Data Imputation Under Similarity Rule Constraints Using Fuzzy Multi-Objective Programming

Missing or incomplete data poses a significant challenge during data collection for forecasting, estimation, and decision-making purposes. Given the profound impact of data quality on the performance of machine learning algorithms, data imputation plays a crucial role in many applications. Considering potential dependencies between data attributes enhances the reliability of the imputation process. In this paper, we address this by incorporating fuzzy relaxation in the differential dependencies (DDs) among attributes and propose a novel fuzzy multi-objective linear (FMOL) model to achieve optimal imputation performance. The proposed model aims to maximize the imputation rate while minimizing violations of crisp DDs. We employ the Improved Zimmermann Method to solve the FMOL model effectively. Experimental results on the Kaggle dataset demonstrate that our proposed approach outperforms existing methods in terms of imputed fields and imputation accuracy.