Gene Pattern Extraction Based On PICA And Classification Using Boosting

Gene chip is one new technique for obtaining large scale gene expression data. One can get and analysis the expression levels of thousands of genes and the relation with diseases in the organism. But, because of the Partial Volume Effect of the organism, directly processing of the gene Microarray data will make that the characters extracted are not those of the biology genes, which will reduce the measurement sensitivity of the gene features. Therefore, instead of the traditional biomedicine methods, the Partial Independent Component Analysis (PICA) is proposed in this paper to do the Partial Volume Correction for getting the pure Microarray data. The method is easy in operation and reduces the biomedicine experiment cost as well. At the same time, considering of the character of the gene Microarray data, a new machine learning algorithm, Boosting is introduced. Based on the detailed analysis of the boosting algorithm, a new proof of the convergence of its is given. Then, considering of the non-robustness of the algorithm, a robust Boosting learning algorithm is proposed in which the modification of the combination rules for many weak classifiers is made. Some open testing datasets and real gene Microarray data are applied in experiments. The results have verified the feasibility and validity of the PICA to get the gene Microarray data and the advanced Boosting Algorithm for gene classification.