Integrating the R model for fast batch prediction

Before continuing to the next steps, please follow the Prerequisite - Prepare the data section.

Step 1 – Train and save a real-time scoring model using T-SQL

In this step, you can create a predictive model for real-time scoring and native scoring, and optionally for the AUC, into a table via a stored procedure. The goal is to build a model that is reusable. You can skip this step if there is an existing compatible model that was created and stored in a table in SQL Server already.

The following stored procedure uses rxSerializeModel, which lets you serialize an R model in raw format. This then allows you to save the model in the VARBINARY format, which can be loaded into SQL Server for real-time scoring. To reverse the serialization for use in R, you could use rxUnserializeModel:

CREATE PROCEDURE [dbo].[uspTrainTipPredictionModelWithRealTimeScoring] 
AS 
BEGIN 
   DECLARE @auc FLOAT; 
   DECLARE @model VARBINARY(MAX); 
 
   -- The data to be...