A Mutation Enhanced Water Cycle algorithm and Support Vector Machine for Feature Selection

A mutation enhanced water cycle algorithm based support vector machine approach is proposed for feature selection. One of the most difficult issues in data mining and machine learning is feature selection. To increase classification accuracy, the suggested approach finds and eliminates redundant features. The water cycle process is the basis for the WCA model, in which rivers and streams travel from the mountains to the sea. The streams represent acceptable solutions and a selected few superior streams are transformed as rivers and the river with the best performance is transformed into sea. The evaporation and raining are two the main processes of WCA. To improve the search ability of WCA, three different mutation based strategies i.e., Cauchy, Gaussian, Opposition have been applied. These mutation strategies have diverse characteristics, which enhances the exploration and exploitation capabilities of search algorithm. Eight datasets from UCI, Kent Ridge, and the Gene Expression Model Database are employed to evaluate the proposed approach. The results are compared in terms of performance metrics i.e., average accuracy, Sensitivity, Specificity, precision, and Percentage of selected Feature and it is found that for all the databases suggested technique is able to outperform its counterparts.