PCA for Leukemia Classification
The real cause of leukemia has not been found, and traditional method to recognize the malignant cells has some limitations, and that is very time-consuming. Because microarray gene expression data is few samples, high-dimensional and nonlinear, it brings us “dimensionality disaster”, so data dimensionality reduction becomes a problem we pay attention, and SVM (Support Vector Machine) overcomes the “dimensionality disaster” to a certain extent by means of optimization method, for the decision function of SVM is only decided by part of the support vectors. This paper combines SVM with Laplacian Eigenmaps and PCA (Principal Component Analysis) respectively for Leukemia data classification, compare the results, PCA gets better result.
Long Chen, Yongkang Zhang, Aixing Feng, Zhenying Xu, Boquan Li and Han Shen
J. Li and G. R. Weng, "PCA for Leukemia Classification", Applied Mechanics and Materials, Vol. 43, pp. 744-747, 2011