Improving the Accuracy of Text Classification Using the Over Sampling Technique in the Case of Sinovac Vaccine

The WHO has declared COVID-19 (Coronavirus Disease 2019) a global health emergency. Up to 19 November 2021, the total positive cases in Indonesia reached 4,252,705, of which 4,100,837 recovered, and 143,714 died. Therefore, vaccines have been developed to minimize COVID-19 transmission. There are some kinds of vaccines developed by several companies such as Sinovac, AstraZeneca, Pfizer, and Moderna. The general public has a different opinion on Sinovac vaccine on Twitter, where some people promote it while others reject it. Data used in this study were 1000 tweets about the Sinovac vaccine. During the dataset collection unequal distribution often occurs, where the number of labels is more on one side. Such a situation is called imbalance class. Imbalance class in a dataset can reduce classification performance. To overcome the imbalance class, this study used the Synthetic Minority Over-sampling Technique (SMOTE). The classification methods used were K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Random Forest, and TF-IDF was used to determine the weight of the words. The average rise of the accuracy value of the three algorithms after SMOTE optimization was 14%. The results of sentiment analysis for the Sinovac vaccine revealed a positive sentiment of 81%. Thus, it can be concluded that the Sinovac vaccine received a positive response from the public.