Analysis of the Effects Using BERT Feature Extraction Method on Case of Sentiment Classification on Twitter Social Media
Diaz Adha Asri Prakoso, Denny Hermawan, Ardiansyah Musa Efendi · 2023
This study will analyze the effect of the BERT feature extraction method on the evaluation results of the logistic regression, SVM, K-NN, and MLP algorithms. The parameters compared to each algorithm evaluation result are accuracy, recall, precision, specificity, and F1 score. We compared the evaluation results from BERT with TF-IDF feature extraction from several classification algorithms. Based on the results, the effect of using BERT feature extraction causes classification evaluation resulting from the SVM, logistic regression, K-NN, and multilayer perceptron algorithms to decrease. Our best evaluation result is from logistic regression using TF-IDF with 99% accuracy.