Classification of Covid-19 Vaccine Sentiment Using K-Nearest Neighbor and Fasttext on Twitter

Safrizal, Afri Naldi
Surya Agustian

Eksakta : Berkala Ilmiah Bidang MIPA
Vol. 25 No. 03 (2024): Eksakta : Berkala Ilmiah Bidang MIPA (E-ISSN : 2549-7464)

 In late 2019 came a flu-like illness that infected the lungs in the city of Wuhan. It is suspected that the disease is suspected to have originated in bats. WHO named this disease Covid-19 and the virus spread throughout the world, causing a pandemic. The government took a vaccination drive to overcome this virus, but received a response of pros and cons from the public. There are many studies that discuss people's sentiments towards vaccination, one of which is the classification of sentiments. This study discusses the classification of sentiment towards covid-19 vaccines using the K-Nearest Neighbor and Fasttext algorithms on twitter. Data is obtained by crawling using the pyton programming language and Twitter API.  Data labeling is carried out by crowdsourcing and majority voting techniques.  The data used after the balancing process are 6000 training data, 778 development data and 400 test data.  The test results after various experiments and feature engineering got the best results with an accuracy value of 69% and an f1-score of 60%. This result is the best result compared to previous studies with the same dataset.