The current level of cooperation between industries and education in the sphere of college and university integration can no longer meet modern social needs due to the continued growth of social economy and artificial intelligence. As a response to this issue, this paper explores a digital twin system designed to enhance the collaborative performance of industry-education integration. Utilizing the theory behind digital twin systems, the research defines the main functional modules necessary to implement this application, obtains research data using web crawling, and creates a resource recommendation module of industry-education integration using the combination of LDA model with a content-based recommendation algorithm. Through the support of development software and image processing methods, the practical training module is developed and implemented. Based on this, an enhanced LSTM network is proposed and a back propagation Rprop algorithm is used to develop a student management early-warning module, thus finalizing the whole system module design. Then, the experimental simulation is made taking into consideration the system functional modules. The simulation outcomes show that the prediction errors of red, orange, yellow, and green warnings are 0.0315, 0.0233, 0.0309 and 0.0045 respectively, which are all less than 0.05. These results confirm the usefulness of the student management early-warning module of the digital twin system, as well as offer the foundation of similar optimization measures. The study helps in the enhancement of practical competence of students and problem-solving abilities, and its ultimate aim is to contribute to the improvement of the collaborative performance of industry-education integration.
