Comparison Of Classification Methods On Sentiment Analysis Of Political Figure Electability Based On Public Comments On Online News Media Sites

Abstract

Elections are an important part of the political process so that not a few figures from political parties begin preparing to participate in the process. Electability is one of the issues of concern, various things are done to be able to increase the electability of political figures who participated in the general election. Media has become one of the important tools used to increase electability, one of which is online news media. News coverage in online news media with its real-time nature will very quickly get comments from readers and can be used as an assessment of political figures in the form of sentiment analysis. However, it is not easy to analyze sentiments from various comments on online news media, because comments that contain text have irregularities, especially in Indonesian texts. Text mining is one way that can be used to overcome this. Pre-processing of text in text mining is an important part of getting basic information contained in comments. This study uses the Indonesian text pre-processing using the Gata text-mining framework. Then proceed with extracting more in-depth information using the Naïve Bayes classification method and Support Vector Machine which is optimized using Particle Swarm Optimization. The tests carried out with both methods get the results that, the Particle Swarm Optimization based Support Vector Machine method is the best method in the process of classifying sentiments analysis of political figures with an accuracy of 78.40% and AUC 0.850. The results of this study get an effective algorithm in classifying positive and negative comments related to political figures from online news media.