Anomaly detection on in-home activities data based on time interval

Abstract

The world population of the elderly is expected to have a continuous growth and the number of elderly living in solitude is also expected to increase in the coming years. As our health decline with age, early detection of possible deterioration in health becomes important. Behavioral changes in in-home activities can be used as an indicator of health decline. For example, changes in routine of in-home activities. Past research mainly focused on detecting anomalies in routine of each type of in-home activities individually. In this paper, an anomaly detection model to detect changes in routine of in-home activities collectively for a day is proposed. The experiment was evaluated with an existing public dataset. The experimental results demonstrated that the anomaly detection model performed well on unseen testing data with an accuracy of 94.44%.