Comparison of The Use of Bigrams and Stopword Removal for Classification Using Naive Bayes (Case Study on Sentiment Analysis of By.U Internet Users)
Manaarul Hidayat, Rahmat Hidayat, Dwi Otik Kurniawati · 2021
With the rapid growth of the number of youth, several providers have launched digital service provider innovations. By.U as a pioneer has been serving this segment for the past year, and its presence has received various responses from netizens. This response, if researched, can improve services and lead to other innovations. However, no research addresses this. Therefore, a final project was written regarding Sentiment Analysis on the provider by.U. In this reearch, a classification model was made with 3804 data using Naive Bayes Classifier and TF-IDF with Bi-Grams. The comparison was made by eliminating one of the preprocessing steps: stopword removal. The comparison found that TF-IDF and Bi-Grams without the application of stopword removal had the best performance values. The best performance value is obtained in the split validation scenario 90:10 with accuracy 86.88%, Precision 88.24%, recall 83.33%, and f1-score 85.71%.