The efficacy of “busyness” as a measure for behaviour pattern analysis using unlabelled sensor data: a case study

The integration of Internet of Things (IoT) into our daily lives has the potential to significantly improve the quality of life and well-being of citizens, especially older adults who want to continue to live independently in their own home. In order to realise the potential of IoT in the real-world, there are four major concerns that should be considered: ensuring sensors are un-intrusive, the resources required for the installation and maintenance of sensors, the large amounts of unlabelled data produced and the noise in the data that is typical of real environments. In this research we investigate a measure of the interaction with everyday objects -“busyness”, generated from a set of discrete wireless sensors that addresses these issues. We present a case study to assess the feasibility and advantages of using a minimal number of sensors to develop a “busyness” measure for recognising specific activities (hot drink making) over a short learning period. The results of these case studies show that the “busyness” metric at different levels of granularity can be used to monitor the user's habits using a commercial IoT system and aggregate data.