基于数据变换的GM(1,1)误差校正方法

提高预测方法的预测效果具有重要意义,但是仅靠建立单一的预测模型来提高预测精度是非常困难的.本文对当前预测方法存在的不足进行了阐述,在此基础上提出将误差校正方法引入预测以提高预测精度的新思路.首先,采用预测方法(文中以T-S模糊神经网络方法为例)对训练样本进行拟合,再对预测对象进行初始预测;其次,引入加速平移变换和加权均值变换对误差序列进行处理,再以处理后的数据为样本构建基于数据变换的GM(1,1)误差预测模型,并对该序列后续点进行预测;最后,利用误差预测结果对初始预测值进行校正.文章最后以上证综合指数的收盘价的预测为例,算例分析表明,与校正前的预测精度相比,校正后的预测精度有显著提高,进而验证了该模型的有效可行.

Research on error correction method based on data transformational GM(1,1)

It is of great significance to improve the prediction effect of forecasting method, however experience shows that it is very difficult for improving the accuracy of forecasting by setting up single forecasting model. This article describes the deficiencies of the present forecasting methods and puts forward a new approach for the improvement of prediction accuracy by introducing error correction. First, the fuzzy neural network forecasting model is established for a preliminary prediction by using the training sample data. Second, the data transformation is introduced to process the error sequence. On the basis of the processed data, the GM(1,1) model is constructed and is used to predict the subsequent error. Third, the correction of preliminary prediction values is calibrated. The numerical example based on the historical data of the Shanghai composite index shows that the presented approach improves the accuracy of forecasting significantly compared with the prediction accuracy before correction, and then the validity of the model is verified.