Knowledge tracing utilizes historical learner response records to diagnose learners’ knowledge states and predict their subsequent responses. Existing models have not effectively extracted the learner-centric information embedded in the learning data, thereby neglecting the impact of learner-specific features on learning outcomes. To accurately model learners’ knowledge mastery, we propose Learner Empowered Knowledge Tracing model (LEKT). First, the model derives learner-centric attributes from complex behavioral data, providing a more comprehensive depiction of the learning process. Second, a gated fusion mechanism is employed to explore the intrinsic relationships among different learner-centric attributes, resulting in a fused feature representation of the learners’ knowledge state. Third, the model reconstructs the information propagation process of graph neural networks by concatenating global and local features, thereby integrating the overall knowledge level of the learner. Finally, experiments on multiple public datasets show that the LEKT model can better model learner subject characteristics and track learners’ learning status.