基于动态窗口的灰色加权填充算法及应用

Grey Weighted Imputation Method Based on Dynamic Window and Its Application

为解决复杂系统中单属性缺失数据填充困难问题,提出了基于动态窗口的灰色加权填充算法。该算法通过建立双向灰色预测模型,并采用模型精度评价系数加权填充缺失数据,有效增强了算法的准确性;提出基于灰色模型评价系数反馈的动态伸缩窗口概念,寻找产生最优模型的训练数据序列,使算法具有良好的鲁棒性和适应性。实验结果表明,该方法在RMSE和MA两项指标上均优于传统的双向灰插值、灰插值、多项式插值等方法。

In order to solve the difficulty of single attribute missing values imputation in complex systems, a grey weightedimputation algorithm based on dynamic window is proposed. The algorithm imput missing values with two-way grey predictionmodel and model accuracy evaluation coefficient weighting method, which enhanced the algorithm precision effectively. It introducesa dynamic flexible window concept based on grey model evaluation coefficient feedback for finding the optimal model, which makes the algorithm be robust and adaptable. Numerical experiments results show that comparing several existingmethod, the algorithm can achieve better performance on root-mean-square error and mean accuracy.