Detection of Anomalous Vital Sign of Elderly Using Hybrid K-Means Clustering and Isolation Forest

Age-related changes to vital signs indicate the possibility of a health condition which requires attention. A deviation from normal in vital signs might be an anomaly and indicate an important warning sign of changing health and indicators of the severity of illness that needs an immediate and reflexive response. Once anomalies are detected at the right time, the system will result in reflexive stimulus and will inform it to the care giver. It leads to the good and fast response of outcomes to save the patient’s life. In this study, we proposed hybrid technique of K-Means clustering and isolation forest to detect anomalies. To evaluate the reliability of proposed hybrid technique, we compare existing isolation forest algorithm and hybrid technique of K-Means clustering and isolation forest on labeled datasets obtained from public. The results show that our hybrid technique is more sensitive in detecting anomalies. Applied on some labelled data, hybrid technique has lower error rate. For some labelled data, the hybrid technique has high error rate.