Adaptive Bayesian Knowledge Tracing Based on Personalized Characteristics

Bayesian Knowledge Tracing (BKT) model is a student knowledge state model that currently widely used in intelligent teaching system. However, BKT still has limitations in capturing dynamic changes during the student learning process and handling multiple characteristics as well as individualized needs. To address the above problems, this paper proposes an adaptive BKT model based on students' personalized features. The core of the model lies in constructing a feature set based on students' previous personalized question-answering data, and dynamically classifying students' current learning situation accordingly. The dynamic classification method classifies students by analyzing their question-answering performance in real time and adjusts the BKT parameters accordingly to more accurately reflect their current knowledge status. On this basis, it predicts students' future learning status and question-answering performance. Comparative experiments show that the model can better adapt to individual learning differences and significantly improve the accuracy of predicting students' future question-answering correctness.