Natural Language Processing for Prediction of Multi-Occupancy Activities of Daily Living

Prediction of Activities of Daily Living (ADLs) is a common problem, predominantly with its use-case in healthcare. Close monitoring of patients is essential when they are incapable of being independent. Several ADL studies have been conducted on single residents. However, in this study, we will consider multi-occupancy activities with two residents. We implement a Natural Language Processing (NLP) methodology that uses pattern recognition to predict the activity of two residents simultaneously. We rely on sensor sequences as input for both residents within the smart home. We use an encoder-decoder architecture to perform pattern recognition of the sensor sequences. The results of this method are satisfactory for both predicted tasks. Finally, we also propose to generalize the procedure for the n-resident case and discuss the possible challenges while scaling for multiple residents.