Effect of Missing Data Types and Imputation Methods on Supervised Classifiers: An Evaluation Study

Data completeness is one of the most common challenges that hinder performance data analytics platforms. Different studies have assessed effect missing values on different classification models based a single evaluation metric, namely, accuracy. However, accuracy its own misleading measure classifier because it does not consider unbalanced datasets. This paper presents an experimental study assesses incomplete datasets five models. The analysis was conducted with ratios in six vary size, type, and balance. Moreover, for unbiased analysis, classifiers measured using three metrics, Matthews correlation coefficient (MCC), F1-score, results show sensitivity supervised to differs according set factors. significant factor pattern ratio, followed by imputation method, then balance dataset. when are due Missing Completely At Random (MCAR) less than their Not (MNAR) pattern. Furthermore, MCC as better reflects variation data.