基于历史误差模型和局域临域校正的预测融合算法

针对预测模型的精度随机性较大,一些数据存在残缺和不确定性,需要融合多模型的组合体来减少预测误差的问题,提出了使用改进模糊RBF神经网络(FRNN)的误差模型的“局域临域校正”的组合体预测融合方法:通过预处理的FRNN误差模型来表征每个预测模型的实时性能,在预测时,利用误差模型,使用基于权值和偏好的融合作为预测模型组合体的聚合方法,将多个预测模型在线融合并给出最终预测结果.该方法将较大的权重赋予局部性能较优的预测模型,同时消除预测偏好的影响,并在预测前预处理所需误差模型.应用结果表明,构建的预测模型其预测精度相比单一预测模型有明显优势,且具有良好的在线预测实时性,具有推广和应用的价值.

New algorithm for predicting ensemble aggregation fusion based on fuzzy error model and local neighbor revision

Prediction models have randomness and sample data have uncertainty and deformity.Ensemble method can reduce the prediction error caused by randomness and uncertainty.This article proposes a method,which uses a revised fuzzy RBF neuron network(FRNN) algorithm to produce a local neighborhood revision fusion of ensemble;an FRNN error-model is pre-processed to represent the runtime performance for each prediction model.In prediction,with the output of error model the ensemble aggregation method is used to guide...