Efficient Gene Expression Analysis by Linking Multiple Data Mining Algorithms

The set of gene micro-arrays, which consists of two leukemia types, was used as a target to evaluate the efficiency of novel integrated data mining classification process. Discovering the most relevant subset of genes among few thousands of analyzed genes is necessary to get accurate disease classification. Dimensional complexity of the classification process was reduced by a filter based on mutual information feature selection coupled with the support vector machines classifier in the leave-one-out loop. The result was an efficient and reliable tool named MIFS/SVM hybrid. Optimal procedure parameters that enable accurate classification and attribute selection could be determined within an acceptable time frame