Earticle

초록

영어

Based on the problem that data simple weight is unchanged in current feature selection algorithms, we proposed a new boosting feature selection algorithm framework using data sampling technique to describe the change of data sample weight in the process of feature selection which can solve the problem that dynamic mutual information is sensitive to noise data. This method evaluates mutual information value among features by dynamic adjust data sample weights in the process of feature selection so that every selected feature can describe characteristics of data classification more precisely. Efficiency of this algorithm has been testified by experience results on UCI common test data sets.

목차

Abstract
 1. Introduction
 2. Related Works
  2.1. Feature Selection
  2.2. Boosting
 3. Feature Selection Based on Dynamic Weight
  3.1. Weight Updating
  3.2. The Proposed Algorithm
 4. Experiments and Analysis
  4.1. Test Dataset
  4.2. Experiment Setting
  4.3. Experimental Results
 5. Conclusion
 References

키워드

저자

• Cheng Zengping [ Hubei University of Automotive Technology, ShiYan City, Hubei Province 442000, China ]

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  International Journal of Signal Processing, Image Processing and Pattern Recognition

• 간기

  격월간

• pISSN

  2005-4254

• 수록기간

  2008~2016

• 십진분류

  KDC 505 DDC 605