Enhancing Boosting by Feature Non-Replacement for Microarray Data Analysis

We have investigated strategies for enhancing ensemble learning algorithms for DNA microarray data analysis. By using modified versions of AdaBoost, LogitBoost and BagBoosting we have shown that feature non-replacement provides an effective enhancement to the performance of all three algorithms, and overall, BagBoosting with feature non-replacement had the lowest error rates when used on six commonly-used cancer datasets.