To accurately assess students’ potential knowledge mastery and improve the quality of English test paper generation, this paper combines a knowledge tracking model and a test paper generation model to propose a dynamic resource allocation model for English teaching. The paper first integrates student learning behavior and learning ability into the Knowledge Tracking Machine (KTM) to propose a knowledge tracking model, KTM-LC. Next, based on the improved English knowledge tracking model, a personalized test paper generation model is established under test paper generation constraints, and the improved artificial fish school algorithm is applied to intelligent test paper generation for solution. The paper validates the effectiveness of the KTM-based multi-feature fusion exercise recommendation model on three datasets, finding that with only 10 knowledge state dimensions, KTM-LC achieves an 83.24% test AUC. Additionally, in the fitness function value experiment, 95% of the fitness values based on the method proposed in this paper were above 99, indicating that the algorithm can effectively find test papers that meet the constraints and complete the dynamic allocation of teaching resources.
