Skip to content

On-premise users: click in-app to access the full platform documentation for your version of DataRobot.

NextGen experience > AI experimentation > Predictive experiments > Evaluate with model insights

Evaluate with model insights

Model insights help to interpret, explain, and validate what drives a model’s predictions. Using these tools can help to assess what to do in your next experiment. Available insights are dependent on experiment type as well as the experiment view (single versus comparison).

Available insights

To see a model's insights, click on the model in the left-pane Leaderboard—the Model Overview opens. From here, all available experiment insights are available, grouped by purpose and answering:

  • Explanations: What did the model learn?
  • Performance: How good is the model?
  • Details: How was the model built?
  • Artifacts: What are the assets from the model?

Use search to filter insights by name and/or description. The results also mark which group the insight belongs to.

Note that different insights are available for time-aware experiments.

Insight / tab Description Problem type Sliced insights? Compare available?
Accuracy Over Space
Performance tab		 Reveals spatial patterns in prediction errors and visualizes prediction errors across data partitions on a map visualization. Geospatial
Activation Maps
Explanations tab		 Highlights regions of an image according to its importance to a model's prediction. Visual AI, time-aware predictive
Anomaly Over Space
Performance tab		 Maps anomaly scores based on a dataset's location features. Geospatial
Blueprint
Details tab		 Provides a graphical representation of data preprocessing and parameter settings. All
Cluster Insights
Explanations tab		 Visualizes the groupings of data that result from modeling with learning type set to clustering. Predictive clustering
Coefficients
Explanations tab		 Provides a visual indicator of the relative effects of the 30 most important variables. All; linear models only
Compliance documentation Generates individualized documentation to provide comprehensive guidance on what constitutes effective model risk management. All
Confusion matrix Performance tab
Compares actual with predicted values in multiclass classification problems to identify class mislabeling. Classification, time-aware
Downloads
Artifacts tab		 Download model artifacts in a single ZIP file. All
Eureqa Models
Details tab		 Uses a proprietary Eureqa machine learning algorithm to construct models that balance predictive accuracy against complexity. All, no multiclass
Feature Effects
Explanations tab		 Conveys how changes to the value of each feature change model predictions All ✔
Feature Impact
Explanations tab		 Shows which features are driving model decisions. All ✔ ✔
Image Embeddings
Explanations tab		 Shows projections of images in two dimensions to see visual similarity between a subset of images and help identify outliers. Visual AI, time-aware predictive
Individual Prediction Explanations
Explanations tab		 Estimates how much each feature contributes to a given prediction, with values based on difference from the average. Binary classification, regression ✔
Lift Chart
Performance tab		 Depicts how well a model segments the target population and how capable it is of predicting the target. All ✔ ✔
Log
Details tab		 Lists operational status results for modeling tasks. All
Metric Scores Performance tab
Displays results for all supported metrics. All
Model Info
Details tab		 Provides general model and performance information. All
Model Iterations
Details tab		 Compares trained iterations in incremental learning experiments. Binary classification, regression
Multilabel: Per-Label Metrics
Performance tab		 Summarizes performance across different label values of the prediction threshold. Multilabel classification
Neural Network Visualizer
Details tab		 Provides a visual breakdown of each layer in the model's neural network. Visual AI, time-aware predictive
Related Assets
Artifacts tab		 Lists all apps, deployments, and registered models associated with the model; launches no-code apps creation or model registration. All
Residuals
Performance tab		 Provides scatter plots and a histogram for understanding model predictive performance and validity. Regression ✔
ROC Curve Performance tab
Provides tools for exploring classification, performance, and statistics related to a model. Binary classification ✔ ✔
SHAP Distributions: Per Feature
Explanations tab		 Displays, via a a violin plot, the distribution of SHAP values and feature values to aid in the analysis of how feature values influence predictions. Binary classification, regression ✔
Word Cloud
Explanations tab		 Visualize how text features influence model predictions. Binary classification, regression
Insight Description Problem type Sliced insights? Compare available?
Explanations
Activation Maps Highlights regions of an image according to its importance to a model's prediction. Visual AI, time-aware predictive
Cluster Insights Visualizes the groupings of data that result from modeling with learning type set to clustering. Predictive clustering
Coefficients Provides a visual indicator of the relative effects of the 30 most important variables. All; linear models only
Feature Effects Conveys how changes to the value of each feature change model predictions All ✔
Feature Impact Shows which features are driving model decisions. All ✔ ✔
Image Embeddings Shows projections of images in two dimensions to see visual similarity between a subset of images and help identify outliers. Visual AI, time-aware predictive
Individual Prediction Explanations Estimates how much each feature contributes to a given prediction, with values based on difference from the average. Binary classification, regression ✔
SHAP Distributions: Per Feature Displays, via a a violin plot, the distribution of SHAP values and feature values to aid in the analysis of how feature values influence predictions. Binary classification, regression ✔
Word Cloud Visualize how text features influence model predictions. Binary classification, regression
Performance
Accuracy Over Space Reveals spatial patterns in prediction errors and visualizes prediction errors across data partitions on a map visualization. Geospatial
Anomaly Over Space Maps anomaly scores based on a dataset's location features. Geospatial
Confusion matrix Compares actual with predicted values in multiclass classification problems to identify class mislabeling. Classification, time-aware
Lift Chart Depicts how well a model segments the target population and how capable it is of predicting the target. All ✔ ✔
Metric Scores Displays results for all supported metrics. All
Multilabel: Per-Label Metrics Summarizes performance across different label values of the prediction threshold. Multilabel classification
Residuals Provides scatter plots and a histogram for understanding model predictive performance and validity. Regression ✔
ROC Curve Provides tools for exploring classification, performance, and statistics related to a model. Binary classification ✔ ✔
Details
Blueprint Provides a graphical representation of data preprocessing and parameter settings. All
Eureqa Models Uses a proprietary Eureqa machine learning algorithm to construct models that balance predictive accuracy against complexity. All, no multiclass
Log Lists operational status results for modeling tasks. All
Model Info Provides general model and performance information. All
Model Iterations Compares trained iterations in incremental learning experiments. Binary classification, regression
Neural Network Visualizer Provides a visual breakdown of each layer in the model's neural network. Visual AI, time-aware predictive
Artifacts
Compliance documentation Generates individualized documentation to provide comprehensive guidance on what constitutes effective model risk management. All
Downloads Download model artifacts in a single ZIP file. All
Related Assets Lists all apps, deployments, and registered models associated with the model; launches no-code apps creation or model registration. All

What's next?

After selecting a model, you can, from within the experiment:

Updated March 12, 2025