Twin and Multiple Black Holes Algorithm for Feature Selection

Over the years, nature-inspired meta-heuristic algorithms have been proposed to solve complex computational problems. Evolutionary algorithms are simple to implement and can be adapted to any kind of optimization algorithms because of their flexibility. Finding best subset of attribute in order to collect interpretable results is an important step in classification and regression tasks. Several stochastic and heuristic algorithms with a suitable classifier have been employed in the past for attribute selection with considerable success. In this work we have proposed a novel modification to the recently proposed Black hole (BH) meta-heuristic swarm optimization algorithm and employed it along with Support Vector Machines (SVM) for simultaneous attribute selection and classification. This attribute selection algorithm can be implemented in different domains. Experiments demonstration on publicly available benchmark datasets shows the algorithm compares favorably with existing algorithms.