基于加速扩散模型的缺失值插补算法
王圣举,张赞

Missing value imputation algorithm based on accelerated diffusion model
Shengju WANG,Zan ZHANG
表 7 不同缺失率下不同插补方法在2数据集中的均方根误差
Tab.7 Root mean square error of different imputation methods under different missing rates in two datasets
方法RMSEWRRMSEFIR
${p_{{\mathrm{mis}}}}$=10%${p_{{\mathrm{mis}}}}$=30%${p_{{\mathrm{mis}}}}$=50%${p_{{\mathrm{mis}}}}$=10%${p_{{\mathrm{mis}}}}$=30%${p_{{\mathrm{mis}}}}$=50%
平均值0.238±0.0020.238±0.0010.238±0.0010.133±0.0020.136±0.0040.135±0.003
ICE0.188±0.0040.194±0.0020.202±0.0010.036±0.0020.047±0.0050.059±0.003
EM0.188±0.0040.194±0.0020.202±0.0010.039±0.0090.050±0.0040.057±0.002
GAIN0.233±0.0040.232±0.0040.268±0.0030.080±0.0010.098±0.0050.188±0.003
MissForest0.175±0.0030.184±0.0010.195±0.0010.052±0.0020.061±0.0040.070±0.003
MIWAE0.257±0.0020.255±0.0020.256±0.0010.145±0.0030.147±0.0050.147±0.004
TabCSDI0.193±0.0040.198±0.0030.204±0.0030.047±0.0030.054±0.0050.060±0.004
PNDM_Tab0.170±0.0030.178±0.0020.190±0.0020.022±0.0020.032±0.0040.042±0.003