Generate Anomaly Detection Predictions
- Ensure you have at least 24 months of actuals data in the accounts that you want to populate with predictions.
- To define your own upper and lower bounds for anomaly detection, create a modeled sheet. See Define Anomaly Thresholds.
- Security:Model includes sheets, accounts, dimensions, and formulaspermission.
The
Anomaly Detection
algorithms use machine learning to generate predictions into a dedicated version based on historical patterns in your actuals data. By building more precise models with machine learning, you can identify risks to avoid and profitable opportunities up to 24 months into the future.When you generate predictions for
Anomaly Detection
:
- The forecasted data populates in the prediction version.
- Anomaly detection in sheets highlights anomalies.
You can only generate machine learning predictions when your default calendar uses months as the default time strata.
- SelectModelingfrom the main menu.
- ClickAnomaly Detection.
- ClickGenerate Predictions.You can leave this area ofModelingwhile the predictions generate.
- To check the status, return to :
- Complete: Predicted data has populated your prediction version for the accounts.
- Failed: Review and complete the actions listed in theStatussection. Then clickGenerate Predictionsagain.
- In Progress: Check back later. The time it takes to complete the predictions depends on the size and complexity of your data.
- Review the prediction data against your plan data with conditional formatting. See Steps: Build Outlier Reports.