Feature Extraction and Dimensionality Reduction of Cancer Data Using Folded LDA

Linear Discriminant Analysis is a less commonly applied dimensionality reduction technique in cancer data classification. This could be due to the inability of LDA to achieve good classification results when applied on small training data-a common characteristics of cancer data. F-LDA is an extension of LDA and was recently proposed in another application to overcome the challenge posed by the lack of enough samples for training. This paper therefore evaluates the effectiveness of F-LDA as a dimensionality reduction technique in cancer data classification. Experimental results obtained are promising and demonstrate the ability of F-LDA to effectively reduce the dimensionality of cancer data in small training sample scenarios.