A Study on Intentional-Value-Substitution Training for Regression with Incomplete Information

Abstract

This paper focuses on a method to train a regression model from incomplete input values. It is assumed in this paper that there are no missing values in a training data set while missing values exist during a prediction phase using the trained model. Under this assumption, Intentional-Value-Substitution (IVS) training is proposed to obtain a machine learning model that makes the prediction error as minimum as possible. Through a mathematical analysis, it is shown that there are some meaningful substitution values in the IVS training for the model. It is shown through a series of computational experiments that the substitution values estimated by the extended mathematical analysis help the models predict outputs for inputs with missing values even though there is more than one missing value.