Analisis Sentimen pada Media Sosial Twitter terhadap Aturan Pemberlakuan Pembatasan Kegiatan Masyarakat Kota Malang dengan Metode K-Nearest Neighbor

Hanif Nabila Muflih (Fakultas Ilmu Komputer, Universitas Brawijaya)
Widhy Hayuhardhika Nugraha Putra (Fakultas Ilmu Komputer, Universitas Brawijaya)
Issa Arwani (Fakultas Ilmu Komputer, Universitas Brawijaya)

Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer
Vol 5 No 10 (2021): Oktober 2021

The increasing number of Corona Virus spreads in early 2021 because there were many violations of health protocols became the basis for the rules for the Implementation of Community Activity Restrictions (PPKM). PPKM rules are implemented with the aim of reducing the transmission of the virus. PPKM until March 2021 was successfully implemented according to its objectives, marked by a decrease in the number of virus spread from Malang City initially being in the red zone to the green zone. Information development related to PPKM can be followed through social media. Twitter is a social media platform that connects users through instant and short message exchange communications. From social media Twitter users can send text messages called tweets with a maximum limit of using characters up to 280. Twitter as a social media to find out information and public opinion that develops about PPKM rules. Each tweet has a positive and negative sentiment value. Sentiment analysis aims to classify an opinion written through a tweet. This study analyzes tweets with the topic of PPKM Malang City. Sentiment analysis is carried out by classifying text data in the form of tweets into positive and negative sentiment classes. The classification process is implemented in RapidMiner Studio software with the K-Nearest Neighbor classification method and Chi Square feature selection. The main processes in sentiment classification are preprocessing, term or text weighting, feature selection and classification. Testing and analysis were carried out using three methods, namely testing based on comparison with manual calculations, testing based on variations in the value of k, and testing based on variations in the percentage of features. The best accuracy result from the classification process is 82%. The accuracy results were obtained during testing with a value of k = 3 and a feature percentage of 100%. The percentage of features used in the classification process affects the accuracy where the smaller the feature value used, the smaller the classification accuracy value.