Tutorial: Categorizing customers using k-means clustering with SQL Server Machine Learning Services

APPLIES TO: yesSQL Server noAzure SQL Database noAzure Synapse Analytics (SQL DW) noParallel Data Warehouse

In this four-part tutorial series, you'll use Python to develop and deploy a K-Means clustering model in SQL Server Machine Learning Services to cluster customer data.

In part one of this series, you'll set up the prerequisites for the tutorial and then restore a sample dataset to a SQL database. Later in this series, you'll use this data to train and deploy a clustering model in Python with SQL Server Machine Learning Services.

In parts two and three of this series, you'll develop some Python scripts in an Azure Data Studio notebook to analyze and prepare your data and train a machine learning model. Then, in part four, you'll run those Python scripts inside a SQL database using stored procedures.

Clustering can be explained as organizing data into groups where members of a group are similar in some way. For this tutorial series, imagine you own a retail business. You'll use the K-Means algorithm to perform the clustering of customers in a dataset of product purchases and returns. By clustering customers, you can focus your marketing efforts more effectively by targeting specific groups. K-Means clustering is an unsupervised learning algorithm that looks for patterns in data based on similarities.

In this article, you'll learn how to:

  • Restore a sample database into a SQL Server instance

In part two, you'll learn how to prepare the data from a SQL database to perform clustering.

In part three, you'll learn how to create and train a K-Means clustering model in Python.

In part four, you'll learn how to create a stored procedure in a SQL database that can perform clustering in Python based on new data.


Restore the sample database

The sample dataset used in this tutorial has been saved to a .bak database backup file for you to download and use. This dataset is derived from the tpcx-bb dataset provided by the Transaction Processing Performance Council (TPC).

  1. Download the file tpcxbb_1gb.bak.

  2. Follow the directions in Restore a database from a backup file in Azure Data Studio, using these details:

    • Import from the tpcxbb_1gb.bak file you downloaded
    • Name the target database "tpcxbb_1gb"
  3. You can verify that the dataset exists after you have restored the database by querying the dbo.customer table:

    USE tpcxbb_1gb;
    SELECT * FROM [dbo].[customer];

Clean up resources

If you're not going to continue with this tutorial, delete the tpcxbb_1gb database from SQL Server.

Next steps

In part one of this tutorial series, you completed these steps:

  • Restore a sample database into a SQL Server instance

To prepare the data for the machine learning model, follow part two of this tutorial series: