Heterogeneous Ensemble for Feature Drifts in Data Streams

  • Hai-Long Nguyen
  • Yew-Kwong Woon
  • Wee-Keong Ng
  • Li Wan
Part of the Lecture Notes in Computer Science book series (LNCS, volume 7302)

Abstract

The nature of data streams requires classification algorithms to be real-time, efficient, and able to cope with high-dimensional data that are continuously arriving. It is a known fact that in high-dimensional datasets, not all features are critical for training a classifier. To improve the performance of data stream classification, we propose an algorithm called HEFT-Stream (Heterogeneous Ensemble with Feature drifT for Data Streams) that incorporates feature selection into a heterogeneous ensemble to adapt to different types of concept drifts. As an example of the proposed framework, we first modify the FCBF [13] algorithm so that it dynamically update the relevant feature subsets for data streams. Next, a heterogeneous ensemble is constructed based on different online classifiers, including Online Naive Bayes and CVFDT [5]. Empirical results show that our ensemble classifier outperforms state-of-the-art ensemble classifiers (AWE [15] and OnlineBagging [21]) in terms of accuracy, speed, and scalability. The success of HEFT-Stream opens new research directions in understanding the relationship between feature selection techniques and ensemble learning to achieve better classification performance.

Keywords

Feature Selection Data Stream Feature Subset Concept Drift Ensemble Learning 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Bifet, A., Holmes, G., Kirkby, R.: Moa: Massive online analysis. The Journal of Machine Learning Research 11, 1601–1604 (2010)Google Scholar
  2. 2.
    Bifet, A., Holmes, G., Pfahringer, B., Kirkby, R., Gavald, R.: New ensemble methods for evolving data streams. In: 15th ACM SIGKDD, pp. 139–148. ACM (2009)Google Scholar
  3. 3.
    Breiman, L.: Bagging predictors. The Journal of Machine Learning Research 24(2), 123–140 (1996)MathSciNetMATHGoogle Scholar
  4. 4.
    Breiman, L.: Random forests. The Journal of Machine Learning Research 45(1), 5–32 (2001)MATHCrossRefGoogle Scholar
  5. 5.
    Domingos, P., Hulten, G.: Mining high-speed data streams. In: The Sixth ACM SIGKDD, pp. 71–80. ACM (2000)Google Scholar
  6. 6.
    Domingos, P., Pazzani, M.: On the optimality of the simple bayesian classifier under zero-one loss. The Journal of Machine Learning Research 29(2-3), 103–130 (1997)MATHGoogle Scholar
  7. 7.
    Eibl, G., Pfeiffer, K.-P.: Multiclass boosting for weak classifiers. The Journal of Machine Learning Research 6, 189–210 (2005)MathSciNetMATHGoogle Scholar
  8. 8.
    Freund, Y., Schapire, R.: Experiments with a new boosting algorithm. In: The 13th ICML, pp. 148–156 (1996)Google Scholar
  9. 9.
    Friedman, J.H.: Stochastic gradient boosting. Computational Statistics & Data Analysis 38(4), 367–378 (2002)MathSciNetMATHCrossRefGoogle Scholar
  10. 10.
    Fumera, G., Roli, F.: A theoretical and experimental analysis of linear combiners for multiple classifier systems. IEEE Transactions on Pattern Analysis and Machine Intelligence 27(6), 942–956 (2005)CrossRefGoogle Scholar
  11. 11.
    Hsu, K.-W., Srivastava, J.: Diversity in Combinations of Heterogeneous Classifiers. In: Theeramunkong, T., Kijsirikul, B., Cercone, N., Ho, T.-B. (eds.) PAKDD 2009. LNCS, vol. 5476, pp. 923–932. Springer, Heidelberg (2009)CrossRefGoogle Scholar
  12. 12.
    Hulten, G., Spencer, L., Domingos, P.: Mining time-changing data streams. In: ACM SIGKDD, pp. 97–106. ACM (2001)Google Scholar
  13. 13.
    Lei, Y., Huan, L.: Feature selection for high-dimensional data: A fast correlation-based filter solution. In: The 20th ICML, pp. 856–863 (2003)Google Scholar
  14. 14.
    Liu, H., Yu, L.: Toward integrating feature selection algorithms for classification and clustering. IEEE Transactions on Knowledge and Data Engineering 17(4), 491–502 (2005)CrossRefGoogle Scholar
  15. 15.
    Oza, N.C.: Online bagging and boosting. In: 2005 IEEE International Conference on Systems, Man and Cybernetics, vol. 3, pp. 2340–2345. IEEE (2005)Google Scholar
  16. 16.
    Sattar, H., Ying, Y., Zahra, M., Mohammadreza, K.: Adapted one-vs-all decision trees for data stream classification. IEEE Transactions on Knowledge and Data Engineering 21, 624–637 (2009)CrossRefGoogle Scholar
  17. 17.
    Shen, C., Li, H.: On the dual formulation of boosting algorithms. IEEE Transactions on Pattern Analysis and Machine Intelligence 32(12), 2216–2231 (2010)MathSciNetCrossRefGoogle Scholar
  18. 18.
    Street, W.N., Kim, Y.: A streaming ensemble algorithm (sea) for large-scale classification. In: The 7th ACM SIGKDD, pp. 377–382. ACM (2001)Google Scholar
  19. 19.
    Tin Kam, H.: The random subspace method for constructing decision forests. IEEE Transactions on Pattern Analysis and Machine Intelligence 20(8), 832–844 (1998)CrossRefGoogle Scholar
  20. 20.
    Tumer, K., Ghosh, J.: Linear and order statistics combiners for pattern classification. Springer (1999)Google Scholar
  21. 21.
    Wang, H., Fan, W., Yu, P.S., Han, J.: Mining concept-drifting data streams using ensemble classifiers. In: ACM SIGKDD, pp. 226–235. ACM (2003)Google Scholar
  22. 22.
    Woods, K., Philip Kegelmeyer, J.W., Bowyer, K.: Combination of multiple classifiers using local accuracy estimates. IEEE Transactions on Pattern Analysis and Machine Intelligence 19(4), 405–410 (1997)CrossRefGoogle Scholar
  23. 23.
    Zhenyu, L., Xindong, W., Bongard, J.: Active learning with adaptive heterogeneous ensembles. In: The 9th IEEE ICDM, pp. 327–336 (2009)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2012

Authors and Affiliations

  • Hai-Long Nguyen
    • 1
  • Yew-Kwong Woon
    • 2
  • Wee-Keong Ng
    • 1
  • Li Wan
    • 3
  1. 1.Nanyang Technological UniversitySingapore
  2. 2.EADS Innovation WorksSingapore
  3. 3.New York UniversityUSA

Personalised recommendations