Comparison of Accuracy and Time Of Naïve Bayes Algorithm with Support Vector Machine Algorithm in Twitter Sentiment Analysis of Peduli Lindungi Application

Abstract

The Peduli Protect application is a government program to monitor and cope with the spike in Covid-19 in Indonesia. However, this application also reaps controversy in the community so that it invites various opinions among the community, one of which is on Twitter. Twitter has become the most popular web and smartphone-based media with more than 200 million users and more than 10.6 billion tweets that have been generated. This information can be used to analyze public opinion and opinion on the Peduli Protect application through sentiment analysis using a classification of income into two classes, namely positive and negative. Sentiment analysis is needed to find out the general assessment of an object. Support vector machine algorithm defines a good hyperplane that gives the SVM algorithm a good accuracy level compared to other algorithms. In contrast, the Naive Bayes algorithm is a machine learning classification algorithm with probability reasoning that is not inferior to other algorithms. From the results of this study, the accuracy shows that the Naïve Bayes method is superior by 80% without validation testing using K - Fold Cross-Validation and 85% with K-Fold Cross Validation on the fold-3. While the calculation of Naïve Bayes processing time is also superior by the getting time of 0.009365 seconds, the Support Vector Machine algorithm is tested for K-Fold-Cross Validation getting 86% in the 3rd iteration.