The Internet is needed for everyday life, whereas in Indonesia there are many internet service providers, one of which is indihome. Sentiment analysis itself aims to classify a text into Negative, Positive and Neutral classes. On the twitter platform, there are many reviews about internet providers, one of which is indihome, because of poor service or just to appreciate the services provided. Based on the calculation of the results obtained 71.1% negative, 21.1% positive and 7.7% neutral. The data obtained is not balanced, therefore the classification process is assisted using Smote. The results of the comparison of the four methods used are Support Vector Machine, Naïve Bayes, Random forest, Decision tree. From the overall comparison, the highest accuracy without smote or using smote is Support Vector Machine with an accuracy level of 89% AUC level of 89% if using smote gets 93% accuracy and 97% AUC level with 80% training data and 20% testing.
Copyrights © 2023