Cited 0 time in webofscience Cited 0 time in scopus

Comparison evidential fusion network with decision tree for reliable contextual information

Title
Comparison evidential fusion network with decision tree for reliable contextual information
Authors
Lee, HyunSon, Byung RakKim, Jung EunLee, Sang Cheol
DGIST Authors
Lee, Hyun; Son, Byung Rak; Kim, Jung Eun; Lee, Sang Cheol
Issue Date
2011
Citation
2011 International Conference on Automation and Robotics, ICAR 2011, 122 LNEE, 513-520
Type
Conference
Article Type
Conference Paper
ISBN
9780000000000
ISSN
1876-1100
Abstract
In intelligent environments, detecting possible errors and completing missing values then deciding about the quality and validity of the sensed data are roles of context reasoning in order to improve the reliability of information. Particularly, learning to associate operations and roles with dynamic contexts is necessary for adapting user needs and making optimal classification to context-aware computing. However, a top-down based inference has difficulty to deal with various kinds of uncertainty such as unknown, ambiguous, imprecise, and erroneous related problems. Since misclassification by high-level classification to choice for learning the role acceptance may lead to wrong decision. Thus, we propose Evidential Fusion Network (EFN) that uses a belief or confidence level of information as a bottom-up based inference. In order to analyze advantage and disadvantage of the proposed EFN, we compare our approach (bottom-up) with decision tree approach (top-down). Finally, we suggest the combination of top-down based analysis and bottom-up based inference as a context reasoning method for obtaining reliable information. © 2011 Springer-Verlag.
URI
http://hdl.handle.net/20.500.11750/3910
DOI
10.1007/978-3-642-25553-3_63
Publisher
Springer
Related Researcher
Files:
There are no files associated with this item.
Collection:
Convergence Research Center for Future Automotive Technology2. Conference Papers


qrcode mendeley

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE