Home // DBKDA 2016, The Eighth International Conference on Advances in Databases, Knowledge, and Data Applications // View article

A Novel Reduced Representation Methodology for Provenance Data

Authors:
Mehmet Güngören
Mehmet Sıddık Aktaş

Keywords: scientific workflows, scientific data provenance

Abstract:
Learning structure and concepts in provenance data have created a need for monitoring scientific workflow systems. Provenance data is capable of expanding quickly due to the catch level of granularity, which can be quite high. This study examines complex structural information based provenance representations, such as Network Overview and Social Network Analysis. Further examination includes whether such reduced provenance representation approaches achieve clustering effective for understanding the hidden structures within the execution traces of scientific workflows. The study applies clustering on a scientific dataset from a weather forecast to determine its usefulness, compares the proposed provenance representations against prior studies on reduced provenance representation, and analyzes the quality of clustering on different types of reduced provenance representations. The results show that, compared to prior studies on representation, the Social Network Analysis based representation is more capable of completing data mining tasks like clustering while maintaining more reduced provenance feature space.

Pages: 15 to 22

Copyright: Copyright (c) IARIA, 2016

Publication date: June 26, 2016

Published in: conference

ISSN: 2308-4332

ISBN: 978-1-61208-486-2

Location: Lisbon, Portugal

Dates: from June 26, 2016 to June 30, 2016