【摘要】 针对复杂时间序列全局预测模型建模效率低、预测性能不佳等问题,提出一种基于局部RBF神经网络的新型预测模型.该模型采用K最近邻搜索方法得到待预测样本的K个近邻,用近邻样本进行RBF神经网络建模,用训练好的RBF神经网络对待预测样本进行预测.实验结果显示该模型在复杂时间序列预测上有良好的性能.更多还原

【Abstract】 A novel complex time series predictor based on local RBF network is proposed to overcome the shortages of the global predictor such as low modeling efficiency,low prediction accuracy and so on.This predictor takes the K nearest neighbor searching method to gain the K neighbors of the sample to be predicted,trains a RBF network with these neighbors,and predicts the sample with well trained RBF network.Experimental results show that the proposed predictor perform very well in the task of predicting complex time series.更多还原