Identifying Varying Health States in Smart Home Sensor Data: An Expert-Guided Approach

The aging population is growing and innovative solutions are needed to address older adults’ complex health needs while concurrently extending the reach of the nurse. One emerging solution is the healthassistive smart home. The smart home uses ambient sensors to monitor the movement of older adults and intelligent algorithms to detect changes in health states. Alerts are provided to patients, family and caregivers so older adults can receive timely interventions. Adding a clinician-in-the-loop when training machine learning algorithms may improve the machines ability to accurately identify and predict changes in health states that have clinical relevance. At Washington State University, the CASAS team uses a clinical nurse-expert in a guided approach to machine learning. Here, we describe the expert guided approach, discuss current challenges and offer suggestions for future machine learning research in the area of health-assistive smart homes.