A feature subset selection algorithm based on feature activity and improved GA

Feature subset selection is an important research branch in the field of pattern recognition. Due to the traditional feature selection algorithms do not take into account the feature updating case, the paper analyzes the relationship between dataset and features, proposes a new feature activity measurement that is used to determine the influence among different features on some certain conditions. Based on the feature activity measurement, to cope with the premature convergence and the weak local search ability of classic genetic algorithm, the paper proposes a feature set selection algorithm based on adaptive feature activity and improved genetic algorithm. The proposed algorithm can dynamic guidance feature selection process, and then accelerate from multidimensional characteristics in the collection to find the optimal feature subset. Experimental results indicate the proposed method can obtain small scale feature set on the basis of higher classification accuracy and faster running time than those compared algorithms. The proposed algorithm can be better applied to the field of feature selection application.