The invention discloses a mine gas monitoring abnormal data identifying method. The influence of production factors on gas emission is considered. A real-time monitoring data sample is reconstructed. An initial centroid vector is set for clustering. Discriminant samples in clustering are analyzed. If a sudden situation occurs and is beyond the 95% confidence interval of historical data, a small probability event is determined. A monitoring anomaly is determined, otherwise detection data are normal. According to the method, abnormal data of gas monitoring in a coal mine can be effectively identified; from gas emission at different underground locations and flow and accumulation characteristics, a ventilation factor, the flowing law of a fluid in a ventilation network, a gas accumulation source in an upper corner and other factors are taken into account; based on historical monitoring data statistical analysis and through on-line analysis with a safety monitoring and control system, abnormal gas monitoring data are identified; and the problems of low computational accuracy and false monitoring alarm, which are caused by the fact that monitoring data processing is affected by a false signal in gas monitoring information, are solved.