An evolutionary correlation-aware feature selection method for classification problems

As global search techniques, population-based optimization algorithms have provided promising results in feature selection (FS) problems. However, their major challenge is high time complexity associated with the exploration of a large space and consequently number fitness function evaluations. Moreover, interaction between features another key issue FS problems, directly affecting classification performance through selecting correlated features. In this paper, an estimation distribution algorithm (EDA)-based method proposed three important contributions. Firstly, as extension EDA, each iteration generates only two individuals competing based on function, evolving during using our update procedure. Secondly, we provide guiding technique to determine be selected for iteration. result, final solution would optimized evolution process. These lead increasing convergence speed algorithm. Thirdly, main contribution addition considering importance alone, can consider features, being able deal complementary increase performance. To do this, conditional probability scheme that considers joint The introduced probabilities successfully detect Experimental synthetic dataset proved approach facing these types Furthermore, 13 real-world datasets obtained from UCI repository showed superiority comparison some state-of-the-art approaches. evaluate effectiveness subset, support vector machines are used classifier. efficiency analysis experimental non-parametric statistical tests had significant advantages other