Analysis Of Microarray Data Based On Dimension Reduction

The development of microarray technology is very important in the history of biology. Measuring for thousands of genes synchronously become possible because of the technology. Meanwhile, the amount of microarray data is increased exponentially. Facing such huge data, it is easy to lead from"Data Resources"to"Data Disaster"without effective processing methods. Furthermore, the data is high-dimension and there is lack of such samples. How to choose fewer feature genes with high classification capacity is very difficult. This paper studied the microarray data analysis approach based on the dimension reduction technique on the issue.The paper is divided into two parts: Firstly, linear dimension reduction algorithms include PCA, MDS and FA has been tried on microarray data. Then, the results of dimension reduction were compared with that of SVM classification algorithm. It shows that the classification results based on PCA is superior to the other two linear dimension reductions; secondly, nonlinear dimension reduction algorithms include LLE, ISOMAP and LTSA has been tried on microarray data. The results of dimension reduction were compared with that of SVM classification algorithm. It shows that the classification results based on LLE is superior to the other two nonlinear dimension reductions. Get the better classification performance based on PCA and LLE by comparing the results of liner and nonlinear dimension reduction. To better understand the effect of dimension reduction, this paper also analyzes the gene expression data visually.