A Hybrid Feature Selection Algorithm For Classification Unbalanced Data Processsing

Abstract

The performance and accuracy of classifier are affected by the result of feature selection directly. Based on the one-class F-Score feature selection and the improved F-Score feature selection and genetic algorithm, combined with machine learning methods like the K nearest neighbor, support vector machine, random forest, naive Bayes, a hybrid feature selection algorithm is proposed to process the two classification unbalanced data problem and multi classification problem. Compared with the traditional machine learning algorithm, it can search in wider feature space and promote classifier to deal with the characteristics of unbalanced data sets according to heuristic rules, which can handle the problem of unbalanced classification better. The experiment results show that the area under receiver operating characteristic curve for two classifications and the accuracy rate for multi classification problem have been improved compared with other models.