The game Mobile legends: Bang Bang generates millions of user reviews
containing valuable sentiment for developers, but its analysis is hindered by large data
volumes and the use of complex informal language. This research aims to design and
implement an automated sentiment analysis system using the deep learning
Transformers BERT model to classify user reviews into positive, negative, and neutral
categories. The research method includes collecting 548,250 review data from Kaggle,
text preprocessing to handle noise, and training the IndoBERT model through a finetuning
technique. The model's performance was evaluated using accuracy, precision,
recall, and F1-score metrics. The results indicate that the fine-tuned IndoBERT model
achieved an accuracy of 83.2%, a significant improvement of +62.8% compared to the
pre-trained base model, which only reached 20.4%. This success demonstrates that the
fine-tuning process effectively adapts the model to understand the unique jargon and
context of game reviews. The entire research workflow is implemented in an interactive
web application using Streamlit as a proof of concept and a tool for visualizing the
analysis results.