Knowledge Tracing Modeling Method for Programming Task

Knowledge Tracing is used to assess student learning progress and predict student performance. It analyzes the student's learning history to infer the degree to which the student has mastered the knowledge points. Existing knowledge tracking models often have various limitations. First, most knowledge tracking models take question number and student's binary answer as the input mode, ignoring other information of the student's answer, and second, most KT models are based on few attempts, so the current KT model cannot adapt to More complex answering situations in multiple task. This paper studies the KT model in the field of computer education. By analyzing the characteristics of the programming task KT data set, a modeling method for programming task is proposed. First, we got the students' accurate score information based on the code text's similarity. Then, we extracted student's multiple attempt sequence features based on the sliding window, which improved the model's ability to solve long-sequence problems. Experiments on three standard KT models showed that this modeling method effectively improved model performance. model performance.