Implementation of Text Mining for Sentiment Analysis of Online Lectures During the Covid-19 Pandemic

Abstract

Strategy against the spread of the Covid-19 virus in Indonesia by enacting Large-Scale Social Restrictions. The implementation of the Scale Social Restrictions forced all universities in Indonesia to close their institutes and conduct lectures online. Online lectures are considered as a solution to continue the teaching process during a pandemic. However, the lack of adaptation and sudden changes caused various responses and public opinions on social media. For this reason, this study aims to conduct text mining on Twitter in order to analyze public sentiment on the topic of "online lectures" the data obtained are classified into 2 classes (positive and negative). The results of the accuracy of the nave Bayes test with the cross validation technique obtained a result of 81.57%. For class precision, positive predictions are 100%, while for negative predictions the results are 73.06% and recall from true positive is 63.13% for true negative is 100%. And for the accuracy of K-Nearest Neighbor 62.10%, for class precision positive prediction is 62.06% while for negative prediction results are 62.13% and recall from true positive is 62.24% for true negative is 61.95%