Transfer Function Based Feature Selection Using Harris Hawk Optimization for Biomedical Applications

Feature Selection plays key role to truncate the high dimensionality in the datasets through eradicating the unwanted and noisy attributes to improve the performance of classification. The Meta-heuristic technique is one of the most optimistic search techniques to travel the feature space to locate finest feature subset. The Harris Hawk Optimization (HHO) excited from hunting approach of hawk in nature. The cost effectiveness and simplicity formulates the algorithm prominent and superior in solving optimization problems. The binary version of the optimization is accountable for delineating continuous searching space to discrete feature space. Here, 6 transfer functions with S-Shaped and V-Shaped families were interposed and analyzed. Twelve benchmark datasets from UCI repository has been introduced to examine the transfer functions and choose the best feature pertaining to mitigate local minima and high speed of convergence. To analyze the efficacy of transfer function, the proposed method was compared over exiting algorithms and the proposed results attain higher classification accuracy in many of the datasets.