Analysis of User Complaints for Telecommunication Brands on X (Twitter) using IndoBERT and Deep Learning

Hakim, Valianda Farradillah
Riana, Dwiza

Jurnal Nasional Pendidikan Teknik Informatika (JANAPATI)
Vol. 13 No. 2 (2024)

Tweeting on different official accounts is what users of Twitter (X) do most frequently. These tweets ranging from compliments to critiques. One of the official accounts that gets a lot of tweets from its customers is Telkomsel, an Indonesian telecom company. This study aims to find the maximum accuracy that can be obtained by combining CNN and Bi-LSTM algorithms with IndoBERT embeddings. A considerable accuracy level above 90% is demonstrated by the study, with CNN obtaining the greatest accuracy of 99% at a learning rate of 6*10^-5, along with scores of 98%, 97%, and 97% for precision, recall, and F1 correspondingly.