Classification in the presence of missing data

Missing data is quite common when dealing with real world datasets. There are several ways to improve prediction accuracy when missing data in some predictors without completely discarding the entire observation. This example shows how decision trees with surrogate splits can be used to improve prediction accuracy in the presence of missing data. Bagging (bootstrap aggregating), is an ensemble approach which involves training several weak learners to create a strong classifier. Decreasing value with number of trees indicates good performance.