A hybrid gene selection approach for microarray data classification using cellular learning automata and ant colony optimization

- This paper proposes a three stage scheme for gene selection from microarray data.
- Fisher measure ranking is used to reduce the features and hence the search space.
- Ant colony optimized Cellular Learning Automata is used as the wrapper approach.
- Final gene(s) are selected so that the area under accuracy curve is maximized.
- Evaluations show that the smallest set of genes with maximum accuracy is selected.

This paper proposes an approach for gene selection in microarray data. The proposed approach consists of a primary filter approach using Fisher criterion which reduces the initial genes and hence the search space and time complexity. Then, a wrapper approach which is based on cellular learning automata (CLA) optimized with ant colony method (ACO) is used to find the set of features which improve the classification accuracy. CLA is applied due to its capability to learn and model complicated relationships. The selected features from the last phase are evaluated using ROC curve and the most effective while smallest feature subset is determined. The classifiers which are evaluated in the proposed framework are K-nearest neighbor; support vector machine and naïve Bayes. The proposed approach is evaluated on 4 microarray datasets. The evaluations confirm that the proposed approach can find the smallest subset of genes while approaching the maximum accuracy.