An adjustable fuzzy classification algorithm using an improved multi-objective genetic strategy based on decomposition for imbalance dataset

摘要

In this paper, we propose an adjustable fuzzy classification algorithm using multi-objective genetic strategy based on decomposition (AFC_MOGD) to solve imbalance classification problem. In AFC_MOGD, firstly, an improved multi-objective genetic strategy based on decomposition is adopted as the basic optimization algorithm in which a new updating pattern getting good solutions is designed. Then, an adjustable parameter which is ranged in the interval [0, 1] is used to adjust complexity of each classifier artificially. Finally, a normalized method which takes class percentage into account to determine class label and rule weight of each rule is introduced so as to obtain more reasonable rules. The proposed algorithm is compared with three typical algorithms on eleven imbalance datasets in terms of area under the ROC of convex hull. The Wilcoxon signed-rank test is also carried out to show that our algorithm is superior to other algorithms.