Research on Multi-Dimensional Bayesian Network Classifiers Based on ICA Dimension Reduction

Abstract:

Multi-dimensional Bayesian network classifiers (MBCs) are probabilistic graphical models proposed to solve classification problems. However, in data analysis and preprocessing tasks, one is often confronted with the problem of selecting features from very high dimensional data. To resolve this problem, the covariance analysis and the FastICA algorithm are applied to decrease the dimension and remove redundant information. And then, we only need to construct class subgraph and bridge subgraph of the MBC model with algorithm and mutual information from the processed data, since the new feature variables satisfy independence assumption. The experiment was tested on three benchmark data sets. The theoretically and experimental results show that our method outperforms other state-of-the-art algorithms for multi-dimensional classification in accuracy.

• Periodical: Applied Mechanics and Materials (Volumes 380-384)
• Pages: 2593-2596
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.380-384.2593
• Online since: August 2013
• Authors: Xiu Fang Zhang, You Long Yang, Xing Jia Tang
• Keywords: Covariance Analysis, Independent Component Analysis (ICA), Multi-Dimensional Bayesian Network Classifier

References

[1] Tsoumakas G, Katakis I, Vlahavas I. Mining multi-label data. In: Data mining and knowledge discovery handbook. Springer, (2010).

DOI: 10.1007/978-0-387-09823-4_34

[2] C. Bielza, G Li, and P Larranaga. Multi-dimensional classification with bayesian networks. International Journal of Approximate Reasoning, 52(2011), 705-727.

DOI: 10.1016/j.ijar.2011.01.007

[3] Jollife I., 1986. Principal Component Analysis, vol. 1, Springer Series in Statistics. Springer, Berlin.

[4] Battiti, R., 2002. Using Mutual Information for Selecting Features in Super-vised Neural Net Learning. IEEE Trans. Neural Networks 5, 537-550.

DOI: 10.1109/72.298224

[5] Peng. H., Long, F., Ding, C. 2005. Feature selection based on mutual information: Criteria of max-dependency, max-relevance, and min-redundancy. IEEE Trans. Pattern Anal. Machine Intell. 27(8), 1226-1238.

DOI: 10.1109/tpami.2005.159

[6] He, X., Cai, D., Niyogi, P., 2006. Laplacian score for feature selection. Adv. Neural Inform. Process. Systems 18.

[7] Ji-min Ye, Ting Huang. New Fast-ICA Algorithmsfor Blind Source Separation Without Prewhitening [J]. Communications in Computer and Information Science, 2011, 255(2): 579-585.

DOI: 10.1007/978-3-642-23220-6_73

[8] Pearl J. Probabilistic reasoning in intelligent systems: networks of plausible inference. Morgan Kaufmann Publishers; (1988).

[9] Koller D, Friedman N. Probabilistic graphical models principles and techniques. Cambridge: MIT Press; (2009).

[10] L. van der Gaag, P.R. de Waal, Multi-dimensional Bayesian network classifiers, in: Third European Conference on Probabilistic Graphical Models , 2006, p.107–114.