Missing value estimation by local minor components

摘要

In this paper, we propose a method of estimating missing values by local minor components. We assume that a data point, some attributes of which are missing, is on a hyper plane. The normal of the hyper plane is a local minor component vector. Thus, the data point is projected onto the hyper plane and its missing attributes are estimated. We extract the local minor components by executing fuzzy clustering and neural principal component analysis (PCA) approach simultaneously. The extracted local minor components take the substructures of the data set into consideration. Missing values are estimated simply and easily.