Professional certification requires that the diagnosis, training implementation and continuous improvement of teaching ability of geography normal students form a clear evidence chain. This paper constructs an analysis framework that integrates graph attention, contrastive learning and multi-source training records for the major of Geographic Science (normal) in Suzhou University. The model organizes the information such as course performance, microtraining, educational practice, competition performance, tutor evaluation and graduate tracking into a teaching ability relationship graph, identifies the ability status, and generates a hierarchical training path and resource allocation scheme. The implementation basis shows that the major has 22 full-time teachers and 11 doctors, 1 training center and 6 professional practice and training rooms, with an experimental area of about 2220 square meters. A total of 3478 graduates have been trained, and 65.2% of them are engaged in the education industry over the years. The framework can transform scattered evidence into traceable, analysis-able and traceable certification diagnosis basis under the goal-oriented and trinity training system, and provide a path for the organization, support and improvement of teaching ability training.
