Input Vector Normalization Methods in Support Vector Machines for Automatic Incident Detection

Daehyon Kim, Seungjae Lee and Seongkil Cho

Transportation Planning and Technology, 2007, vol. 30, issue 6, 593-608

Abstract: It is known that support vector machines (SVMs), based on statistical learning theory, are an efficient approach to solving the pattern recognition problem because of their remarkable performance in terms of prediction accuracy. When applying SVMs, the input vectors should be normalized. The prediction performance would differ according to the normalization method used. Thus, it is important to choose an efficient method for normalizing input vectors as this could improve the prediction performance of the SVMs. In this paper, various normalization methods for input vectors have been studied and the best normalization method proposed to achieve the best performance in automatic incident detection. The experimental results show that the performance of an automatic incident detection system using SVMs can be highly dependent on the method used in normalizing the input vectors, and that the proposed normalization method is the most efficient method for automatic incident detection.

Date: 2007
References: View complete reference list from CitEc
Citations: View citations in EconPapers (1)

Downloads: (external link)
http://hdl.handle.net/10.1080/03081060701698235 (text/html)
Access to full text is restricted to subscribers.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:taf:transp:v:30:y:2007:i:6:p:593-608

Ordering information: This journal article can be ordered from
http://www.tandfonline.com/pricing/journal/GTPT20

DOI: 10.1080/03081060701698235

Access Statistics for this article

Transportation Planning and Technology is currently edited by Dr. David Gillingwater

More articles in Transportation Planning and Technology from Taylor & Francis Journals
Bibliographic data for series maintained by Chris Longhurst ().