• Blog
  • DataRobot Predictor application

DataRobot Predictor application

April 3, 2020
· 4 min read

This post was originally part of the DataRobot Community. Visit now to browse discussions and ask questions about DataRobot, AI Cloud, data science, and more.

This article is a walkthrough of the DataRobot Predictor application. Here we describe how to access it, what it does, and how to derive the most value out of using it.

To find the Predictor application, you can go to the DataRobot header and click on the Applications tab and you’ll immediately be taken to the Applications Gallery where you can find a list of all the applications available to your organization (See Figure 1).

Figure 1 DataRobot Applications Gallery
Figure 1. DataRobot Applications Gallery

The Predictor application helps anybody in your organization make predictions now using a deployed model. Some of the uses and advantages of this application include:

  • Make predictions one data point at a time by providing the required inputs with features that match what the deployed model is expecting.
  • Make a large batch of predictions by importing a file.
  • Compare the resulting prediction result to historical data from your training dataset and use what you learned to judge if the prediction score is high, low, or typical.
  • View the prediction explanations and adjust input values to see how they affect the score.
  • Enable your business stakeholders to experience your deployed model instantly without needing to spend any effort building a UI for the deployed model.

For this walkthrough we are going to use the 10K Diabetes dataset that comes with the Predictor application. This dataset contains a list of diabetes patients with their probability of being readmitted to the hospital after being discharged. Each patient also has other information such as what medications they were taking, age, gender, and so on. The variable we are going to predict is readmitted. You can download that dataset from the Predictor application demo page shown in Figure 2. Once that’s done you can use this data to launch the Predictor application through two approaches:

  • Build and deploy a model using the 10K dataset. In the Deployments tab find the deployed model, click the associated hamburger icon, click on create application, and launch the Predictor application as depicted in Figure 3.
  • Launch the Predictor application from the Applications Gallery as shown in Figure 4. Fill out the information on the Launch Predictor Application tab and click Launch.

Figure 2 Predictor app demo page
Figure 2. Predictor application demo page

Figure 3 Launching the predictor app from the Deployment page
Figure 3 Launching the predictor app from the Deployment page

Figure 4 Launching the Predictor app from the Applications Gallery
Figure 4. Launching the Predictor application from the Applications Gallery

After launching the Predictor application, DataRobot will open the Current Applications page shown in Figure 5. In this tab you can view all applications that you’ve deployed already. You can view details about each application by clicking the hamburger button next to each application’s name. Clicking that menu shows information such as rotating your authentication token, deactivating the application, sharing the application, or deleting it.

Figure 5 Current Applications tab displaying all the deployed apps
Figure 5. Current Applications tab displaying all the deployed applications

Figure 6 shows the Predictor application interface after opening the application. Here you can experiment with creating a new record. Alternatively, you could upload a file with one or more records that need to be scored in bulk. The Predictor application will submit those records to your deployed model and return the results.

Figure 6 The Predictor app interface when the application is first opened
Figure 6. The Predictor application interface when the application is first opened

Let’s create a record for a hypothetical Mr. Jones and pretend that we are Mr. Jones’s doctor (See Figure 7). We would have information about his medical history and his current status. For the purpose of this demo, we will fill in average values for these fields, random values for the text features, and press GET PREDICTION!

Figure 7 Entering values for a new record
Figure 7. Entering values for a new record

The Predictor application returns the prediction depicted in Figure 8, which shows that Mr. Jones has a 34.52% chance of re-admissions. We can also see the historical distribution for the training data (Figure as well as the prediction explanations showing what impacted the predicted value the most (Figure 9). If you want to explore the effect of changing Mr. Jones’s record, you can do that by entering other values for one or more variables and resubmit Mr. Jones’s record for prediction.

Figure 8 Prediction scores and historic distribution from the predictor app
Figure 8. Prediction scores and historic distribution from the Predictor application

Figure 9 Prediction explanations from the predictor app
Figure 9. Prediction explanations from the Predictor application

The Predictor application returns all predictions as shown in Figure 10. You can scroll down the predictions and/or download them. If you need to share this application with your business stakeholders, you can do so by clicking the share icon on the top right corner of the interface depicted in Figure 10.

Figure 10 Returned predictions after submitting them to the respective deployed model
Figure 10. Returned predictions after submitting them to the respective deployed model

See DataRobot in Action
Request a Demo
About the author
Linda Haviland
Linda Haviland

Community Manager

  • Listen to the blog
  • Share this post