School registration management is an important part of the school management process, which is the basis of other work. This paper proposes an innovative school registration management system that integrates the school registration early warning system into the one-stop school registration management model. The data mining method is utilized to mine and process students’ school registration information to extract useful information. On this basis, the academic registry management system and the academic registry early warning system are constructed, and the graph theory-based subspace clustering algorithm is introduced to realize the innovative academic registry management model integrating academic registry early warning. Behavioral data provided by the information center of University A is used as the basis for the analysis of students’ campus card usage data. The subspace clustering algorithm proposed in this paper can obtain accurate results for clustering students’ usage behaviors, and the accuracy of the algorithm is consistently maintained between 80% and 90%. Then the academic data of the 2018 to 2021 graduates in University A is used as an experimental analysis object to predict the clusters to which the performance points belong, and the clusters generated by clustering are mined to find out the common characteristics of outstanding students and the relationship between the student data and the academic data. Finally, feature data clustering is performed on students’ academic characteristics to obtain information on students with abnormal academic status, and the implementation of dynamic monitoring of academic status reveals that 95% of students do have abnormal academic characteristics.