Pendeteksian Penjiplakan Kode Program C dengan K-Means

Abstract

The practice of plagiarism in programming assignments can be easily done by students. Thus, a system that can detect this practice, i.e., by grouping similar programs into a cluster, is needed. K-Means clustering is a method of flat clustering that has been used to determine similar codes based on the structure of the source codes. The objective of this research is to make the checking process automatic. This research uses a sample of 92 source codes in C comprising 9 groups based on the source code structure similarity. In this study we conducted two experiments, i.e., by determining the number of clusters (manual) and without determining the number of clusters (automatic). The results indicate that the manual experiment resulted in a higher Rand Index (91.35%) and a faster execution time. The system is able to perform automatic clustering process with a Rand Index of 90.63%. The automated clustering does not have significant difference of Rand index, implying the performance is good enough for the clustering without determining the K value first