Novel framework for imputation of missing values in databases.

摘要

Many of the industrial and research databases are plagued by missing values problem. One of common ways to cope with this problem is to perform imputation (filling in) of the missing values through variety of statistical and machine learning (ML) procedures. This study concentrates on performing experimental comparison between several algorithms for imputation of missing values, which range from simple statistical algorithms such as mean and hot deck imputation to imputation algorithms that work based on application of inductive ML algorithms. The thesis also proposes a new framework that can be used to improve the accuracy of existing imputation method while maintaining the same asymptotic computational complexity. Extensive experimental test were performed and the results show that a significant improvement of imputation accuracy can be achieved by applying the proposed framework, and that the accuracy of the framework based methods is, on average, the highest among the considered methods.