Comparison of support vector machine and K-Nearest Neighbor algorithms in Indonesia hoax classification

Abstract

Today, more and more people spread the word and read news from social media. This is because on social media it is easier to spread the news and to get news. So, to overcome these problems, it is necessary to classify the news contained on social media whether the news is hoax news or facts because it is important to assess the level of truth of a news spread on social media. To classify which news is hoax or fact, manually will spend a lot of time, so one solution to make it easier to classify hoaxes or facts is to use a machine learning approach. In this study, researchers used the SVM and KNN algorithms to classify hoax news in Indonesian. From the two algorithms, we will look for which model is the best in classifying hoax news. Researchers use four categories of data that have their own characteristics, in this study several stages were carried out starting from the preprocessing stage and using TF-IDF calculations in measuring the weight of each news, then the learning process was carried out using the Support Vector Machine algorithm and K-Nearest Neighbor and evaluation of the model by using the confusion matrix. Based on the results of the model evaluation, it was found that the classification with the SVM algorithm obtained a higher accuracy value than KNN, with an accuracy value of 84% and f-measure of 84% in the third data category.