Modeling and detecting urinary anomalies in seniors from data obtained by unintrusive sensors

In this project, we use unintrusive sensors to collect data about toilet attendance of seniors as a proxy for micturition, in order to detect anomalous behaviour. Firstly, we identify and address challenges associated with building a robust dataset of normal toilet-attendance behaviour from sensor logs. Next, since our users are healthy, we leverage medical information to build personalized simulated models of abnormal toilet attendance on the basis of users’ normal behaviour. We then compare the performance of two anomaly-detection models in detecting abnormal increases in toilet visits.