Krill Herd Optimization algorithm for cancer feature selection and random forest technique for classification

The Cancer Feature Selection and classification problem is one of the prevalent tasks in computational molecular biology. Detecting a gene or list of genes which cause cancer can be acknowledged using the feature selection and classification which leads to giving a faultless treatment for patient and drug discovery of the particular gene. The feature selection and classification of cancer using microarray gene expression data is a computationally difficult task. Even now, the computation of gene selection and classification is a challenging area to provide an exact biological related gene that causes cancer. In this work, three methods have been proposed. One is the Fish Swarm Optimization algorithm along with both Support Vector Machine and Random Forest technique for cancer feature selection and classification. But the above methods have reduced very few features from the datasets. Thus, they are considered as an existing method for this work. Now, the second proposed method namely an enhanced Krill Herd Optimization (KHO) technique was employed for selecting the genes and Random Forest (RF) Technique was employed to classify the cancer types. The Random Forest classification has been used because of its accurate classification accuracy. First, the subset of features is selected using KHO and the Random Forest classification is applied to the selected features. Ten different gene microarray cancer datasets were used to evaluate the efficiency of the proposed. The proposed KHO/RF method is compared with other well-known existing methods like PSO/SVM, PSO/RF, FSO/SVM and FSO/RF. As an outcome, the proposed method outperforms the other existing methods with 100% accuracy of results for most datasets.