Regulation Document Search Based on Themes using Cosine Similarity and Naive Bayes

Ministry of Administrative and Bureucratic Reform main task is to make policies. The policies that are produced are Regulations that can be in the form of Undang-Undang (Act), Peraturan Pemerintah (Government Regulation), Peraturan Presiden (Presidential Regulation), and Peraturan Menteri (Ministerial Regulation). In making these regulations, one of the main concern is to look for regulation that has the same theme as the regulation that is still being drafted, so that the regulation that will be produced later, did not conflict or overlap with existing regulation. So far the search process take a long time because the search process is only based on the title, type and year in which the regulation created. Furthermore, the search process will not produce results if you don't use the right keywords. And after we have succeeded in finding the regulations that we are looking for, in order to facilitate the consultation on these regulations, each regulation document must be classified based on the unit that made it.This paper discusses the search for regulation based on themes using the Cosine Similarity and then classify the regulation document using the Naïve Bayes. The result is that the searching process using Cosine Similarity only requires less than 1 second and the classification of the regulations documents using Naïve Bayes has an accuracy of 82%. The conclusion is that Cosine Similarity method can be used in the Regulation Document searching process. And Naïve Bayes method can also be used in Regulation document classifying process.