Akademik Veri Yönetim Sistemi
30th Signal Processing and Communications Applications Conference, SIU 2022, Safranbolu, Türkiye, 15 - 18 Mayıs 2022
© 2022 IEEE.In this study, it was aimed to detect sepsis before it occurs and thus reduce the mortality rate by enabling early treatment. The inference performance of the machine learning method was increased by monitoring the demographics and vital signs of the patients and various laboratory test results collected at regular intervals with time-based feature modeling. A sliding window approach was proposed for this modeling so as to include the past six hours, and this feature vector has been used to classify whether or not sepsis will develop in patients. Model performance was evaluated from different perspectives on the basis of numerous studies proposed in recent years, and it was concluded that the proposed approach resulted in a significant performance improvement.