Exploiting Marked Temporal Point Processes for Predicting Activities of Daily Living

The increasingly large availability of sensors in modern houses, due to the establishment of home assistants, allow to think in terms of smart houses where behaviours can be automatized based on user habits. Common tasks required to this aim include activity prediction, i.e., the task of forecasting what is the next activity a human is going to perform in the smart space based on past sensor logs. In this paper, we propose a novel activity prediction method for smart houses based on the seminal probabilistic method named Marked Temporal Point Process Prediction.