Training teachers to manage their instruction and the class is known to be important and challenging. In traditional teacher training programmes, micro-teaching is frequently adopted; however, its effectiveness could be limited if the learners do not have good opportunities to immerse themselves in and interact in the learning contexts. Virtual Reality (VR) technology could be a good approach to coping with this problem. Nevertheless, the development cost of a VR scene is high from the perspectives of time and money. Hence, a spherical view VR-based training system of classroom misbehaviour management (TrainCM2) is proposed in this study. TrainCM2 can situate learners in a real situation at a low cost. To verify the effectiveness of TrainCM2, an experiment was conducted. The experimental results showed that the TrainCM2 learners outperformed the micro-teaching learners in terms of misbehaviour detection rate and processing performance, showing the benefits of the proposed approach.