Predict Real-Time

Predict real-time output value by changing the input value.

Note: To reset the simulation, click Refresh > Hard Refresh option in the Setup tab. The Hard Refresh option will purge the current data and reset to the beginning.

Predict real-time output values by changing the input values using the What-If analysis method.

  1. Click Predict to view the Model Quality values.
    Figure 1. Predict Real-Time

    Predict Real-Time
    The Model Quality and Predict values are displayed.
    Figure 2. Predict Real-Time Information

    Predict Real-Time Information
  2. Click next to Model Quality to view its values.
    Figure 3. Model Quality
    Model Quality
    The model field prediction values are displayed.
    Figure 4. Model Quality - Field Prediction Values
    Model Quality - Field Prediction Values
    The field prediction metrics and its description are as follows:
    • r2 (coefficient of determination) - Generally has a useful lower bound of zero. < 0 indicates a simple mean would provide better predictions. The desired value limit is 1 and the range is [-inf,1]
    • r2f (flattened coefficient of determination) - Generally has a useful lower bound of zero. < 0 indicates a simple mean would provide better predictions. Flattening done over time axis. The desired value limit is 1 and the range is [-inf,1]
    • mse (mean square error) - The desired value limit is unit^2 and the range is [0,inf]
    • mae (mean absolute error), mdae (median absolute error), and mxae (maximum absolute error) - The desired value limit is 0 and the range is [0,inf]
    • rmae (relative mean absolute error), rmdae (relative median absolute error), and rmxae (relative maximum absolute error) - Normalized relative to the observed range (max-min). The desired value limit is 0 and the range is [0,inf]
  3. Click Scalar to view the details.
    Figure 5. Model Quality - Scalar Values
    Model Quality - Scalar Values

    R-Square is a measure of the quality of the machine learning model. If the R-Square value is close to 1.0, the model is more accurate in predicting known data points. You can proceed with Predict Study or Optimization if R-square values are greater than 0.7. The values in the R-Square Test column are the most accurate indicators. In the absence of R-Square Test values, the values from the R-Square Cross Validation column can be used.

  4. Click Predict to view the predict real-time output values.
    Figure 6. Predict Real-Time Values
    Predict Real-Time Values
    The output values are displayed.
    Figure 7. Predict Real-Time Output Values
    Predict Real-Time Output Values
  5. Click in the output to display the curve.
    Figure 8. Curve Prediction

    Curve Prediction
    The curve for the output is displayed.
    Figure 9. Output Curve

    Output Curve
  6. Click View to view the predicted model preview.
    Figure 10. Predicted Model Preview - Information
    Predicted Model Preview - Information
    The Info panel is displayed with the predicted model preview.
    Figure 11. Predicted Model Preview - Info Panel
    Predicted Model Preview - Info Panel
  7. Click the play button in the Info panel.
    The predicted model preview will be generated and displayed.
    Figure 12. Predicted Model Preview - Generated Preview
    Predicted Model Preview - Generated Preview
  8. Click Save to save the predicted model.
    Figure 13. Predicted Model Preview - Save
    Predicted Model Preview - Save
    The save dialog box is displayed.
    Figure 14. Predicted Model Preview - Save Dialog
    Predicted Model Preview - Save Dialog
  9. Enter the file name of the predicted model preview and click Save.

    The predicted model is saved in the Altair One Drive location. By default, a folder predicted is created under the project folder name. You can change the folder name by navigating the Altair One drive location.

  10. Slide or double click and edit the values of trained model input value in the Predict panel to generate a new predicted model.
    Figure 15. Predict Real-Time Values

    Predict Real-Time Values
  11. Click Predict to get the updated predict real-time output values.
    Figure 16. Predict Real-Time Updated Values
    Predict Real-Time Updated Values
    Note: Generate multiple prediction model files by changing the input values and save them to compare the predicted model files.
Now you can proceed to Optimization.