Evaluating User Satisfaction in The Halodoc Application Using a Hybrid CNN-BiLTSM Model for Sentiment Analysis

The growing demand for digital healthcare services in Indonesia has driven the adoption of Online Healthcare Applications (OHApps) such as Halodoc. Despite over 65 million users, maintaining user satisfaction remains a challenge. This study employs sentiment analysis using a hybrid Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) model to classify user review ratings. A dataset of 10,000 Google Play Store reviews was divided into COVID-19 and post-pandemic segments. The methodology includes data collection, pre-processing, and dataset segmentation for training, validation, and testing. Results indicate that the CNN-BiLSTM model surpasses traditional machine learning by combining CNN’s feature extraction with BiLSTM’s long-term dependency capture, achieving 98.71% accuracy on COVID-19 data and 98.16% post-pandemic. Additionally, the model demonstrates strong performance across other key evaluation metrics, with precision, recall, and F1-score. Misclassification analysis highlights minor errors, particularly in ratings 4 and 5. These findings help healthcare providers enhance digital services by identifying user concerns, improving platform features, and optimizing customer engagement. Beyond healthcare, this approach has real-world applications in e-commerce and financial services, where sentiment analysis informs user experience improvements.