Mining Periodic Changes in Complex Dynamic Data Through Relational Pattern Discovery

The empowerment of the information technologies in many real-world applications has opened to the possibility of tracking complex and evolving phenomena and gather information able to describe such phenomena. For instance, in bio-medical applications, we can monitor a patient and collect data that range from his clinical picture to the laboratory studies on biological products. In this scenario, studying the possible alterations manifested over time becomes thus relevant and, in life sciences, even determinant. In this paper, we investigate the task of determining changes which are regularly repeated over time and we propose a method based on two notions of patterns, emerging patterns and periodic changes. The method works on a time-window model to the end of i capturing statistically evident changes and ii detecting their periodicity. The method was applied to two typical real-world scenarios with complex dynamic data, that is, Virology and Meteorology.