Affinity Propagation Clustering Algorithm Based on PCA

Abstract:

Overlap information usually exits in the high-dimensional data. Misclassified points may be more when affinity propagation clustering is applied to these data. Concerning this problem, a new method combining principal components analysis and affinity propagation clustering is proposed. In this method, dimensionality of the original data is reduced on the premise of reserving most information of the variables. Then, affinity propagation clustering is implemented in the low-dimensional space. Thus, because the redundant information is deleted, the classification is accurate. Experiment is done by using this new method, the results of the experiment explain that this method is effective.

• Periodical: Applied Mechanics and Materials (Volume 590)
• Pages: 688-692
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.590.688
• Online since: June 2014
• Authors: Bei Chen*, Kun Song
• Keywords: Affinity Propagation Clustering, Dimensionality Reduction, PCA, PCA-AP

References

[1] Jain A K, Murty M N, Flynn P J: ACM Computing Surveys, Vol. 31(1999), pp.264-323.

[2] Frey B J,D. Dueck: Science, Vol. 315(2007), pp.972-976.

[3] He Xiaoqun, Multivariate statistical analysis, Second ed., China Renmin University Press, Beijing, (2008).

[4] Dunia R, Qin S J: AIChE Journal, Vol. 44(1998), pp.1813-1831.

[5] Wang Kaijun, Zhang Junying, Li Dan, Zhang Xinna, Guo Tao: ACTA AUTOMATICA SINICA, Vol. 33(2007), pp.1242-1246.

[6] Wang Kaijun, Li Jian, Zhang Junying, Tu Chongyang: Computer Engineering, Vol. 33(2007), pp.197-201.

[7] Howley T, Madden M G: Knowledge-Based Systems. Vol. 19(2006), pp.363-370.

[8] Kaufman L, Rousseeuw P J, Finding groups in Data: An introduction to Cluster Analysis, John Wiley and Sons, Malden, (1990).

[9] Dudoit S, Fridlyand J: Genome Biology, Vol. 3(2002), pp.1-21.