Booster in High Dimensional Data Classification

Abstract

Classification problems specified in high dimensional data with smallnumber of observation are generally becoming common in specific microarray data. In the time of last two periods of years, manyefficient classification standard models and also Feature Selection (FS) algorithm which isalso referred as FS technique have basically been proposed for higher prediction accuracies. Although, the outcome of FS algorithm related to predicting accuracy is going to be unstable over the variations in considered trainingset, in high dimensional data. In this paperwe present a latest evaluation measure Q-statistic that includes the stability of the selected feature subset in inclusion to prediction accuracy. Then we are going to propose the standard Booster of a FS algorithm that boosts the basic value of the preferred Q-statistic of the algorithm applied. Therefore study on synthetic data and 14 microarray data sets shows that Booster boosts not only the value of Q-statistics but also the prediction accuracy of the algorithm applied.