Analisis Sentimen Twitter menggunakan Metode Naive Bayes dengan Relevance Frequency Feature Selection (Studi Kasus: Opini Masyarakat mengenai Kebijakan New Normal)

Kresentia Verena Septiana Toy (Fakultas Ilmu Komputer, Universitas Brawijaya)
Yuita Arum Sari (Fakultas Ilmu Komputer, Universitas Brawijaya)
Imam Cholissodin (Fakultas Ilmu Komputer, Universitas Brawijaya)

Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer
Vol 5 No 11 (2021): November 2021

COVID-19 first appeared in wuhan, China and spread around the world. The virus spread very rapidly, including in Indonesia. The Indonesian government seeks to implement policies to suppress the COVID-19 case increase. Implemented policies have a new impact on communities, such as job downsizing, layoffs of work relationships and other effects on the country's economy. Consequently, the government adopted a new policy called new normal. New normal has become a topic of debate among the public on twitter's social media. Public opinion can be classified into positive, negative, and neutral opinions and require analytic sentiments. The sentiment analysis process is based on pre-processing for opinion processing, Relevance Frequency Feature Selection to reduce the number of features, and the classification using Naive Bayes methods. The dataset is 300 public opinion data, with the distribution of data using k-fold validation in k=5. The results of 5 tests using Naive Bayes classification, obtained an average accuracy of 62,6%, while the results of classification accuracy tests with the addition of Relevance Frequency Feature Selection obtained an average accuracy of 65,3%.