AI Report¶
Once you complete an Autopilot run, you can generate an AI Report, which communicates the most important findings of your modeling project to stakeholders. The AI Report provides a high-level overview of your modeling results and insights, with particular focus on Trusted AI insights that fall under the dimensions of quality, accuracy, and interpretability.
The report provides accuracy insights for the top performing model, including its speed and cross-validation scores. It also captures interpretability insights in the model's Feature Impact histogram, which helps to show which features are driving model decisions.
The AI Report provides these summary details:
- The target and the number of models trained.
- The type of problem, for example, binary or regression.
- The modeling mode, for example, Quick Autopilot.
It also provides details about your data and its quality, including:
- The number of rows of data.
- The number and type of features (for example, categorical, numeric, text, etc.).
- The number of features determined to be informative—DataRobot removes the features that are not informative.
Availability is as follows:
- Reports can be created for projects built with all modeling modes except manual mode.
- The option is unavailable (the functionality is disabled) for time-aware, unsupervised, and multiclass/multilabel projects.
Generating an AI Report¶
To generate an AI Report for your Autopilot run, on the Leaderboard, click Menu and select AI Report > Generate new report.
DataRobot automatically downloads the AI Report as a Word document after it is generated.
Viewing an AI Report¶
To view a previously generated AI Report, on the Leaderboard, click Menu and select AI Report > View report.
Downloading an AI Report¶
To download a previously generated AI Report, on the Leaderboard, click Menu and select AI Report > Download report.
DataRobot downloads the AI Report as a Word document.