Ensemble Fuzzy Min-max Neural Network

In this paper, we propose an ensemble fuzzy min-max neural network (EFMNN) for data classification. EFMNN is a five-layer structure neural network based on bagging technology that is a classical method of ensemble learning. The second layer is a preprocessing of the input samples, which is achieved by bootstrap sampling. The third layer is consists of fuzzy min-max neurons (FMNs), while the fourth layer is made up of vote neurons (VNs). FMN is a typical fuzzy min-max neural network, and VN is created based on voting mechanism. VNs are used to vote on the results which are output by FMN. Compared with the traditional fuzzy min-max neural network (FMM), EFMNN has better performance of classification. FMM is very sensitive to the input order of data, and EFMNN can overcome this limitation. The performance of EFMNN is evaluated by several benchmark data sets. The experimental results show that EFMNN has higher classification accuracy and lower sensitivity to the expansion coefficient θ than other classical FMM models.