摘要
为解决时间序列的一步预测问题,提出了一种基于混沌算子的预测网络。混沌算子具有复杂的动力学行为,根据各算子所处的不同状态,利用加权方法计算出时间序列下一时刻的预测值。根据预测值与实际值的误差,利用混沌优化方法动态地调节混沌算子的参数,逐渐提高网络的预测精度。利用该方法分别对混沌以及实际股票价格等复杂时间序列进行了仿真预测。仿真结果表明,该方法可以对具有内在确定性的系统进行有效的预测。
A novel method of time series prediction based on chaotic operators was proposed. Prediction network is composed of chaotic operators. Chaotic operators have complex dynamic behaviors. When the parameters are changed, chaotic operators can keep different states respectively. Prediction result can be obtained by weighted average method. According to the error between prediction value and factual value, the parameters of chaotic operators are adjusted by chaos optimization algorithm to raise the precision. The method can predict lots of complex time series, such as chaotic time series, stock price time series, etc. Simulation results prove that the method can validly predict time series which have inherent certain property.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第2期507-509,共3页
Journal of System Simulation
基金
国家自然科学基金(10402003)
天津市高等学校科技发展基金(20060613)
关键词
时间序列
预测
混沌算子
网络
time series
prediction
chaotic operator
network