摘要
根据数据序列具有宏观变化、微观波动、相近性和相依性,建立了基于BP神经网络与模糊加权马尔可夫链的数据预测模型.首先利用BP神经网络对数据拟合,对残差进行模糊C均值聚类得到马尔可夫链的状态区间,以此求出状态转移概率矩阵,并利用归一化后的自相系数对其进行改进,确定出预测数据所属状态区间,根据状态区间求出预测值.从实例分析表明该文算法具有较高的精确度和可靠性,应用前景广阔.
According to the data sequence with macroscopic changes,microscopic fluctuations,close sex and dependency,data forecast model based on BP neural network and the fuzzy weighted Markov chain was established.The first use of BP neural network in data fitting,the residual fuzzy C mean clustering by Markov chain state interval,in order to calculate the state transition probability matrix, the normalized auto-correlation coefficient was used to improve it,according to the state interval calculated predictive value.The prediction data of state interval were determined,the examples show that the proposed algorithm has high accuracy and reliability,and have wide application prospect.
出处
《甘肃联合大学学报（自然科学版）》
2013年第2期8-13,共6页
Journal of Gansu Lianhe University :Natural Sciences
基金
甘肃省教育厅科研资助项目(00330715-01)
关键词
BP神经网络
马尔可夫链
模糊聚类
自相关系数
BP neural network
Markov chain
fuzzy clustering
autocorrelation coefficient
