Frequency Count And Joint Effect Based Gene Selection

Feature selection is the hotspot of the research in the information field,especially in Pattern Recognition. Gene selection is a application of the feature selection and has a very important role as a predictor for diseases.As is known to everyone,diseases can often result from some genes combined together. There are some approaches which can find the disease genes effectively, but can't find the combined genes.To resolve the above-mentioned problem, this article introduced a new method of gene selection---Frequency Count and Joint Effect based Gene Selection. Following is the main idea. First, combine the genes ralated closely to be some branches by using Fisher, all of which make up a branch matrix. Then, select the combined genes which has the highest Frequency Count and the biggest Joint Effect value from the branch matrix, that is, the most essential disease genes.This method has its special advance, that is, it is more suitable to the disease caused by combined genes.This method could also be used to solve other problems of feature selection. By compared with other methods, such as, mRMR[1], SPCA[2], Ruck[3], Tarr[4] and so on, this method is more effective then others in finding the essential feature and reducing the feature space dimension.