ANALISIS SENTIMEN ULASAN PENGGUNA TERHADAP APLIKASI LINKEDIN MENGGUNAKAN PENDEKATAN BERT
No Thumbnail Available
Date
2023-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Elfitra
Abstract
Google Play Store, one of the largest digital distribution platforms for downloading and
uploading developed applications, was the focus of this research aimed at obtaining
sentiment analysis results of LinkedIn application users. Sentiment analysis was used to
evaluate the response of LinkedIn users and enable developers to improve the quality of
LinkedIn applications in the future. User review data was collected through the scraping
method and included 4,333 reviews, of which 2,296 were positive, 543 were neutral, and
1,494 were negative. Positive sentiments dominated the proportion of sentiments, at 53%.
The model and method used for analysis were the pre-trained BERT model, known as
IndoBERT, with average accuracies of 84% for validation data, respectively. An
evaluation of the model through a confusion matrix yielded an accuracy of 80%, precision
of 90%, recall of 84%, and F1-Score of 87%.
Description
Keywords
BERT, Confusion Matrix, IndoBERT, LinkedIn, NLP, Sentimen Analysis
Citation
Perpustakaan