Implementation of Support Vector Machine (SVM) Algorithm for Spam Comment Classification on Instagram

Abstract

Instagram is a photo and video sharing application that allows users to take photos, take videos, apply digital filters, and share them on various social networking services. Public figures,  specially actors, actresses, influencers and Instagram celebrities, use Instagram as a platform to promote their various activities or works to their fans or followers. However, there are a lot of spam comments on these posts, both good and bad. This research obtained comment data from Instagram to be used as a dataset, namely from public figures, actresses and actors who have more than 5 million followers. By using preprocessing (data cleansing, tokenization, case folding and stopword removal) and TF-IDF for weighting, selecting the kernel you want to use and training the SVM model in the classification process. We need a system that can classify whether comment are spam or not spam. The method used is Support Vector Machine is a machine learning method that has the working principle of structural risk minimization aimed at finding hyperplanes or separators between two classes in an input space. From the test results it was found that the performance of the proposed an accuracy level of 96.82%, precision 94%, recall 92%, and f1-score 93%.