时间序列神经网络预测方法
TIME SERIES NEURAL NETWORK FORE- CASTING METHODS
-
摘要: 本文从信息论的角度出发,讨论了利用神经网络理论构造时间序列预测模型的可能性和关键问题,并在此基础上提出3种时间序列神经网络预测方法,它们是:神经网络非线性时间序列模型、神经网络多维时间序列模型和神经网络组合预测模型,将上述模型应用于实例的结果表明,在非线性信息的处理能力和预测精度方面都有很大提高。进一步,对今后智能信息预测方法的发展方向进行了探讨,提出了智能信息预测系统的结构模型。
-
关键词:
- 信息论; 信息处理; 神经网络预测方法
Abstract: This paper discussed the possibility and key problem of constructing the neural network time series model, and three time series neural network forecasting methods has been proposed. That is, the neural network nonlinear time series model, the neural network multi-dimension time model and the neural network com bining predictive model. These three methods are applied 10 real prcblems. The results show that these methods are better than the traditional ones. Furthermore, the comparison of the neural network with the traditional methods and the constructed model of intellectual information forecasting system are given. -
翁文波.预测论基础.北京:石油工业出版社,1984,第二章.[2]文新辉,陈开周.西安电子科技大学学报,1994,21(1): 73-78.[3]文新辉,陈开周.预测,1993,12(6): 48-51.[4]文新辉,牛明洁.预测,1992,26(4): 58-61.[5]文新辉,陈开周.神经网络在经济管理中的应用之二:神经网络广义组合预测模型,中国神经网络1993年学术大会,西安:1993,1034-1042.[6]Bates J M, Granger C W J. Operations Research Quarterly, 1969, 20(2): 319-324.
计量
- 文章访问数: 2862
- HTML全文浏览量: 210
- PDF下载量: 470
- 被引次数: 0