Parallel Fuzzy Reasoning Models with Ensemble Learning

This paper proposes a new learning algorithm and a parallel model for fuzzy reasoning systems. The proposed learning algorithm, which is based on an ensemble learning algorithm AdaBoost, sequentially trains a series of weak learners, each of which is a fuzzy reasoning system. In the algorithm, each weak learner is trained with the learning data set that contains more data misclassified by the previous weak one than the others. The output of the ensemble system is a majority vote weighted by weak learner accuracy. Further, the parallel model is proposed in order to enhance the ensemble effect. The model is made up of more than one ensemble system, each of which consists of weak learners. In order to show the effectiveness of the proposed methods, numerical simulations are performed. The simulation result shows that the proposed parallel model with fuzzy reasoning systems constructed by AdaBoost is superior in terms of accuracy among all the methods.