A New Maximum-Relevance Criterion for Significant Gene Selection
Gene (feature) selection has been an active research area in microarray analysis. Max-Relevance is one of the criteria which has been broadly used to find features largely correlated to the target class. However, most approximation methods for Max-Relevance do not consider joint effect of features on the target class. We propose a new Max-Relevance criterion which combines the collective impact of the most expressive features in Emerging Patterns (EPs) and some popular independent criteria such as t-test and symmetrical uncertainty. The main benefit of this criterion is that by capturing the joint effect of features using EPs algorithm, it finds the most discriminative features in a broader scope. Experiment results clearly demonstrate that our feature sets improve the class prediction comparing to other feature selections.
KeywordsSupport Vector Machine Feature Selection Gene Selection Feature Subset Target Class
- 2.Das, S.: Filters: Wrappers and a Boosting-Based Hybrid for Feature Selection. In: Proc. 18th Intl. Conf. Mach. Learn., pp. 74–81 (2001)Google Scholar
- 4.Chung, H.Y., Liu, H., Brown, S., McMunn-Coffran, C., Kao, C.Y., frank Hsu, D.: Identifying Significant Genes from Microarray Data. In: Proc. of the fourth IEEE symp. on BIBE, vol. 358 (2004)Google Scholar
- 8.Dong, G., Li, J.: Efficient Mining of Emerging Patterns: Discovering Trends and Differences. In: Proc. of the fifth ACM SIGKDD Inter. Conf. on Know. Disc. and Data min., pp. 43–52 (1999)Google Scholar
- 11.Fayyad, U., Irani, K.: Multi-interval discretization of continuous-valued attributes for classification learning. In: Proc. of the 13th Inter. J. Conf. on Arti. Intel., pp. 1022–1029 (1993)Google Scholar
- 13.Yu, L., Liu, H.: Redundancy Based Feature Selection for Microarray Data. In: KDD 2004, pp. 22–25 (2004)Google Scholar