DataRobot PartnersUnify all of your data, ETL and AI tools in our open platform with our Technology Partners, extend your cloud investments with our Cloud Partners, and connect with DataRobot Services Partners to help you build, deploy or migrate to the DataRobot AI Platform.
This tutorial shows how to integrate Tableau extensions with DataRobot. By the end, you will be able to:
Generate automated insights using the DataRobot Insights extension.
Run sensitivity analysis using the DataRobot What-if extension.
Requirements
You must have a full version of Tableau (2018.2 and later) to use these extensions. You cannot use a Tableau Public account to access the DataRobot extensions. See Tableau help for using extensions.
For Tableau Server and Tableau Online, you need to ensure that DataRobot’s app and prediction server URLs are whitelisted:
If you are using an on-premise version of DataRobot:
Tableau expects all extension communication to use HTTPS. You need to make sure that DataRobot is set up with SSL and ideally with a certificate from a trusted Certificate Authority; otherwise, additional steps are required.
Update the DataRobot application server URL in the Tableau Extension:
Open the DataRobot_Insights.trex file with a text editor.
Edit the URL of the DataRobot application server for your organization.
DataRobot Insights Extension
Open a new Tableau book and add a new data source. You can upload a local Excel file as well.
Create a new worksheet and try to visualize some interactions between the columns. For this example, we use AVG(Target) in columns and the Purpose of the loan in rows to see how the outcome is dependent on different categories.
Create a new dashboard using the worksheet created above.
Establish a connection between DataRobot and Tableau by providing the API Token.
Click the account profile link to go to your DataRobot settings.
From the Account Profile page, select Developer Tools.
Copy the API token.
Paste the API token in the Tableau extension (1) and click Login (2).
Select the worksheet, set the target column (Find insights related to), and click Start to launch automated insights.
DataRobot will analyze all the features, and find associations between these features and the target column. This process could take from a few seconds to a few minutes depending on the volume of data.
After analysis
Once the computation is finished, you will see the correlations between the features and the outcome in an interactive manner.
These insights can be incredibly important to get a sense of what the data looks like before starting to build models. It can give an early indication of whether or not there’s any any predictive signal in the data.
In situations where there too many columns, these insights could be used for reducing the dimensionality of the data by eliminating columns that do not have any association with the target.
Important:Tableau restricts this extension to operate on 10,000 rows. If you think that more data is required for meaningful insights, then it is recommended that you connect DataRobot directly to the source data.
DataRobot What-if Extension
Note: The What-If extension requires a DataRobot Prediction Server and works for binary classification and regression models; time series and multiclass are currently not supported.
Open a new workbook and create a new dashboard. This time there is no need to add data sources because you will access the DataRobot model deployment directly.
tableau 15
Drag the Extension button to the dashboard again.
tableau 16
This time, upload the DataRobot_What-if.trex file when asked for the extensions file.
tableau 17
Connect to DataRobot by providing the API key (as you did for the DataRobot Insights extension).
Paste the API token in the Tableau extension (1) and click Login (2).
tableau 18
Once the connection is made, you will see all the deployments associated with your DataRobot account. Choose the deployment on which you want to run the scenarios (1) and click Start Simulation.
In this area of the page you can create scenarios by entering values for different variables that the model uses for computing predictions. After entering the values, click Updateprediction. This will display the predicted probability along with Prediction Explanations.
Click + Add to comparison to add this scenario to the comparison list.
tableau 19
Create more scenarios by tweaking the values of variables of interest and click Update prediction and + Add to comparison for each of these scenarios.
On the Compare tab (next to the Simulate tab), you now see a side-by-side comparison of these different scenarios.
tableau 21
The consumers of these predictions can modify attributes and analyze how the outcome changes. In addition to providing a glimpse into the future, the what-if analysis also leads to faster decisions.
How to Enable Debug Mode on Tableau
If you have issues connecting Tableau to DataRobot, it can be very useful to enable Tableau’s debug mode, which allows you to gather more information so that you can narrow down the issue.
The below steps describe how to enable Tableau’s Debug Mode for Tableau 2019.1 or greater. If you are using Tableau 2018.2 or 2018.3, refer to Tableau’s trex debugger for the Chromium browse (available for download).
For Tableau 2019.1 or greater:
Start Tableau Desktop and enable debugging: from a command prompt, start Tableau using –remote-debugging-port=8696
For example: C:\Program Files\Tableau\Tableau 2019.3\bin>tableau.exe –remote-debugging-port=8696
Open the Tableau Dashboard that includes the DataRobot Extension.
Open Chrome browser and browse to http://localhost:8696.
The image below will appear, listing the pages to inspect.
Select the extension to inspect; in this example, it is the DataRobot What-If extension.
Complete the fields in the Tableau dashboard and select the dashboard button that calls the DataRobot integration. (For the What-If extension, this is the Update Predictions button.)
Switch back into the Chrome inspect page to view the network requests from Tableau Desktop and the associated status. Note, this will be in the Network tab on the page.
The two images below show success cases:
The first image provides the details around the call (the request URL, response headers, and status) and the second image provides a preview of the response message.
The next two images show an error condition and the resulting message back from DataRobot:
The first image is showing a status code of 422 while the second image from the Preview tab displays the message back. In this case, the Tableau Extension is not properly handling the formatting of currency data types.
Looking for More Information on Tableau Extensions?