<![CDATA[Along with the development of technology, social media is now often used to provide an assessment, one of which is Twitter is one of the most popular microblogging. On Twitter, users don't just review a product, but they often complain or share their experiences about their level of satisfaction while using Tokopedia. Sentiment analysis on tokopeida provides useful indicators for various purposes which can be found in comments, feedback or criticism. This study uses the Naïve Bayes and Random Forest algorithms to provide the expected classification results, the analysis will be carried out by comparing several combinations of algorithms that will be tested on twitter tweets about Tokopedia, including the combination with Synthetic Minority Oversampling Technique (SMOTE) to optimize unbalanced data. In the Tokopedia Twitter tweet test, SMOTE was able to increase the accuracy of the Naive Bayes algorithm to 86.93%, an increase of 3.4% from the previous 83.53%. Meanwhile, Random Forest with SMOTE has an accuracy value of 88.44%, an increase of 1.55% from the previous Random Forest test of 86.89%.]]>
Copyrights © 2022
