基于频繁闭合序列模式挖掘的学生程序雷同检测

Plagiarism detection in student programs based on frequent closed sequence mining

针对学生程序抄袭导致考核可信度降低而人工检测抄袭工作量巨大的问题,提出了程序雷同检测模型,首先通过词法分析将程序转换成token序列,并将其散列映射为数字序列;然后采用BIDE挖掘算法挖掘频繁闭合序列;在此基础上,识别相似代码片段,并计算程序之间的相似度,进而判定程序是否雷同。实验结果表明,与目前应用广泛的雷同程序检测工具MOSS相比,本文方法提高了雷同检测的准确性,不但可以准确地给出雷同统计信息,还能够较为直观地显示雷同代码片段。

Plagiarism in student programs is a common phenomenon, which decreases the credibility of assessment. However, manual detection loads a heavy burden on the teachers. To solve this problem, a plagiarism detection model is proposed. First, student programs are converted into token sequences through lexical analysis. Then, the token sequences are hashed to digital sequences. Then, the frequent closed sequences are mined by the BIDE algorithm. On this basis, the similar code fragments are detected and the plagiarism programs are identified by the calculated similarity. Experimental results show that, compared with the commonly used toll MOSS, the proposed method is more precise. It can not only give accurate statistical information of similar programs, but also explicitly display the plagiarized code fragments.