Denying Evolution Resampling: An Improved Method for Feature Selection on Imbalanced Data

Abstract

Imbalanced data classification is an important problem in the field of computer science. Traditional classification algorithms often experience a decrease in accuracy when the data distribution is uneven. Therefore, measures need to be taken to improve the balance of the dataset and enhance the classification accuracy of the model. We have designed a data resampling method to improve the accuracy of classification detection. This method relies on the negative selection process to constrain the data evolution process. By combining the CRITIC method with regression coefficients, we establish crossover selection probabilities for elite genes to achieve an evolutionary resampling process. Based on independent weights, the feature analysis improves by 3%. We evaluated the resampled results on publicly available datasets using traditional logistic regression with cross-validation. Compared to the other resampling models, the F1 score performance of the logistic regression five-fold cross-validation is more stable than the other methods using the two sampling results of the proposed method. The effectiveness of the proposed method is verified based on F1 score evaluation results.