부산대학교 도서관

Diagnosing patients in a post-operative situation, more accurate and timely, is one of the health care challenges where computer science should help. According to studies conducted in some countries, patients are concerned about to stay longer time in the hospital after surgery, mainly for financial reasons. Reduction on hospital stay also reduces the risks of hospital-acquired infections. In this context, we tried to verify the impact of Internet of Things (IoT) devices and inference techniques in order to improve the diagnosis procedure. This paper presents METAPHOR (MultiagEnT Architecture POstoperative Referral) as an evolution of a previous proposed model applying machine learning classifier algorithms in order to indicate the best referral for postoperative patients based on on-line medical information. The best prediction classification algorithm reached an accuracy of almost 90% in the evaluated cases, evidencing the possibility of IoT medical devices can be used to improve post-operative patients management. To confirm the results, the Precision, Recall and F1 Score of each estimator where analyzed as well.