Comparison of Seven Machine Learning Algorithms in the Classification of Public Opinion

Abstract

Sentiment analysis is one way that is widely used to identify the beginning of public opinion in various fields of life which are associated with very massive and a lot of information through social media. This study aims to compare several algorithms in machine learning to see the best ability in sentiment classification. The research dataset uses a dataset of public opinion related to tourism in Indonesia. The number of datasets used is 10,228 twitter data that have been cleaned and labelled. The machine learning algorithm used is Logistic Regression, KNN, AdaBoost, Decision Tree, SVM, Random Forest and Gaussian. The seven algorithms for sentiment classification from the Twitter public opinion each produce a Gaussian accuracy of 0.52; SVM 0.78; KNN 0.98; Logistic Regression, Random Forest, Decision Tree, AdaBoost of 0.99. This study shows that the selection of the right machine learning algorithm will have a very good impact on the classification of public opinion through social media