Feature embedding and conditional neural processes for data imputation

Abstract

This Letter addresses the missing value prediction problem in data. Existing missing value prediction studies apply their algorithms after replacing missing values with a constant value, which does not represent the missing values clearly and has disadvantages in processing data with a high missing rate. The authors propose a model that generalises a single data sample from its observations and predicts the value of a query feature. The authors' model can deal with an incomplete data sample independently of missing values. In this process, the feature embedding matrix and feature importance weights are introduced to represent each feature and learned jointly with the proposed model. Through experiments, they verify that the proposed model effectively predicts the missing values of data.