Tutorial: Predict automobile price with the visual interface
In this tutorial, you take an extended look at developing a predictive solution in the Azure Machine Learning service visual interface. By the end of this tutorial, you'll have a solution that can predict the price of any car based on technical specifications you send it.
This tutorial continues from the quickstart and is part one of a two-part tutorial series. However, you don't have to complete the quickstart before starting.
In part one of the tutorial series you learn how to:
- Import and clean data (the same steps as the quickstart)
- Train a machine learning model
- Score and evaluate a model
In part two of the tutorial series, you'll learn how to deploy your predictive model as an Azure web service.
A completed version of this tutorial is available as a sample experiment. From the Experiments page, go to Add New > Sample 1 - Regression: Automobile Price Prediction(Basic)
Create a workspace
If you have an Azure Machine Learning service workspace, skip to the next section. Otherwise, create one now.
Sign in to the Azure portal by using the credentials for the Azure subscription you use.
In the upper-left corner of the portal, select Create a resource.
In the search bar, enter Machine Learning. Select the Machine Learning service workspace search result.
In the ML service workspace pane, select Create to begin.
In the ML service workspace pane, configure your workspace.
Field Description Workspace name Enter a unique name that identifies your workspace. In this example, we use docs-ws. Names must be unique across the resource group. Use a name that's easy to recall and differentiate from workspaces created by others. Subscription Select the Azure subscription that you want to use. Resource group Use an existing resource group in your subscription, or enter a name to create a new resource group. A resource group holds related resources for an Azure solution. In this example, we use docs-aml. Location Select the location closest to your users and the data resources. This location is where the workspace is created.
To start the creation process, select Review + Create.
Review your workspace configuration. If it is correct, select Create. It can take a few moments to create the workspace.
To check on the status of the deployment, select the Notifications icon, bell, on the toolbar.
When the process is finished, a deployment success message appears. It's also present in the notifications section. To view the new workspace, select Go to resource.
Open the visual interface webpage
Open your workspace in the Azure portal.
In your workspace, select Visual interface. Then select Launch visual interface.
The interface webpage opens in a new browser page.
Import and clean your data
The first thing you need is clean data. If you completed the quickstart, you can reuse your data prep experiment here. If you haven't completed the quickstart, skip the next section and start from a new experiment.
Reuse the quickstart experiment
Open your quickstart experiment.
Select Save As at the bottom of the window.
Give it a new name in the pop-up dialog that appears.
The experiment should now look something like this:
If you successfully reused your quickstart experiment, skip the next section to begin training your model.
Start from a new experiment
If you didn't complete the quickstart, follow these steps to quickly create a new experiment that imports and cleans the automobile data set.
Create a new experiment by selecting +New at the bottom of the visual interface window.
Select Experiments > Blank Experiment.
Select the default experiment name "Experimented Created on ..." at the top of the canvas and rename it to something meaningful. For example, Automobile price prediction. The name doesn't need to be unique.
To the left of the experiment canvas is a palette of datasets and modules. To find modules, use the search box at the top of the module palette. Type automobile in the search box to find the dataset labeled Automobile price data (Raw). Drag this dataset to the experiment canvas.
Now that you have your data, you can add a module that removes the normalized-losses column completely. Then, add another module that removes any row that has missing data.
Type select columns in the search box to find the Select Columns in Dataset module. Then drag it to the experiment canvas. This module allows you to select which columns of data you want to include or exclude in the model.
Connect the output port of the Automobile price data (Raw) dataset to the input port of the Select Columns in Dataset.
Select the Select Columns in Dataset module and select Launch column selector in the Properties pane.
On the left, select With rules
Next to Begin With, select All columns. These rules direct Select Columns in Dataset to pass through all the columns (except those columns we're about to exclude).
From the drop-downs, select Exclude and column names, and then type normalized-losses inside the text box.
Select the OK button to close the column selector (on the lower right).
Now the properties pane for Select Columns in Dataset indicates that it will pass through all columns from the dataset except normalized-losses.
Add a comment to the Select Columns in Dataset module by double-clicking the module and entering "Exclude normalized losses.". This can help you see, at a glance, what the module is doing in your experiment.
Type Clean in the Search box to find the Clean Missing Data module. Drag the Clean Missing Data module to the experiment canvas and connect it to the Select Columns in Dataset module.
In the Properties pane, select Remove entire row under Cleaning mode. These options direct Clean Missing Data to clean the data by removing rows that have any missing values. Double-click the module and type the comment "Remove missing value rows."
Train the model
Now that the data is ready, you can construct a predictive model. You'll use your data to train the model. Then you'll test the model to see how closely it's able to predict prices.
Classification and regression are two types of supervised machine learning algorithms. Classification predicts an answer from a defined set of categories, such as a color (red, blue, or green). Regression is used to predict a number.
Because you want to predict price, which is a number, you can use a regression algorithm. For this example, you'll use a linear regression model.
Train the model by giving it a set of data that includes the price. The model scans the data and looks for correlations between a car's features and its price. Then test the model by giving it a set of features for automobiles it's familiar with and see how close the model comes to predicting the known price.
Use your data for both training the model and testing it by splitting the data into separate training and testing datasets.
Type split data in the search box to find the Split Data module and connect it to the left port of the Clean Missing Data module.
Select the Split Data module you just connected to select it. In the Properties pane, set the Fraction of rows in the first output dataset to 0.7. This way, we'll use 70 percent of the data to train the model, and hold back 30 percent for testing.
Double-click the Split Data and type the comment "Split the dataset into training set(0.7) and test set(0.3)"
To select the learning algorithm, clear your module palette search box.
Expand the Machine Learning then expand Initialize Model. This displays several categories of modules that can be used to initialize machine learning algorithms.
For this experiment, select Regression > Linear Regression and drag it to the experiment canvas.
Find and drag the Train Model module to the experiment canvas. Connect the output of the Linear Regression module to the left input of the Train Model module, and connect the training data output (left port) of the Split Data module to the right input of the Train Model module.
Select the Train Model module. In the Properties pane, Select Launch column selector and then type price next to Include column names. Price is the value that your model is going to predict
Now the experiment should look like.
Run the training experiment
An experiment runs on a compute target, a compute resource that is attached to your workspace. Once you create a compute target, you can reuse it for future runs.
Select Run at the bottom to run the experiment.
When the Setup Compute Targets dialog appears, if your workspace already has a compute resource, you can select it now. Otherwise, select Create new.
The visual interface can only run experiments on Machine Learning Compute targets. Other compute targets will not be shown.
Provide a name for the compute resource.
The compute resource will now be created. View the status in the top-right corner of the experiment.
It takes approximately 5 minutes to create a compute resource. After the resource is created, you can reuse it and skip this wait time for future runs.
The compute resource will autoscale to 0 nodes when it is idle to save cost. When you use it again after a delay, you may again experience approximately 5 minutes of wait time while it scales back up.
Score and evaluate the model
Now that you've trained the model using 70 percent of your data, you can use it to score the other 30 percent of the data to see how well your model functions.
Type score model in the search box to find the Score Model module and drag the module to the experiment canvas. Connect the output of the Train Model module to the left input port of Score Model. Connect the test data output (right port) of the Split Data module to the right input port of Score Model.
Type evaluate in the search box to find the Evaluate Model and drag the module to the experiment canvas. Connect the output of the Score Model module to the left input of Evaluate Model. The final experiment should look something like this:
Run the experiment using the same compute target used previously.
View the output from the Score Model module by selecting the output port of Score Model and select Visualize. The output shows the predicted values for price and the known values from the test data.
To view the output from the Evaluate Model module, select the output port, and then select Visualize.
The following statistics are shown for your model:
- Mean Absolute Error (MAE): The average of absolute errors (an error is the difference between the predicted value and the actual value).
- Root Mean Squared Error (RMSE): The square root of the average of squared errors of predictions made on the test dataset.
- Relative Absolute Error: The average of absolute errors relative to the absolute difference between actual values and the average of all actual values.
- Relative Squared Error: The average of squared errors relative to the squared difference between the actual values and the average of all actual values.
- Coefficient of Determination: Also known as the R squared value, this is a statistical metric indicating how well a model fits the data.
For each of the error statistics, smaller is better. A smaller value indicates that the predictions more closely match the actual values. For Coefficient of Determination, the closer its value is to one (1.0), the better the predictions.
Manage experiments in Azure Machine Learning service workspace
The experiments you create in the visual interface can be managed from the Azure Machine Learning service workspace. Use the workspace to see more detailed information such as individuals experiment runs, diagnostic logs, execution graphs, and more.
Open your workspace in the Azure portal.
In your workspace, select Experiments. Then select the experiment you created.
On this page, you'll see an overview of the experiment and its latest runs.
Select a run number to see more details about a specific execution.
The run report is updated in real time. If you used an Execute Python Script module in your experiment, you can specify script logs to output in the Logs tab.
Clean up resources
You can use the resources that you created as prerequisites for other Azure Machine Learning service tutorials and how-to articles.
If you don't plan to use anything that you created, delete the entire resource group so you don't incur any charges:
In the Azure portal, select Resource groups on the left side of the window.
In the list, select the resource group that you created.
On the right side of the window, select the ellipsis button (...).
Select Delete resource group.
Deleting the resource group also deletes all resources that you created in the visual interface.
Delete only the compute target
The compute target that you created here automatically autoscales to zero nodes when it's not being used. This is to minimize charges. If you want to delete the compute target, take these steps:
In the Azure portal, open your workspace.
In the Compute section of your workspace, select the resource.
Delete individual assets
In the visual interface where you created your experiment, delete individual assets by selecting them and then selecting the Delete button.
In part one of this tutorial, you completed these steps:
- Reuse the experiment created in the Quickstart
- Prepare the data
- Train the model
- Score and evaluate the model
In part two, you'll learn how to deploy your model as an Azure web service.