Analisis Sentimen Media Sosial Twitter Dengan Algoritma K-Nearest Neighbor Dan Seleksi Fitur Chi-Square (Kasus Omnibus Law Cipta Kerja)

Abstract

The use of information technology is growing rapidly, marked by public opinion that can be conveyed indefinitely through social media. One of the social media that is Twitter. Twitter is considered easier to retrieve information related to existing opinions and sentiments due to the limited character in a tweet made by users and there is the hashtag # which can be searched easily regarding the current hotly discussed situation. Some time ago, there was a lot of discussion regarding the ratification of the omnibus work copyright law. Tweets in the form of lively sentiments adorn the hashtag “#omnibuslaw” and other related hashtags. This study discusses reviews in the form of tweets related to omnibus law with the Chi Square selection feature and the K-Nearest Neighbor algorithm using R Studio. Data were taken as many as 500 tweets related to the omnibus law. The methodology used is Knowledge Discovery in Databases. Data labeling is carried out by experts who are divided into positive and negative sentiments. The results of the modeling using K-Fold Cross Validation, the highest accuracy is obtained with a 25% feature use scheme (Chi Square feature selection), and the value of k = 5 in KNN is 81.4%. Testing on the model was carried out using 100 random data and obtained 83% accuracy, 100% precision, 15% recall and 26.08% F-Measure value. Of the 500 data taken, the word people is the most dominating word. Of the 500 data taken, 78.8% negative sentiments and 21.2% positive sentiments.