关键词: 深度模糊神经网络, 自适应聚类, 灰狼算法, Box-Jenkins燃气炉, 交通流预测

Abstract:

A deep fuzzy neural network (DFNN) is proposed to solve the problem that the deep neural network has poor interpretability and the correction of the model is not targeted when dealing with the big data regression prediction problem. The proposed deep fuzzy neural network adopts an adaptive fuzzy C-means (AFCM)clustering algorithm in structural learning. The structure of the model, namely the number of rules and the antecedent parameters of the rules, is determined by calculating the introduced validity function. The identification of consequent parameters uses an improved grey wolf optimization (IGWO) algorithm. By replacing the linear decreasing strategy in the IGWO with an exponential convergence factor, and using an adaptive position update strategy combined with dynamic weight update, the consequent parameters of the deep fuzzy neural network and the initialization parameters in the adaptive fuzzy mean clustering are optimized. The proposed deep fuzzy neural network and related algorithms are applied to Box-Jenkins gas furnace and short-term traffic flow prediction problems. The experimental results prove the feasibility of the proposed model and algorithm.

Key words: deep fuzzy neural network(DFNN), adaptive clustering, grey wolf algorithm, Box-Jenkins gas furnace, traffic flow prediction