ANALISIS SENTIMEN PENGGUNA TWITTER TERHADAP APLIKASI MYPERTAMINA MENGGUNAKAN METODE K-NEAREST NEIGHBOR DENGAN SELEKSI FITUR CH-SQUARE

Abstract

Pertamina announced a trial policy for using the MyPertamina application for targeted distribution of subsidized fuel in June 2022. This trial was carried out in 11 regions started on July 1, 2022. This policy had been widely discussed on social media Twitter, so there are pros and cons from the community. The large number of public responses in the form of tweets on Twitter requires a sentiment analysis to see an overview of the public's perception of the MyPertamina application. This study used the K-Nearest Neighbor algorithm to classify the public's response to the policy in the form of positive or negative sentiment. This study also added the Chi-Square selection feature to determine the effect on the performance of the K-Nearest Neighbor algorithm. This study used the Python programming language with 2923 tweets as a dataset which had a proportion of 73% positive sentiment and 27% negative sentiment. The research used ratio of 80:20 to the dataset, namely 80% of the training data is 2338 and 20% of the testing data is 585. The resulted of the research on the K-Nearest Neighbor algorithm used K-Fold Cross Validation for 10 broadcasts produced the same average accuracy, namely 77% for both methods, which had the highest accuracy of 88% for K-Nearest Neighbor with Chi-Square and 87% accuracy without used Chi-Square.