|
Title:
|
DEVELOPMENT AND EVALUATION OF A METHOD FOR IDENTIFYING LOGIC ERRORS USING MACHINE LEARNING WITH A FOCUS ON PROGRAM STRUCTURE |
|
Author(s):
|
Yuta Harada, Soichiro Sato, Shoichi Nakamura and Youzou Miyadera |
|
ISBN:
|
978-989-8704-62 |
|
Editors:
|
Paula Miranda and Pedro Isaías |
|
Year:
|
2024 |
|
Edition:
|
Single |
|
Keywords:
|
Programming Learning Support, Logic Error, Identifying Stumbling, Source Code Editing History, Machine Learning |
|
Type:
|
Full |
|
First Page:
|
219 |
|
Last Page:
|
227 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
In programming exercise classes at universities, instructors must quickly determine the learner's situation and provide
support. However, it is difficult to deal with learners making logic errors. This is because there is more than one way to
solve a task that a learner is working on, and the method of program logic coding (hereinafter, "coding method") differs
for each student. Therefore, when instructors help learners solve logic errors, they must be supported by considering each
learner's coding method. This study aims to develop a method for identifying logic errors that learners are currently making,
considering each learner's coding method. In this study, we assume that the learner's coding method is reflected in the
program structure. We then attempt to analyze the program focusing on its structure. Specifically, using clustering, we tried
to categorize source code by the learner's coding method and identify logic errors in it. Results of evaluating our method
using the Hold-out Validation show it can identify logic errors with 70% accuracy. This result indicates a new approach
for instructors to help learners when they are stuck during programming practice. |
|
|
|
|
|
|
