Analisis Sentimen Larangan Impor Pakaian Bekas Menggunakan Metode Support Vectore Machine dan Lexicon Based

Abstract

X as a social media platform, provides real-time tweets after the event, making it very relevant for sentiment analysis related to certain topics. This research discusses the prohibition on imports of used clothing in Indonesia using the SVM and Lexicon-based methods. The research aims to determine public sentiment regarding this government policy. The SVM method achieved 85.87% accuracy, with 93.83% recall for positive sentiment and 62.00% recall for negative sentiment. There were 76 wrong positive predictions with a precision of 88.11% and 37 wrong negative predictions with a precision of 77.02%. Meanwhile, the Lexicon Method achieved 60% accuracy, with a positive precision of 69% and a negative precision of 31%. Recall for the negative class is 25%, while for the positive class it is 75%. The results of sentiment analysis applying the Support Vector Machine method to build a classification model resulted in 753 data being successfully classified as positive sentiment, while 147 data were classified as negative sentiment. With different accuracies, it shows that sentiment analysis using the Support Vector Machine method has a higher level of accuracy than Lexicon.