HMVI: Unifying Heterogeneous Attributes with Natural Neighbors for Missing Value Inference

TL;DR

HMVI tackles missing value inference in heterogeneous tabular data by unifying cross-type attribute dependencies within a single framework. It introduces a unified heterogeneous metric, cross-type imputation that leverages incomplete data, and a clustering-with-imputation loop guided by natural neighbors. Empirical results demonstrate superior imputation accuracy and clustering robustness against strong missingness, outperforming baseline methods such as MMS, MF, and KNNMI. The approach offers practical benefits for real-world systems by exploiting both inter-sample and inter-attribute relationships to improve data completeness and downstream analytics.

Abstract

Missing value imputation is a fundamental challenge in machine intelligence, heavily dependent on data completeness. Current imputation methods often handle numerical and categorical attributes independently, overlooking critical interdependencies among heterogeneous features. To address these limitations, we propose a novel imputation approach that explicitly models cross-type feature dependencies within a unified framework. Our method leverages both complete and incomplete instances to ensure accurate and consistent imputation in tabular data. Extensive experimental results demonstrate that the proposed approach achieves superior performance over existing techniques and significantly enhances downstream machine learning tasks, providing a robust solution for real-world systems with missing data.

Figures (3)

• Figure 1: Mechanism of the proposed HMVI.
• Figure 2: Evaluation of imputation accuracy for heterogeneous data (Lower value indicates better imputation performance).
• Figure 3: Evaluation of mRMSE for different ablation variants of HMVI on heterogeneous datasets.