Evaluation of Machine Learning Performance Based on BERT Data Representation with LSTM Model to Conduct Sentiment Analysis in Indonesian for Predicting Voices of Social Media Users in the 2024 Indonesia Presidential Election

Abstract

In this era, the innovation of science and technology has changed rapidly, such as Artificial Intelligence (A.I.) which has helped a lot in human life. Deep learning (DL) as part of A.I. is the development of one of the machine learning models, namely the neural network. With many neural network layers, deep learning models can perform feature extraction and classification processes in a single architecture. This model has proven to perform state-of-the-art machine learning techniques in text classification, pattern recognition, speech, and imagery. Various text classification tasks, including sentiment analysis, have gone beyond AI-based approaches in Deep Learning models. Text data can come from multiple sources, including social media, such as comments on videos on YouTube. Sentiment analysis, one of the opinion mining, is a computational study that analyzes people's opinions from texts. In this research, machine learning performance analysis is carried out on a deep learning method based on BERT data representation with the LSTM method. The implementation of the model uses youtube commentary data on political videos related to the 2024 presidential election in Indonesia; performance analysis is carried out using accuracy, precision, and recall metrics. In this study, the accuracy of the BERT-LSTM model outperformed the BERT model with an accuracy of 0.8783.