GCM data analysis using dimensionality reduction

Abstract

Extracting useful information from gene microarray becomes a research in great demand, but because the microarray sample size is small, high-dimensional, nonlinear, traditional statistical learning methods face a challenge of "dimensionality disaster" and "problem of small sample size", therefore, dimensionality reduction becomes a key to pattern recognition. This article uses Principal Component Analysis (PCA) and Local Tangent Space Alignment (LTSA) to reduce the dimensionality of the Global Cancer Map data, and then utilizes Support Vector Machine to classify the data, PCA getting the better result. © 2012 Springer-Verlag GmbH.