This study applied VR technology to the ERP simulation experiment teaching, introducing the optimized collision detection algorithm. In the established virtual ERP simulation experiment environment, the data visualization operation is carried out on the very important data content in the ERP experiment, so that the system can predict the students’ operation in real time when the students interact. Thus, it reduces the unreality of the interaction. In addition, we improve the AABB bounding box based on B+ tree storage to improve the efficiency of inter-model collision detection and reduce the occupation and consumption of system memory during detection. Finally, the framework of Bayesian predictive filtering algorithm is introduced to predict the students’ simple interactive operation and the orientation of object model. This model optimizes the whole immersive experience teaching of ERP based on VR technology, and makes a certain contribution to meet the talent demand in the context of big data.