useR! 2024

Missing data poses a pervasive issue in statistical analysis across various domains. Ignoring missing values or using incongruous imputation methods can introduce bias and decrease the validity of statistical results. To overcome the challenge of missing data imputation, we propose the use of novel machine learning algorithms: The R package VIM (Visualization and Imputation of Missing Values) has incorporated machine learning (ML)-based imputation methods, including xgboost and transformer models. This presentation will elucidate the recent advancements in VIM, with a special emphasis on the performance of these ML models in handling missing data, comparing them to more conventional imputation methods, and highlight their advantages and disadvantages. Through real-world examples, we aim to demonstrate the effectiveness of our models in improving accuracy and reliability.