How to find insights from successfully trained supervised machine learning models?

This article outlines the different available insights in the Engine in order to successfully evaluate trained supervised machine learning models.

Step 1 - Go to a model page

From the model leaderboard or the model list, select a successfully trained model

From model leaderboard From model leaderboard 

From model listFrom model list

Step 2 - View all the available insight tools from the model summary tab

All the available tools to examine a model are listed under the “Insight” section of the model summary tab. Currently, the available tools are:

  • Performance - metrics to evaluate prediction quality

  • Feature importance - impact score of different features on a model prediction

  • Prediction explanation - prediction explanation examples taken from the training data

  • What-if analysis tool - individual predictions to explore different input value scenarios

The insight section under model summary tabThe insight section under model summary tab

Step 3 - Navigate to the insights tab and explore different tools

  • From the model summary tab, click the tab “Insights”

  • Select the tool of interest and view the details


To quickly go to specific tools from the model summary tab, click on the link in a particular tool card. For example, this is how you can quickly access the different insights and the batch prediction tool:

Summary page