Measurement of the Similarity of Indonesian Papers on One Journal Topic with the Naive Bayes Algorithm and Vector Space Model

Abstract

Abstract— One way to maintain the quality of scientific work in Indonesia is by checking articles before they are published. Checking before the publication was done so that the similarity level is not high because the published papers can be quoted to cause a high level of similarity. The next problem is the importance of grouping topic papers, where papers to be checked should have the same category as comparative papers. In this study, to classify the topic of the journal using the Naïve Bayes algorithm and to measure the similarity of papers using the Vector Space Model method. Naïve Bayes algorithm can better classify the test data with the .docx file format than to the test data in the .pdf file format. The results of the calculation of text similarity detection by the Vector Space Model can reach 90% and above for test data with the .docx file format, while for test data with the .pdf file format the calculation results by the Vector Space Model are on average less than 90%. The results of the calculation of text similarity detection by the Vector Space Model method are also strongly influenced by training data. The more complete and complex of the training data, then more valid the results of the Vector Space Model performance testing