Integrated Analysis of Missing-Value Imputation Techniques and Class-Imbalance Handling for CKD Prediction

Missing values and class imbalances are major factors that significantly degrade the performance of classification models in medical-data analysis.This comprehensive study systematically analyzes the specific effects of combining various advanced missing-value imputation techniques and traditional class-imbalance handling methods on overall classification performance.To achieve this, we utilized a real-world chronic kidney disease dataset obtained from the UCI Machine Learning Repository.The empirical results demonstrated that iterative imputation achieved a noticeably higher classification performance than simple mean imputation or k-nearest neighbor-based imputation, successfully attaining an impressive F1 score of 0.9980 and an area under the curve of 1.0000.In contrast, the subsequent application of oversampling techniques, such as the synthetic minority oversampling technique, provided an extremely limited effect on performance improvement.These robust findings strongly suggest that the initial missing-value handling strategy may exert a substantially greater impact on model performance than the chosen resampling technique.