Deep Learning-Based Approach for Missing Data Imputation

Abstract

The missing values in the datasets are a problem that will decrease the machine learning performance. New methods arerecommended every day to overcome this problem. The methods of statistical, machine learning, evolutionary and deeplearning are among these methods. Although deep learning methods is one of the popular subjects of today, there are limitedstudies in the missing data imputation. Several deep learning techniques have been used to handling missing data, one of themis the autoencoder and its denoising and stacked variants. In this study, the missing value in three different real-world datasetswas estimated by using denoising autoencoder (DAE), k-nearest neighbor (kNN) and multivariate imputation by chainedequations (MICE) methods. The estimation success of the methods was compared according to the root mean square error(RMSE) criterion. It was observed that the DAE method was more successful than other statistical methods in estimating themissing values for large datasets.