Simple rule-based ensemble classifiers for cancer DNA microarray data classification

Abstract

DNA microarray, which is one of the most important molecular biology technologies in post-genomic era, has been widely applied in medical field, especially for cancer classification. However, it is difficult to acquire excellent classification accuracy by using traditional classification approaches due to microarray datasets are extremely asymmetric in dimensionality. In recent years, ensemble classifiers which may obtain better classification accuracy and robustness have attracted more interests in this field but it is more time-consuming. Therefore, this paper proposed a novel ensemble classification method named as SREC(Simple Rule-based Ensemble Classifiers). Firstly, the classification contribution of each gene is evaluated by a novel strategy and the corresponding classification rule is extracted. Then we rank all genes to select some important ones. At last, the rules of the selected genes are assembled by weighted-voting to make decision for testing samples. It has been demonstrated the proposed method may improve classification accuracy with lower time-complexity than traditional classification methods.