时序与回归预测方法比较研究  

作　　者：薛裕颖 段希义 潘玉民[1] 

机构地区：[1]华北科技学院电子信息工程学院,北京101601

出　　处：《黑龙江科技信息》2015年第19期129-131,共3页Heilongjiang Science and Technology Information

摘　　要：基于神经网络的时序和回归预测方法,对同一数据建立BP、RBF神经网络及小波神经网络(WNN)模型,分别采用两种方法预测,比较实验结果,指出各自的优点和存在的问题。仿真实验表明回归预测法高于时序预测精度,但回归预测法要求一定规模的数据且规律性好,它需提供各影响因子数据才能进行并行递推预测;时序预测一般精度较差,但数据采集方便,其可行性和实用性好,适用于数据信息少的环境等特点。A method based on neural network for sequence and regression forecasting,to establish BP , RBF neural network and wavelet neural network （WNN） model for the same data, using two methods to predict respectively, comparing the results, pointed out their advantages and problems.Simulation results show that the regression prediction method is higher than the temporal prediction accu-racy,but regression prediction method requires a certain amount of data and good regularity, it must provide each impact factor data and then to parallel recursive prediction; Use time-series forecasting to predict accuracy, data acquisition is convenient, good feasibility and practicability,and applicable to less data information etc.

关 键 词：时序预测 回归预测 神经网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]