摘要
采用自适应前馈网络算法(AFN)进行非线性时序预测,对网络结构设计进行详细的探讨,并应用该方法对经典非线性时间序列数据进行预测,与传统预测方法(TAR)比较,结果证明此种方法具有较好的效果,网络的结构得到了简化,不仅满足了误差目标的要求,而且提高了网络的推广能力。且AFN方法可以对时间序列数据间的关系给出一种基于贡献率的解释。
Adaptive feed-forward network algorithm is used to mack non-linear prediction, and the design of network topology is discussed. An example of classical non-linear prediction is given, and compared with TAR method. The results prove the new method is better. The topology of neural network has been simplified. Not only the error goal is satisfied, but also the generalization capability is improved. AFN method can also explain the relation in the time series data based on contribution ratio.
出处
《辽宁工程技术大学学报(自然科学版)》
CAS
北大核心
2004年第1期138-140,共3页
Journal of Liaoning Technical University (Natural Science)
关键词
非线性时间序列
预测
自适应前馈网络算法
门限自回归模型
人工神经网络
adaptive feed-forward network
non-linear prediction
time series
threshold autoregressive model
artificial neural network