DataRobot Connector Tools For Alteryx

DataRobot Connector Tools For Alteryx

By: J. Kyle Roberts, Ph.D
February 19, 2019

 

 

 Fig1

 

As many of you have begun the foray from predictive tools into machine learning, you may have noticed DataRobot recently introduced two  connector tools for Alteryx Designer:  1.  DataRobot Auto-model Tool   2. DataRobot Predictor Tool. 

I wanted to give a quick overview of these tools and their capabilities.

 

DataRobot Automodel Tool:

Like most Alteryx tools, the DataRobot Auto-model tool is relatively straightforward and easy to use. Once a dataset has been cleaned and is ready for analysis inside DataRobot, running the model is as simple as dragging a connector from the data to the tool and completing a couple of setup tasks. In this example, we will utilize the DataRobot web-hosting service (although this could be deployed internally on a local server). We then enter our DataRobot Token (obtained under the user settings inside DataRobot), give the project a name (this will be saved to our DataRobot dashboard), and select the target variable, as can be seen below.

 

Fig2 

 

The DataRobot Auto-model tool has two more options inside the Alteryx interface; of which we selected both. The first is “Use Quick Autopilot”. The difference between the “Quick Autopilot” and normal “Autopilot” is that the “Quick” version runs on a restricted number of models and does not include most of the blended models.  Selecting either of these options immediately keys up DataRobot and runs the Autopilot (default settings) on the target variable and all other features contained in the dataset that are passed forward. The second check box toggles whether or not DataRobot opens the project in a new web browser.

 

DataRobot Predict Tool: 

After the model is run, the DataRobot Predict Tool can be deployed to score new data. Once deployed, the user can select a saved project from all saved projects in DataRobot and then select a model by which the data will be scored. In other words, you do not need to run the “DataRobot Autopilot” tool inside of Alteryx to run the “DataRobot Predict” tool. A nice feature in the “Model Pulldown” is that Alteryx pulls models in the same rank cross-validation order as is displayed in the DataRobot model leaderboard.

 

 

Revised Fig3-1 

 

 Once the data is scored, the results can then be joined back to the original dataset inside Alteryx.
 

Fig4 

 

Key Take-Aways:

There are a few key take-aways from my interaction with the DataRobot tools for Alteryx.

  • Alteryx now provides a seamless way to load data into a DataRobot session without having to trace down the current version of a dataset, thus eliminating the probability of modeling the wrong data.
  • The DataRobot Predict Tool is quick and powerful and eliminates the need to batch processes inside DataRobot and then import data back into Alteryx.
  • It should be noted that these tools are for people who have a DataRobot license.  Alteryx users will require a DataRobot license to use these tools. 
  • These tools do not eliminate the need for a user to fine-tune and deploy the right model with the right feature list.  You will still need to do those functions inside DataRobot as Alteryx will not aid in making these decisions for you.

In short, the new DataRobot Connector tools contribute greatly to productivity for data scientists. The tools also eliminate a potential source of error by removing steps necessary to load data into and out of DataRobot. In short, these are great tools for users of DataRobot and Alteryx.

To read the official DataRobot announcement click here.