Ensemble Selection for Cancer Diagnosis

Microarrays are known for their wide use in providing expression profiles for thousands of genes. Gene expression profiles provide a rich information for cancer diagnosis. Selecting an efficient classifier is a challenging task due to the presence of several classifier types. Previous studies showed that ensembles of classifiers are more efficient than single classifiers in cancer samples classification. However, designing an efficient ensemble has faced a number of challenges such as the large space of ensembles, increasing the diversity between the ensemble members, and the use of an efficient method to combine the decisions of the ensemble members. In this book, a novel ensemble selection algorithm is proposed. The proposed algorithm addresses the main challenges of the ensemble selection problem taking into consideration the special nature of microarray datasets. A set of experiments has been performed to study the robustness of ensembles of classifiers. This study shows that ensembles of classifiers are more robust than single classifiers. The study also shows that the proposed algorithm performs betten than other ensemble selection algorithms in the literature.