As seen in the previous section, both the model training step and the prediction step call sp_execute_external_script, which invokes the R process. Real-time scoring and native scoring allow you to do predictions without invoking an R process. Therefore, these scoring methods improve the performance of prediction operations.
In addition, real-time scoring and native scoring let you use a machine learning model without having to install R. As long as you obtain a pretrained model in a compatible format and save it in an SQL Server database, you can call prediction operations easily.
