Modified Binary Ant Colony Optimization for Drift Compensation

In data analytics and pattern recognition, feature selection is a critical task to provide a subset of features with minimum redundancy. This reduces the computation time as well as cost. In this manuscript, a correlation based feature selection approach based on a modified binary ant colony optimization algorithm (MBACO) is proposed. Combined with random forest regression, the proposed MBACO algorithm is customized for a drift compensation application. In this application, the ant road map is initialized to avoid the local optimum. The proposed method is compared with that of binary particle swarm optimization on a well-known UCI dataset. Experimental results show that the proposed method exhibits better performance over the binary particle swarm optimization based approach.