Sentiment Analysis of Super App Polri Reviews Using a Hybrid CNN-BiLSTM Model

Digital transformation in public services has led the Indonesian National Police (POLRI) to introduce the Super App Polri as an integrated online service platform providing services such as Police Clearance Certificate (SKCK), Driving License (SIM), and Vehicle Registration Certificate (STNK). User experiences vary and are reflected in reviews on the Google Play Store. This study analyzes user sentiment toward the Super App Polri using a deep learning approach. A Convolutional Neural Network-Bidirectional Long Short-Term Memory (CNN-BiLSTM) model with FastText word embedding is employed, with hyperparameter optimization conducted using a random search strategy. The proposed model achieves an accuracy of 95.47% and an AUC of 98.34% on the training dataset, while obtaining an accuracy of 92.28% and an AUC of 96.08% on the testing dataset. The small performance gap between training and testing results indicates strong generalization capability. These findings demonstrate that the CNN-BiLSTM model is effective and robust for sentiment analysis of public service applications.