天津工业大学学报2011，Vol.30Issue(4)：85-88,4.
基于迟滞神经网络的商品零售价格指数预测
Prediction of commodity retail price index based on hysteretic neural network
摘要
Abstract
Hysteretic characteristic is brought into neural network, and hysteretic neural network is constructed by changing activation function into hysteretic function. Hysteretic characteristic can enhance the inertia of keeping original state, which can reduce the error rate of state alteration, and enhance the storage and memory ability. Switching characteristic of hysteretic segment response can restrain fault saturation phenomenon during the training. A hysteretic neural network with the structure and training algorithm of feed-forward neural network is proposed to predict the time series. It is applied to predict the commodity retail price index. Prediction results show that the network has good capacity of generalization, and its forecast validity is better than that of conventional neural network.关键词
神经网络/迟滞/泛化能力/记忆能力/商品零售价格指数/预测Key words
neural network/ hysteretic/ capacity of generalization/ memory ability/ commodity retail price index/ prediction分类
信息技术与安全科学引用本文复制引用
赵鑫,修春波..基于迟滞神经网络的商品零售价格指数预测[J].天津工业大学学报,2011,30(4):85-88,4.基金项目
天津市自然科学基金项目(10JCYBJC0004) （10JCYBJC0004）
