Dynamics-Aware Gated Graph Attention Neural Network for Student Program Classification and Knowledge Tracing
Tiancheng Jin, Liang Dou, Guang Yang, Aimin Zhou, Xiaoming Zhu, Chengwei Huang · Atlantis Highlights in Computer Sciences/Atlantis highlights in computer sciences · 2023
Faced with a large number of questions on the programming OJ (Online Judge) system, students are usually mindless when choosing questions, which is not conducive to helping students quickly improve their programming ability.Programming Knowledge Tracing (PKT) is a technology that dynamically traces students' programming knowledge states using their historical learning data including submitted programs.Relying on PKT, OJ can find students' unmastered knowledge points, and recommend questions examining these knowledge points to students, so as to help students overcome their weakness.However, existing program analysis modules in PKT models ignore dynamic information of program.Therefore, this paper proposes Dynamics-Aware Gated Graph Attention Neural Network (DGGANN), which inputs test cases of questions into program, obtains call frequency coefficients of every node in Abstract Syntax Tree (AST) through code coverage statistical tool, and introduces such call frequency information into process of program analysis.This paper applies DGGANN to two tasks in our experiments: classifying programs by functionalities and PKT.Experimental results show that our approach can achieve higher performance than the state-ofthe-art models in both tasks on datasets of two well-known OJ systems named CodeForces and Libre.