Page 34 - MSDN Magazine, July 2017

MACHINE LEARNING
Doing Data Science and
AI with SQL Server
Wee Hyong Tok
Data is an important asset for every business. Whether you’re in the retail, health care, telecommunications, utilities or financial businesses, you’re familiar with the following two use cases:
• In an online transaction processing (OLTP) scenario, trans- actional data is stored in a database. The transaction data is produced by various line-of-business (LOB) applications.
• In a data warehousing scenario, data is acquired from various heterogeneous data sources, transformed and cleansed, and loaded into data warehouses. The consolidated data provides the single source of truth for business reporting and dash- boards. At the same time, it also enables interactive analysis via multi-dimensional online analytical processing (OLAP) cubes, and tabular models.
Getting from raw data to insights empowers business decision makers to gain a deeper understanding into each aspect of the business and helps them react to new business situations quickly. For example, consider a retail scenario. The business analyst notices that sales are dropping for specific retail stores. The busi- ness analyst wants to drill down to understand the details on what’s causing the drop in sales. By being able to run the analysis (aggregating, joining of data from multiple data sources, filtering
and so on) on a large amount of data, it enables deep analysis of customer behavior and trends in the retail stores. Microsoft SQL Server powers these mission-critical applications.
Many companies have started on digital transformation to mod- ernize their data platform to keep pace with the ever-growing requirements on the type of data that needs to be stored and the volume in which the data is being acquired.
As part of this digital transformation, advanced analytics plays an important role. Specifically, companies have been either build- ing up data science teams within their companies or leveraging external resources to do data science. They use data science to distill data assets into nuggets of gold that can help them proac- tively deliver personalized customer experiences (personalized Web sites, product recommendations, customer lifetime value and so on), reduce downtime for equipment (predicting remaining useful lifetime) and more. The potential use of data science and how it can literally change businesses is exciting.
Some common use cases (non-exhaustive) of data science include the following:
Identifying Loan Credit Risk: A lending institution (a credit bureau) might want to leverage loan credit risk models to determine the borrowers that are likely to default and reduce the number of loans given to these high-risk borrowers.
Managing Customer Churn: Customer churn models have been used extensively (in retail and by telecommunication provid- ers). For example, customers leveraging mobile services offered by telecommunication providers have a rich variety of choices and can easily switch between service providers. Managing customer churn is important to reduce customer acquisition costs and main- tain a high-quality service. In addition, retail companies are using
This article discusses:
• Serving AI with data
• Built-in AI capabilities of SQL Server
• How to develop R/Python AI stored procedures Technologies discussed:
SQL Server, R and Python Tools for Visual Studio, Data Science, Artificial Intelligence, R and Python
30 msdn magazine