The pervasiveness of technology in our life and low cost tracking devices has allowed us to keep a record of our daily activities and behaviours in a seamless way. An interesting use of this data is the discovery of patterns of behaviour or Pattern-Of-Life (POL) with many applications in numerous fields that range from personalized healthcare applications, mobile security and a more collective use in traffic analysis.
Changes or abnormalities/anomalies, by which we mean a behaviour not represented by the model, can be an indicator of an interesting event in any domain. When applied to Pattern- Of-Life, preventive and corrective actions can be taken such as the generation of an alarm. For example a Pattern-of-Life approach could be incorporated to an assistive robot to learn the usual behaviours of a person and provide either personalized response based on the user’s preference, or proactive actions when an anomalous conduct is detected.
When revealing human patterns, it is important to understand which factors govern an individual’s behaviour. Patterns emerge at different temporal scales due to the periodic modulation that characterise human nature. This means, it is also relevant to include contextual information such as the type of the week, or the time of the day. This can reveal cyclic behaviour patterns not considered by prior work.