System for user context determination in a network of IoT devices

Abstract

In order to build a user profile using data from various connected IoT smart sensors and devices, determination of the current context of the user is vital. We assume a hierarchy of contexts (such as party, trip, exercise) based on common daily activities of users. Knowing the context can inform about the actual activity being performed by the user and predict what the user might be interested in at a given moment. This can then be used to suggest appropriate services to the user. In this paper, we propose a system to infer the user context from input data from various devices. Our system includes an app classifier, a Points of Interest (POI) classifier and a motion classifier to make sense of the input sensor data. We describe the implementation details of a system and some results on real world data to measure our model performance.