Detection of plagiarism in students´ programs using a data mining algorithm

Studies have shown that many students have similar programs in programming class, most of which due to plagiarism. Students may simply modify others´ programs as their own. This makes the assessment standards for students´ programs with lots of ambiguity and uncertainty, limiting assessment accuracy and efficiency, and reducing the reliability of test results. To solve this problem, a student program plagiarism detection approach is proposed based on a data mining algorithm. Firstly, similar code fragments are mined by the CloSpan algorithm. Then, similarities between programs are calculated. Finally, the plagiarism list is output. Experiments showed that compared with the widely used plagiarism detection tool MOSS, our approach is can not only more accurately give statistical information of the similar program detected, but also be able to visualize the similar code fragments, which can greatly increase detection efficiency.