Optimal feature selection using binary teaching learning based optimization algorithm

Feature selection is a significant task in the workflow of predictive modeling for data analysis. Recent advanced feature selection methods are using the power of optimization algorithms for choosing a subset of relevant features to get better classification results. Most of the optimization algorithms like genetic algorithm use many controlling parameters which need to be tuned for better performance. Tuning these parameter values is a challenging task for the feature selection process. In this paper, we have developed a new wrapper-based feature selection method called binary teaching learning based optimization (FSBTLBO) algorithm which needs only common controlling parameters like population size, and a number of generations to obtain a subset of optimal features from the dataset. We have used different classifiers as an objective function to compute the fitness of individuals for evaluating the efficiency of the proposed system. The results have proven that FS-BTLBO produces higher accuracy with a minimal number of features on Wisconsin diagnosis breast cancer (WDBC) data set to classify malignant and benign tumors. (c) 2018 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Automated summary

Feature selection is an important task in predictive modeling for data analysis. Advanced methods that utilize optimization algorithms have been developed to choose relevant features and improve classification results. In this paper, a new wrapper-based feature selection method called FSBTLBO is proposed, which requires only a few control parameters to obtain an optimal subset of features. The results demonstrate that FS-BTLBO achieves higher accuracy with a minimal number of features for classifying tumors.