基于局部线性重构的近邻填充算法

Nearest Neighbor Filling Algorithm Based on Local Linear Reconstruction

K近邻填充算法(KNNI)对缺失数据填充时,使用K个近邻的训练样本属性值的平均值作为KNNI算法的填充值,该值偏离真实值较大,一种有效的方法是K近邻样本属性值的加权平均值作为填充值。如何确定K近邻样本的权值,是KNNI算法研究热点。为此,提出引入局部线性重构方法用于近邻填充,该方法是用K近邻样本重构测试样本得出权值。实验结果表明提出的算法比KNNI算法填充效果更好,填充值更接近真实值。

K nearest neighbor filling algorithm for filling missing data, the average value of training sample attributes uses K neighbor values as the filling value of KNNI algorithm, the value is far from the real values, an effective method is to sample attribute weighted K nearest neighbor average as filling value. How to determine the weights of K nearest neighbor samples is a hot research topic of KNNI algorithm. For this reason, proposes a method of local linear reconstruction for nearest neighbor filling. The method uses the K nearest neighbor sample to reconstruct the test sample. The experimental results show that the proposed algorithm is better than the KNNI algorithm, and the filling value is close to the real value.