Robert Moskovitch

Proactive fall prevention in care homes is a major healthcare challenge. This study leverages a unique, large-scale database collected via a mobile care monitoring application, capturing the daily living activities of over 68,000 residents in multiple UK care homes. To utilize these irregular and heterogeneous event streams effectively, we propose a continuous prediction framework that transforms raw records into symbolic time intervals using temporal abstraction and mines frequent Time Interval Related Patterns (TIRPs). Based on frequent TIRPs that end with a fall, the Fully Continuous Prediction Model (FCPM) learns a continuous prediction model. Then in real-time, while instances of those patterns unfold, it estimates the probability of the completion of the patterns, meaning that the fall will occur. In this study, we propose enhancement for the FCPM: the use of a supervised state abstraction method. A rigorous evaluation on a large real-life database shows that the use of the TD4C abstraction performs better than EWD, EFD, and SAX. Finally, the FCPM performs better in comparison to the baseline models, including the deep learning sequence-based model (LSTM-FCN) and the feature-based classifiers (XGBoost, ROCKET), achieving above 78%AUC.