Functional data is a type of complex nonlinear structured data,which is often presented and stored in the form of functions(curves).However,in the process of data collection,missing data is inevitable.This paper proposes a Missforest Combin-ing Gaussian Processes(MFGP)method based on cross-sectional and longitudinal information.Inspired by the ensemble models,the method integrates imputation based on Missforest model(MF)with prediction based on Gaussian processes(GP),effectively in-tegrates cross-sectional and longitudinal information of functional data to enhance imputation accuracy.Meanwhile,the results of simulation data interpolation experiment and stock data example analysis show that under the missing ratio of 5%to 55%,MFGP has a significant imputation advantage over seven other imputation methods,namely mean imputation,Hot.deck,SFI,HFI,MICE,MF and GP,and the obtained data is more consistent with the original data.
; 李唯欣1
修改评论