Twitter Sentiment Analysis of Public Space Opinions using SVM and TF-IDF Methods

Abstract

Public space opinion reviews are currently a source of information for interested parties and decision-makers. Twitter is a social media that is a means of expressing themselves for people to express their opinions and criticize the current situation. This becomes information for readers. Information published on Twitter contains elements of commentary on a situation or object Sentiment analysis of public space opinion on Twitter using Machine Learning with the Support Vector Machine (SVM) method with the data weighting process using the Term Frequency-Inverse Document Frequency (TF-IDF) method. Dataset obtained by scraping using the Twitter API as much as 5000 data then labeled where the goal is to get accuracy on positive, negative, or neutral sentiment. The results of research conducted experiments on three Machine Learning algorithms with the extraction function "TF-IDF" obtained an accurate training model with good classification capabilities, especially SVM of 91,6% on data distribution 70: 30; SVM is 92.8% in the case of data distribution of 80: 20; the SVM is 91,8% in the case of 90:10 decomposition data.