AI-Powered Sentiment Analytics in Banking: A BERT and LSTM Perspective.

In recent years, the banking industry has witnessed a surge in digital feedback channels, where customers regularly share their experiences and opinions. Extracting meaningful insights from this unstructured data is vital for enhancing customer satisfaction and service quality. This study presents a comparative analysis of two advanced deep learning models—Long Short-Term Memory (LSTM) and Bidirectional Encoder Representations from Transformers (BERT)—to classify the sentiment of bank customer reviews into positive, negative, and neutral categories. A cleaned and preprocessed dataset consisting of real-world customer reviews was used for model training and evaluation. The LSTM model was able to capture sequential patterns effectively, achieving competitive results in sentiment prediction. However, BERT outperformed LSTM across all evaluation metrics, achieving higher accuracy, precision, recall, and F1-score. Detailed confusion matrix analysis further confirmed BERT’s superiority in handling ambiguous and context-rich sentiment expressions. The findings highlight the practical implications of using transformer-based models in financial text analytics and provide a reliable framework for future sentiment analysis applications in the banking sector.