An complementarity based feature selection method for pattern recognition

For the pattern recognition problem, this paper proposes a feature selection method based on complementarity analysis. Analyse the separability of single feature, search the feature combination with the smallest probability of mixing region and in the mixed region with the greatest separability to reduce the probability of classification error. Compared with other feature selection algorithms, data testing result shows that the feature selection method based on complementarity analysis has a lower error recognition rate than other methods, which has verified the superiority and the advanced nature of the method.