Microarray gene feature classification based on LS-SVM

DNA microarray has the characteristics of the higher dimension and redundancy, they bring into a series of the difficulties for the gene feature clasiffication. Pertaining to the two classical microarray datasets (cancer of colon set and leukemia set), firstly, the preprocess has been taken by the normalizing and the redundant data have been withdrawn; secondly, Principal Component Analysis method has been adopted to reduce the dimension of datasets and the information gene sets have been obtained; Finally, multiple classifiers have been utilized for the simulating tests, such as LS-SVM, SVM, BP, RBF, etc. They demonstrate that LS-SVM classifier has the higher accuracy for classification and show the approached method can make the correct judgment for classifying the feature of gene dataset, and provide the verifying reliance for clinical therapy further.