AbstractsComputer Science

The Application of Sequential Pattern Mining in Healthcare Workflow System and an Improved Mining Algorithm Based on Pattern-Growth Approach

by Qi Zhang




Institution: University of Cincinnati
Department: Engineering and Applied Science: Computer Science
Degree: MS
Year: 2013
Keywords: Computer Science; Workflow; Windows Workflow Foundation; Frequent Itemset; Sequential Patterns; Projected Database; Frequent Pattern Growth
Record ID: 2010970
Full text PDF: http://rave.ohiolink.edu/etdc/view?acc_num=ucin1378113261


Abstract

Workflow technology has broadened substantially into the healthcare industry in recent years. Hospitals and other medical units are advocating this technology as a means to improve operational efficiency, achieve patient safety goals and positively influence the quality of care. As workflow management concepts and workflow-based techniques are being applied in the healthcare information systems, there is also a growing interest in applying the data analysis and knowledge discovery techniques, such as sequential patterns mining techniques, to support the use of large healthcare information databases, which can be made more efficient when synchronized with workflow system.This thesis introduces workflow techniques and their application in Healthcare information system, addresses the opportunities that workflow technology has to make a profound impact on medical care system while examining the challenges that are presented in the healthcare arena. It also addresses the concepts for sequential pattern mining and its widely used algorithms, compares the performance of the algorithms and indicates their preferred application domains. Based on a popular approach, an improved algorithm, called the transformation–based frequent pattern growth (T-FPG), was proposed which has the potential to be more efficient in mining large sequence databases with numerous patterns and/or long patterns than other classic methods. All the features of this algorithm are illustrated, and the experimental results of the T-FPG algorithms show that it outperforms PrefixSpan, the typical FR-growth algorithm for sequential patterns mining.