Title :
Mathematical Analysis on Weight Vectors in Text Classification
Author :
Fengxi Song ; Qinglong Chen ; Zhongwei Guo ; Xiumei Gao
Author_Institution :
Dept. of Autom. & Simulation, New Star Res. Inst. of Appl. Tech. in Hefei City, Hefei, China
Abstract :
By means of rigid mathematical deductions we prove that weight vectors cannot promote the performance of the optimal classifier, i.e. the Bayesian classifier in terms of the error, F-one score, or breakeven point. The conclusion is important in that people used to promote the performance of a classifier by trying various weight vectors in text classification.
Keywords :
Bayes methods; classification; mathematical analysis; text analysis; Bayesian classifier; F-one score; breakeven point; mathematical analysis; mathematical deductions; text classification; weight vectors; Bayesian methods; Classification algorithms; Mathematical model; Support vector machine classification; Text categorization; Training; Vectors; evaluation; optimal classifier; text classification; weight vector;
Conference_Titel :
Intelligent Systems (GCIS), 2012 Third Global Congress on
Conference_Location :
Wuhan
Print_ISBN :
978-1-4673-3072-5
DOI :
10.1109/GCIS.2012.14
