Fine-Tuning BERT for Sentiment Analysis in Public Transportation

Public transportation systems, especially in bustling cities like Bangkok, rely heavily on passenger feedback to enhance service quality and meet rising demands. Traditional sentiment analysis models, such as Logistic Regression, Support Vector Machines (SVM), and Long Short-Term Memory (LSTM) networks, often struggle with the nuances and complexities of human language, particularly in multilingual and context-specific datasets. This study investigates the effectiveness of fine-tuning a pre-trained Bidirectional Encoder Representations from Transformers (BERT) model for sentiment analysis of passenger feedback within Bangkok's public rail transportation system. We collected a diverse dataset from various online platforms frequented by local residents and international tourists, addressing challenges like data imbalance and language translation through rigorous preprocessing. The fine-tuned BERT model was evaluated against baseline models using standard metrics: accuracy, precision, recall, and F1-score. Experimental results demonstrate that the fine-tuned BERT model significantly outperforms traditional models, achieving an accuracy of 90.45% and an F1-score of 90.38%. Visualizations using confusion matrices and ROC curves further corroborate its superior performance. These findings underscore the potential of advanced NLP models like BERT in capturing nuanced sentiments in passenger feedback, providing transportation service providers with more accurate and actionable insights to enhance service quality and passenger satisfaction.