Bridging Data and Behavior in Homecare: Personalized Routine Modelling and Anomaly Interpretation

Area of Science:

• Gerontology
• Computer Science
• Artificial Intelligence

Background:

• Understanding older adults' daily life organization is vital for personalized homecare solutions.
• Smart home sensor data offers a rich source for analyzing independent living patterns.
• Behavioral anomaly detection is key to proactive health monitoring in aging populations.

Purpose of the Study:

• To propose an interpretable framework for modeling daily life activities in older adults.
• To detect behavioral anomalies using smart home sensor data.
• To identify and explain variations in daily routines and lifestyle archetypes.

Main Methods:

• Utilized data from 18 CASAS smart homes with continuous sensor recordings.
• Applied Principal Component Analysis (PCA) to analyze daily activity patterns in 15-minute intervals.
• Defined personal baseline routines and compared deviations against population-level patterns.

Main Results:

• Identified three distinct lifestyle archetypes: active bimodal, stable routine, and early resting.
• Revealed explainable behavioral differences among residents.
• Linked deviation scores to specific activities like sleep, hygiene, and computer use.

Conclusions:

• The proposed framework offers an interpretable method for understanding and monitoring older adults' daily lives.
• It enables the detection and explanation of irregular behaviors, supporting person-centered homecare.
• The identified lifestyle archetypes provide insights into diverse aging-in-place patterns.