Resetting machine learning model

The Reset Model feature enables security teams to restart individual machine learning models from a clean baseline while preserving all custom configurations. This process removes previously learned behavioral data, allowing the model to retrain by using current and accurate information without impacting existing system settings or ongoing operations.

About this task

The Reset Model feature is used in the following scenarios:

  • Resolving issues caused by inaccurate, incomplete, or corrupted historical training data.
  • Enabling rapid model adaptation to significant organizational changes, such as mergers, acquisitions, or structural reorganization.
  • Improving detection accuracy when model performance degrades, including increased false positives or missed threat activity.
  • Supporting compliance and regulatory requirements that mandate periodic model retraining or refresh cycles.

Procedure

  1. Go to the UEBA dashboard and select the machine learning model that you want to reset.
  2. Click (Navigation menu icon) menu next to the selected machine learning model and select Reset Model.
    Machine learning models screen
  3. Review the confirmation dialog and click Yes.
    Machine learning models screen
    The model is temporarily disabled while the reset completes (1–5 minutes).
    Machine learning models screen