Comparison of Machine Learning Algorithms in Analyzing Public Opinion Sentiments Against Fuel Price Increases

Abstract

Twitter is a social media platform that is quite widely used by the world community, especially people in Indonesia. Twitter is one of the social media that provides information, one of which is the increase in the price of crude oil which was recorded at 105 US dollars per barrel. The increase in fuel prices has a negative impact on society, causing pros and cons. Based on these problems, the authors aim to compare the performance of the artificial neural network and naïve Bayes algorithms to determine the best model for sentiment analysis of fuel price hikes. The data used amounted to 1000 datasets in the form of text documents with labeling using the lexicon and split data 90:10, 80:20, 70:30 and 60:40 as a comparison of precision values. The application of word vectorization utilizes TF-IDF in assigning a weight value to each word. Based on the results of the experiments that have been carried out, it is found that the best algorithm using an artificial neural network is capable of producing an accuracy value of 87% for 1000 data on public opinion sentiment on fuel price hikes. Based on the evaluation results, the model built can categorize public opinion sentiment into positive sentiment, negative sentiment, and neutral sentiment automatically and the polarity of public sentiment tends to be positive towards the issue of the fuel price increase that occurred.