A hybrid multi-objective optimization approach with NSGA-II for feature selection

This study introduces a hybrid feature selection technique with a multi-objective algorithm incorporating Information Gain, Random Forest, and Relief F-based approach. We integrate the strengths of filter and wrapper methodologies to enhance the efficacy of addressing feature selection. The information gain, random forest, and relief F-based approach are used to evaluate the significance of features concerning the labels. Subsequently, the information derived from feature scoring is utilized to initialize the population. In addition, the work introduces a new operator for crossover and mutation that uses feature scores to guide these processes. This strategy improves the convergence efficiency and sharpens the search direction of the proposed model within the search space. As part of our empirical research, we compare the suggested model to three different multi-objective feature selection techniques on five different high-dimensional datasets. Our proposed model outperforms state-of-the-art algorithms, as shown by the empirical data. It achieves higher classification accuracy across a range of datasets and exhibits robustness in performance while substantially reducing the feature space. • Introduce a hybrid feature selection method combining filter and wrapper approaches • Use NSGA-II for multi-objective optimization to address big data challenges • Find Pareto-optimal solutions, balancing objectives without major sacrifices • Propose new crossover and mutation operators considering feature weights • Enhance convergence efficiency and guide search to improve algorithm performance