Improving classification performance for heterogeneous cancer gene expression data

In our previous work, we proposed the "impact factors" (IFs) to measure the symmetric errors in different microarray experiments, and integrated the IFs to the Golub and Slonim (GS) and k-nearest neighbors (kNN) classifiers. In this paper, we perform experiments with different cancer types, which are lung adenocarcinomas and prostate cancer, to further validate the efficiency and effectiveness of the IFs integrations in terms of measurements of classification accuracy, sensitivity and specificity. For both cancer types, the IFs integrations can be successfully improved on the classification performance.