论文部分内容阅读
分析人工神经网络预报中的误差来源。推导人工神经网络预报过程中预报误差和训练样本质量之间的关系；讨论训练样本质量对用于时间序列预报人工种经网络性能的影响；并从统计的观声、引入用于评价训练样本质量的数字指标“一致度’（DCT）；还随新指标给出一些模拟结果和相应的建议，以便在人工种经网络训练中准确地选择训练样本。
Analysis of sources of error in artificial neural network forecast. The relationship between the prediction error and the quality of training samples in the process of artificial neural network prediction is deduced; the influence of training sample quality on the performance of the artificial seed network via time series prediction is discussed; and the quality of training samples is introduced from the view of statistics (DCT). Some new simulation results and corresponding suggestions are also given along with the new index, so that training samples can be accurately selected in artificial training through network.