Analisis Kinerja Algoritma Winnowing pada Pendeteksian Plagiarisme

Abstract

– The problem of plagiarism at this time can be solved. This research was carried out by analyzing the performance of the Winnowing Algorithm on plagiarism detection. The choice of the Winnowing Algorithm is because by calculating the hash value of k-grams, the rolling hash function is used to find the hash value, after which a window is formed from the hash value. In each window, the smallest hash value is selected. When more than one hash has the lowest value, the rightmost hash value is selected. Then all selected hash values are stored to be used as document fingerprints. This fingerprint serves as a basis for comparing the similarity of embedded text. The results of the Winnowing Algorithm performance analysis research show that this algorithm is quite good at detecting text similarity or plagiarism with the result of the smallest percentage value of 10 tests shown in the 10th test with n-gram values n = 10, window w = 3, time process sec = 0.0094 with a result of 47% text similarity. This result is better than the results of the similarity of the Rabin Karp Algorithm from the results of previous studies.