Predict Real-Time

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

1. Click Predict to view the Model Quality values.

   
   
   Figure 1. Predict Real-Time

   The Model Quality and Predict values are displayed.

   
   
   Figure 2. Predict Real-Time Information

2. Click next to Model Quality to view its values.

   Figure 3. Model Quality

   The model field prediction values are displayed.

   Figure 4. 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

   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

   The output values are displayed.

   Figure 7. Predict Real-Time Output Values

5. Click in the output to display the curve.

   
   
   Figure 8. Curve Prediction

   The curve for the output is displayed.

   
   
   Figure 9. Output Curve

6. Click View to view the predicted model preview.

   Figure 10. Predicted Model Preview - Information

   The Info panel is displayed with the predicted model preview.

   Figure 11. 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

8. Click Save to save the predicted model.

   Figure 13. Predicted Model Preview - Save

   The save dialog box is displayed.

   Figure 14. 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

11. Click Predict to get the updated predict real-time output values.

    Figure 16. 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.