Measuring Source Code Similarity Using Reference Vectors

This paper discusses on a method of measuring similarities between program source codes. Unlike many of existing similarity measuring method we do not compare a pair of source codes directly but compare them indirectly with using reference source codes. Using reference vectors calculated from each source codes and reference source codes reduced considerable amount of computation time for similarity measurement. To examine our method, we built a system implemented this algorithm and made computational experiments on Java program source codes submitted as assignments for a programming class. From results, we confirmed there are evident similarities between program source codes have close reference vectors each other