A Benchmark of Modeling for Sentiment Analysis of the Indonesian Presidential Election in 2019

Abstract

Researching with a machine learning method approach, the truth is to try to solve a case by using various algorithmic approaches to obtain the most suitable model for a case. In this research, we want to know which process of modelling that has the best accuracy value for classifying emotions in the text. The algorithm used is using the LSTM algorithm, while the benchmarking that we tested is the Random Forest and Naive Bayes algorithm. This research takes public opinion about the 2019 Indonesian Presidential Election by classifying it into four types of emotions: Happy, sad, angry, and afraid. The data we use contains more than 1200 Indonesian tweets. In this experiment, we achieved an accuracy of 68.25% using the Random Forest model, whereas, with the Multinomial Naïve Bayes model, the accuracy was 66%.