Mining productive-periodic frequent patterns in tele-health systems

Recently, tele-health systems have gained attention from vast research fields because they facilitate remote monitoring of patients (e.g. vital sign data, physical activities, etc.) by utlizing various technologies such as body sensor network, wireless communications, multimedia and human-computer interactions without interrupting the quality of lifestyle. As tele-health generates a huge amount of healthcare data consisting of much useful information, finding hidden information from the data is an important task. The purpose of this work is to facilitate a real-time warning alarm in the context of tele-health remote monitoring using data mining techniques. This can be utilized for the e-wellbeing applications, for example, rehabilitation, early identification of therapeutic issues and emergency warning. In particular, we focus on mining Productive Periodic frequent patterns from incremental databases (such as vital sign data of patients) for various decision makings. Exploring the correlations between periodic frequent vital sign data or items is important since the inherent relationships between the items of patterns are relevant. To mine the correlated periodic frequent patterns from incremental databases, we introduce the productive (i.e. useful) periodic frequent patterns (PPFP) as the set of periodic frequent patterns with periodicities that result from the occurrence of correlated items. We finally design and develop an efficient PPFP mining technique that can mine the complete set of useful periodically occurring patterns in incremental databases. Numerous experiments were performed on both real and synthetic data set to judge the effectiveness of the proposed pattern mining procedure when contrasted with existing best in class approaches.