Binary Differential Evolution Based on Individual Entropy for Feature Subset Optimization

The high dimensionality of data brings great challenges to the classification accuracy and complexity of the algorithm. Feature selection technology can improve the classification performance of the algorithm effectively. In this paper, a novel binary differential evolution based on individual entropy (BDIE) is proposed. First, the individual entropy method is constructed to quantify the diversity of the population, and the relationship between population diversity and convergence is analyzed. Then, the objective function based on individual entropy is designed to evaluate the feature subset. A new binary mutation strategy is proposed, and it can effectively search the global optimal solution. In order to validate the BDIE, the datasets with different sizes and the classifiers of different types are used for testing. In addition, the well-known algorithms are introduced for comparison. The experimental results show that the proposed algorithm can effectively improve the classification performance and reduce the time cost without increasing the size of the feature subset.