Akademik Veri Yönetim Sistemi
Sakarya University Journal of Computer and Information Sciences, cilt.8, sa.2, ss.301-311, 2025 (Scopus)
This study focuses on developing a digital twin for baby incubators in neonatal intensive care units to enhance monitoring and care for premature infants. The digital twin employs a hybrid model integrating Long Short-Term Memory (LSTM) and Random Forest (RF) algorithms to predict potential errors and alarms. The LSTM algorithm was trained using sensor data provided by a health technology company to predict future measurements. Subsequently, the RF algorithm classifies these predictions into specific error conditions. The hybrid model demonstrates success with mean squared error and mean absolute error values of 1540533.6 and 160.8 for the LSTM model and an 86.44% accuracy rate for the RF model. The study's key findings emphasize the effectiveness of the hybrid model in predicting future sensor values and classifying errors, representing a significant step towards improving premature baby care. Integrating LSTM and RF algorithms offers an innovative approach to error prediction, minimizing risks and improving premature infant health outcomes. In summary, this study successfully develops a digital twin for baby incubators, offering a promising solution for advancing newborn healthcare services and providing a foundation for future research.