Earticle

초록

영어

In the data stream classification process, in addition to the solution of massive and real-time data stream, the dynamic changes of the need to focus and study. From the angle of detecting concept drift, according to the dynamic characteristics of the data stream. This paper proposes a new classification method for data stream based on the combined use of concept drift detection and classification model. The data stream classification model can’t adapt to concept drift problem to solve. Before the model classification, the use of information entropy to judge the data block concept drift, the concept of history to have appeared, the use of a classifier pool mechanism to save it, to makes the classification model has stronger resistance to concept drift.

목차

Abstract
 1. Introduction
 2. Detection Method for Concept Drift
  2.1. Classification model based on integration learning
  2.2. Classification model based on incremental learning
  2.3. Method for detecting concept drift
 3. Classification Model and Visualization Approach based on Concept Drift Detection Method
  3.1. Classification Model based on KL-distance
  3.2. Visualization Method for Concept Drift
 4. Experimentation and Result Analysis
  4.1. Data for experiment
  4.2. Experiment results of artificial data
  4.3. Experiment results of real data
  4.4. Experiment results of forgetting mechanism
 5. Conclusion
 References

키워드

저자

• Li Xiaofeng [ Department of Informatic Science, Heilongjiang International University, Harbin 150025, China ]
• Gao Weiwei [ Department of Informatic Science, Heilongjiang International University, Harbin 150025, China ]

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  International Journal of Multimedia and Ubiquitous Engineering

• 간기

  월간

• pISSN

  1975-0080

• 수록기간

  2008~2016

• 등재여부

  SCOPUS

• 십진분류

  KDC 505 DDC 605