Analisis Sentimen Pada Media Sosial Twitter Menggunakan Naive Bayes Classifier Dengan Ekstrasi Fitur N-Gram

Abstract

Social media is currently an online media that is widely accessed in the world. Microblogging services such as Twitter allow users to write about various things they experience or write reviews of a product, service, public figures and so on. This can be used to take opinion or sentiment towards an entity that is being discussed on social media such as Twitter. This study utilizes these data to determine public opinion or sentiment regarding public perceptions of the issue of rising electricity tariffs. Opinion taking is based on three classes namely positive, negative and neutral. Users often use non-standard word abbreviations or spelling, this can complicate the process and accuracy of classification results. In this study the authors apply text-preprocessing in handling these problems. For feature extraction, n-gram and classification methods are used using the Naive Bayes classifier. From the results of the research that has been done, the most negative sentiments are formed in response to the issue of the increase in basic electricity tariffs. In addition, from the results of testing with the method of cross validation and confusion matrix it is known that the accuracy of the naïve Bayes method reaches 89.67% before applying n-gram, and the accuracy rate increases 2.33% after applying n-gram characters to 92.00%. It is proven that the application of the n-gram extraction feature can increase the accuracy of the naïve Bayes method.