Mining Patterns in Sensor Data for Personalized Healthcare
Pattern mining in time series data is a widely researched problem domain and has significant impact in the field of healthcare. The huge potential of extracting knowledge from patients' sensor data, along with the ubiquitous use of wearable health devices has encouraged and accelerated innovative solutions for challenges in diagnostic procedures and treatment delivery. Such healthcare solutions are enhanced by customizing them to suit each patient's specific health conditions and requirements. In this work, we address a variety of problems associated with the mining of patterns from sensor generated time series data and propose efficient solutions for application in personalized healthcare. These problems include interactive pattern discovery, multidimensional pattern mining, enhanced query based pattern mining, pattern sequence mining, predictive modeling and detection of anomalous patterns. Each of the studies included in this work provide a detailed overview of the problem addressed, the proposed solution and its evaluation, and its impact in enhancing personalized healthcare. Overall, the work presented here provides an insightful assessment of the benefits of using pattern mining in improving healthcare, and presents a compelling case for personalization of diagnosis and treatment.