regression improved CVA is used for fusion analysis of multi-channel coal mine microseismic monitoring data, and then real-time identification of complex microseismic monitoring data status is achieved. The simulated data and actual coal mine microseismic monitoring data are used for experimental verification of ridge regression improved CVA. In experiments based on simulated data, as the noise variance increases from 5×10⁻⁶ to 5×10⁻², the recognition accuracy of ridge regression improved CVA increases by 0.6%-5.4% compared to CVA, and the sum of false alarm rate and omission rate decreases by 4.8%-17.3% compared to CVA. In experiments based on actual microseismic monitoring data, ridge regression improved CVA can reflect the fluctuation of microseismic signals in the fusion analysis results of 20 channels of microseismic monitoring data. It verifies that this method has the capability to identify microseismic events in real-time. The average identification accuracy is 97.14%, which is 2.39% higher than CVA. The sum of false alarm rate and omission rate is 31.06%, which is 0.07% lower than CVA. The error rate is 2.86%, which is 2.4% lower than CVA.