FREMONT, CA: Microsoft’s recently launched SQL Server 2016, features new services that enable advanced, in-database analytics with R Services, called the SQL Server R Services. The SQL Server R services combine the power of SQL Server and Microsoft R Server (or Open Source R).
With SQL Server R Services, users can develop analytic models in a local R integrated development environment (IDE) while data resides in SQL Server, and computation happens on SQL Server (by setting the compute context to SQL Server).
Once the model is ready for production, it can be operationalized via SQL stored procedures (where R code is encapsulated inside), which can be run within SQL Server Management Studio or called by outside applications to make predictions.
To help familiarize enterprises on building advanced analytics applications with SQL Server R Services, Microsoft announces the release of three new machine-learning data science templates for its SQL Server R Services: Online Fraud Detection, Customer Churn Prediction and Predictive Maintenance. They are available in the Cortana Analytics Gallery and include R code, sample data and SQL stored procedure code. The templates released are:
Online Fraud Detection Template (SQL Server R Services)
The Online Fraud Detection template helps spot transactions made with stolen payment information or compromised accounts. The data-science template teaches developers how to use data from online purchases to identify fraud using R and SQL for data processing and the R integrated development environment (IDE) for training the system and fine-tuning its scoring parameters.
Customer Churn Prediction Template (SQL Server R Services)
The Customer Churn Prediction template is aimed at helping businesses manage and prevent the loss of customers. The code relies on customer demographics and transaction records to help determine if customers are likely to churn or stay put.
Predictive Maintenance Template (SQL Server R Services)
Using simulated sensor measurements for aircraft engines and current operating conditions, it creates three predictive model types; Regression model, classification model and multiclass classification models.
The template's regression models predict an engine's Remaining Useful Life or Time to Failure, while classification and multiclass classification models predict whether a component is likely to fail and when it will fail, respectively.
"These templates are sample advanced analytics solutions that demonstrate best practices and provide building blocks to help users implement a solution quickly," says Xinwei Xue, senior data scientist manager. "Each template is designed to solve a specific problem, and includes sample data, R code (which uses the highly scalable Microsoft R Server ScaleR APIs) and SQL stored procedure code that extends from data preparation and feature engineering to model training and scoring," she added.