MVIAnalyzer: A Holistic Approach to Analyze Missing Value Imputation

TL;DR

This paper introduces MVIAnalyzer, a comprehensive framework for analyzing missing value imputation methods within the broader context of data analysis, supported by software and visualization tools.

Contribution

It presents a novel, holistic framework for evaluating missing value imputation methods, including simulation, analysis, and visualization, with an open-source implementation.

Findings

Demonstrates the application of MVIAnalyzer on diverse datasets

Highlights the strengths and limitations of various MVI methods

Provides insights into the impact of different parameters on imputation quality

Abstract

Missing values often limit the usage of data analysis or cause falsification of results. Therefore, methods of missing value imputation (MVI) are of great significance. However, in general, there is no universal, fair MVI method for different tasks. This work thus places MVI in the overall context of data analysis. For this purpose, we present the MVIAnalyzer, a generic framework for a holistic analysis of MVI. It considers the overall process up to the application and analysis of machine learning methods. The associated software is provided and can be used by other researchers for their own analyses. To this end, it further includes a missing value simulation with consideration of relevant parameters. The application of the MVIAnalyzer is demonstrated on data with different characteristics. An evaluation of the results shows the possibilities and limitations of different MVI methods. Since MVI is a very complex topic with different influencing variables, this paper additionally illustrates how the analysis can be supported by visualizations.