【摘要】 针对少数据、贫信息、非线性、动态性的时间序列,采用遗传算法对Elman神经网络的初始权值进行优化以避免陷入局部最小值.建立灰色GM(1,n)模型对其进行预测,使用优化后的神经网络对预测结果进行修正.通过实例拟合、预测,对比灰色GM(1,n)模型、灰色神经网络模型和基于遗传算法的灰色神经网络模型结果,验证预测模型的有效性.结果表明,基于遗传算法的灰色Elman神经网络预测模型能够扩大搜索范围,稳定网络结构,提高解的精度.更多还原

【Abstract】 To predict the time data sequences of less data, poor information, nonlinearity and dynamic, genetic algorithm is used to optimize the initial weights of the Elman neural network to avoid falling into the local minimum. GM(l,n) model is established to predict original data,and then, the optimized neural network is used to correct the prediction. Experimental results show the effectiveness of the predictive model by comparing the gray GM(1, n) model, the gray neural network model and the gray neural network model based on the genetic algorithm. The model established in this paper can expand the search range, stabilize the network structure and improve the precision of the solution.更多还原