Imputation Analysis of Central Tendencies for Classification

In real-world datasets missing values are so common. Most Machine Learning algorithms won't work with missing values, and so they should be handled before training the model. It is a common practice to impute the missing values with central tendencies (Mean, Median, Mode), but choosing a particular one among them is not an easy choice to make. This paper analyzes the impact of using each central tendency for different distributions of data. Skewness and the presence of outliers are considered for selecting the data for analysis. Certain presumptions have been made before the examination, and performance metrics such as accuracy, AUC-ROC, precision, recall, and F1 score are analyzed to prove/disprove the assumptions.