title: "Python tutorial: Predict NYC taxi fares with binary classification"

description: In this five-part tutorial series, you'll learn how to embed Python code in SQL Server stored procedures and T-SQL functions with SQL machine learning to predict NYC taxi fares using binary classification.

ms.technology: machine-learning

ms.date: 07/28/2020

monikerRange: ">=sql-server-2017||>=sql-server-linux-ver15||>=azuresqldb-mi-current||=sqlallproducts-allversions"

::: moniker range=">=sql-server-2017||>=sql-server-linux-ver15||=sqlallproducts-allversions"

In this five-part tutorial series for SQL programmers, you'll learn about Python integration in [SQL Server Machine Learning Services](../sql-server-machine-learning-services.md) or on [Big Data Clusters](../../big-data-cluster/machine-learning-services.md).

::: moniker-end

::: moniker range=">=azuresqldb-mi-current||=sqlallproducts-allversions"

In this five-part tutorial series for SQL programmers, you'll learn about Python integration in [Machine Learning Services in Azure SQL Managed Instance (preview)](/azure/azure-sql/managed-instance/machine-learning-services-overview).

::: moniker-end

You'll build and deploy a Python-based machine learning solution using a sample database on SQL Server. You'll use T-SQL, Azure Data Studio or SQL Server Management Studio, and a database instance with SQL machine learning and Python language support.

This tutorial series introduces you to Python functions used in a data modeling workflow. Parts include data exploration, building and training a binary classification model, and model deployment. You'll use sample data from the New York City Taxi and Limousine Commission, and the model you'll build predicts whether a trip is likely to result in a tip based on the time of day, distance traveled, and pick-up location.

In this article, you'll:

In [part two](sqldev-py3-explore-and-visualize-the-data.md), you'll explore the sample data and generate some plots.

In [part three](sqldev-py4-create-data-features-using-t-sql.md), you'll learn how to create features from raw data by using a Transact-SQL function. You'll then call that function from a stored procedure to create a table that contains the feature values.

::: moniker range=">=sql-server-2017||>=sql-server-linux-ver15||=sqlallproducts-allversions"

+ Install [SQL Server Machine Learning Services with Python](../install/sql-machine-learning-services-windows-install.md#verify-installation)

::: moniker-end

+ Restore the [NYC Taxi demo database](demo-data-nyctaxi-in-sql.md)

This tutorial series assumes familiarity with basic database operations such as creating databases and tables, importing data, and writing SQL queries. It does not assume you know Python and all Python code is provided.

The process of building a machine learning solution is a complex one that can involve multiple tools, and the coordination of subject matter experts across several phases:

+ obtaining and cleaning data

+ exploring the data and building features useful for modeling

+ training and tuning the model

+ deployment to production

Development and testing of the actual code is best performed using a dedicated development environment. However, after the script is fully tested, you can easily deploy it to [!INCLUDE[ssNoVersion](../../includes/ssnoversion-md.md)] using [!INCLUDE[tsql](../../includes/tsql-md.md)] stored procedures in the familiar environment of Azure Data Studio or [!INCLUDE[ssManStudio](../../includes/ssmanstudio-md.md)]. Wrapping external code in stored procedures is the primary mechanism for operationalizing code in SQL Server.

Whether you are a SQL programmer new to Python, or a Python developer new to SQL, this five-part tutorial series introduces a typical workflow for conducting in-database analytics with Python and SQL Server.

After the model has been saved to the database, you can call the model for predictions from [!INCLUDE[tsql](../../includes/tsql-md.md)] by using stored procedures.

In this article, you:

title: "Python tutorial: Explore and visualize data"

description: In part two of this five-part tutorial series, you'll explore sample data and generate some plots in preparation for using binary classification in Python with SQL machine learning.

ms.technology: machine-learning

ms.date: 07/29/2020

monikerRange: ">=sql-server-2017||>=sql-server-linux-ver15||>=azuresqldb-mi-current||=sqlallproducts-allversions"

In part two of this five-part tutorial series, you'll explore the sample data and generate some plots. Later, you'll learn how to serialize graphics objects in Python, and then deserialize those objects and make plots.

In this article, you'll:

In [part one](sqldev-in-database-python-for-sql-developers.md), you installed the prerequisites and restored the sample database.

In [part three](sqldev-py4-create-data-features-using-t-sql.md), you'll learn how to create features from raw data by using a Transact-SQL function. You'll then call that function from a stored procedure to create a table that contains the feature values.

In [part four](sqldev-py5-train-and-save-a-model-using-t-sql.md), you'll load the modules and call the necessary functions to create and train the model using a SQL Server stored procedure.

In [part five](sqldev-py6-operationalize-the-model.md), you'll learn how to operationalize the models that you trained and saved in part four.

To do this, run the following Python code, replacing the server name, database name, and credentials as appropriate (for Windows authentication, replace the `UID` and `PWD` parameters with `Trusted_Connection=True`). Make sure the Python version is the same on the client and the server. Also make sure that the Python libraries on your client (such as matplotlib) are the same or higher version relative to the libraries installed on the server.

5. If the connection is successful, you should see a message like the following:

*The plots are saved in directory: xxxx*

6. The output file is created in the Python working directory. To view the plot, locate the Python working directory, and open the file. The following image shows a plot saved on the client computer.


> [!div class="checklist"]

> + Reviewed the sample data

> + Created plots using Python in T-SQL

> [!div class="nextstepaction"]

> [Create data features using T-SQL](sqldev-py5-train-and-save-a-model-using-t-sql.md)