Skip to Main Content
In many data stream applications, data segments which are sequential and complicatedly changeable always imply great domain specific value. Especially in the field of medical survey, mining such sequential data segments will help making diagnosis. We discovered that, based on extensive analysis, although containing rich semantics, these data segments are actually composed of some certain basic units, and these units can form different kinds of complex patterns with duplication or lack in certain positions considering a temporal logic. Therefore, we present a scalable pattern mining method. With this method, the Scalable Pattern Tree (SPTree) structure is designed to support the expression of scalable semantics and efficient mining. At last, the experimental results on real datasets prove that our method is feasible and efficient.
Date of Conference: 29-31 May 2012