Cryptocurrency Sentiment Classification Based on Comments On Facebook Using K-Nearest Neighbor

Ramu Will Sandra (Universitas Islam Negeri Sultan Syarif Kasim)
Yelfi Vitriani (Universitas Islam Negeri Sultan Syarif Kasim)
Muhammad Affandes (Universitas Islam Negeri Sultan Syarif Kasim)
Suwanto Sanjaya (Universitas Islam Negeri Sultan Syarif Kasim)

IJISTECH
Vol 6, No 2 (2022): August

Cryptocurrencies continue to develop and have received world attention, price changes that occur every day are influenced by uncertain factors such as political problems and global economic problems. The author will explore the problems discussed by the public regarding positive and negative cryptocurrency comments on Facebook comments using the K-Nearest Neighbor method. This study uses 1000 data comments which are divided into 500 positive data and 500 negative data. The data was obtained manually by using the keyword "bitcoin price" on social media facebook. The results of the testing process using the confusion matrix get the highest accuracy at a comparison of 90: 10 by 62%, recall 70%, error rate 38% and precision 60,34% with k value of 11 and threshold 9.