Implementasi SSD Mobilenet V3 Untuk Deteksi Tingkat Kepuasan Pasien di Klinik Berdasarkan Ekspresi Wajah Secara Real-Time

Abstract

Patient satisfaction is one of the key indicators in assessing the quality of services in healthcare facilities, including clinics. Satisfaction can be measured through both verbal and non-verbal approaches. Verbal methods, such as questionnaires or interviews, may introduce bias and fail to accurately represent a patient's true level of satisfaction. As an alternative, a non-verbal approach—specifically, real-time facial expression detection—offers a more objective means of evaluation. This study aims to detect and classify patient satisfaction levels based on facial expressions by implementing the Single Shot Multibox Detector (SSD) with a MobileNet V3-Large backbone. The model was trained using a balanced dataset of 35,000 images consisting of seven facial expression classes: happy, neutral, angry, sad, fearful, disgusted, and surprised. These were then grouped into three categories: satisfied, neutral, and dissatisfied. Each category included 5,000 images, divided into 3,000 for training, 1,000 for validation, and 1,000 for testing. The model training process involved data augmentation and fine-tuning techniques to enhance performance. Evaluation results show that the model achieved an accuracy of 94.54% on the test data. These results indicate that the SSD MobileNet V3-Large-based facial expression recognition model is effective in automatically and in real-time classifying patient satisfaction levels, and has potential for implementation in clinical environments as a service quality evaluation support system.