Skip to main navigation Skip to search Skip to main content

F-score-like measure: A new measure for spam filtering

  • Yong Han*
  • , Hao Liang Qi
  • , Mu Yun Yang
  • *Corresponding author for this work
  • Heilongjiang Institute of Technology
  • School of Computer Science and Technology, Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Logistic average misclassification percentage (lam%) and I-AVC(area under the ROC curve) are two important and wildly adopted measures. This paper demonstrates that a spam filter can achieve a perfect 0.00% in lam%, the minimal value in theory, by simply setting a biased threshold during the classifier modeling. At the same time, I-AVC is left untouched; and the overall classification performance reaches only a low accuracy. This means that lam% and I-AVC as main measures for spam filtering are not suitable. To solve the problem of measuring spam filtering, F-score-like measure based on ham and spam misclassification is proposed to be a single measure for spam filtering evaluation.

Original languageEnglish
Title of host publicationProceedings of 2012 International Conference on Machine Learning and Cybernetics, ICMLC 2012
Pages2047-2051
Number of pages5
DOIs
StatePublished - 2012
Externally publishedYes
Event2012 International Conference on Machine Learning and Cybernetics, ICMLC 2012 - Xian, Shaanxi, China
Duration: 15 Jul 201217 Jul 2012

Publication series

NameProceedings - International Conference on Machine Learning and Cybernetics
Volume5
ISSN (Print)2160-133X
ISSN (Electronic)2160-1348

Conference

Conference2012 International Conference on Machine Learning and Cybernetics, ICMLC 2012
Country/TerritoryChina
CityXian, Shaanxi
Period15/07/1217/07/12

Keywords

  • Evaluation measure
  • F-score-like measure
  • I-AVC
  • Logistic average misclassification percentage
  • Spam filter

Fingerprint

Dive into the research topics of 'F-score-like measure: A new measure for spam filtering'. Together they form a unique fingerprint.

Cite this