Multigraph-based Deep Programming Ability Tracing Method for Students
Programming has become a crucial ability with the development of artificial intelligence, making it increasingly important to track and improve programming proficiency of students. However, most programming ability tracing methods rely primarily on the final outcomes of programming exercises to meas...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Tamkang University Press
2025-01-01
|
| Series: | Journal of Applied Science and Engineering |
| Subjects: | |
| Online Access: | http://jase.tku.edu.tw/articles/jase-202508-28-08-0019 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Programming has become a crucial ability with the development of artificial intelligence, making it increasingly important to track and improve programming proficiency of students. However, most programming ability tracing methods rely primarily on the final outcomes of programming exercises to measure the programming
proficiency, and neglect the rich behavioral structure information derived from the programming process, leading to a suboptimal solution in the programming ability tracing. To this end, we propose a multigraphbased deep programming ability tracing method for students (MDPAT), which consists of four components. Specifically, MDPAT devises the multigraph programming modelling via conducting the program knowledge graph and the program iteration graph to generate the submission representations of programming exercises, which effectively captures static structure information and dynamic structure information within the programming process. Then, MDPAT designs the programming knowledge gated update and the programming ability
gated update to aggregate information of the programming knowledge and the programming ability that are derived from submission representations, which achieves a balanced and comprehensive tracking of evolving programming proficiency. Meanwhile, MDPAT utilizes the programming answer prediction to generate final programming proficiency measurement of students. Four components of MDPAT collaborate organically to achieve maximum performance in the programming ability tracing. Finally, extensive results on two real world
datasets, especially on the Atcoder_C dataset, ACC shows a 5.06% improvement compared to the second best result, verify that MDPAT conducts a new standard baseline in the programming ability tracing. |
|---|---|
| ISSN: | 2708-9967 2708-9975 |