Why MissForest Fails in Prediction Tasks: A Key Limitation You Need to Keep in Mind

The popular MissForest algorithm contains a critical flaw for predictive modeling because its standard implementation does not save the imputation model. This prevents applying the exact same imputation rules from the training set to the test set, a crucial step for unbiased model evaluation. Common workarounds either introduce data leakage by combining datasets or create inconsistent results by re-learning imputation on the smaller test set. To avoid these issues and ensure valid performance, imputation models must be trained solely on the training data and then applied without modification to any new data.