Predictive Modeling Pane
Last updated
Last updated
The Predictive Analytics control panel lets you create models, edit existing model, or apply models. See for a detailed discussion of the modeling capabilities in ADVIZOR AnalystX.
If the Predictive Model pane is not visible, click on the Predictive Modeling button in the Model Ribbon.
To add a new model:
If a new model was trained, when processing is completed the model is displayed in two pages in the Analyst.
The model that you’ve created must be evaluated for adequacy before it is used. The metric displayed depends on the type of model:
A "Coefficient of Determination" is shown for regression models. This number, often called "R2", gives the amount of variability in the target that is captured by the model.
For classification models, an "Adjusted Count" metric is shown. This is the number of predictions that match the actual values.
The Current Model grouping shows information about the current model. The "Models" combo box shows the name of the current model; this may be used to change to another model if you have more than one built.
Operations on the current model are:
Click the "Edit ..." button to display the Model Properties dialog box. Make changes as required to the information displayed in this dialog box. Click "Train" To implement your changes, ADVIZOR builds the model.
When processing is completed, the model is displayed.
Click the Delete Model button. A message, Delete model , is displayed. Click .
The model is deleted.
Use the "Rebuild" button to re-train the current model with the existing target and explanatory fields.
If the model description pages are not visible, you may show them with the "View" button".
Creates the predicted field values using the current model.
Advanced options are:
Run equations with project: If selected, the model equation is recorded with the project and will be run whenever new data is loaded with the project.
Name bins ...: You may name the binned scores that are produced for classification models to use other than the default bin numbers. Use this to give the groups descriptive names.
By default the model is applied to all rows in the training table, even those that were excluded during training. If you wish to constrain the model to a subset, you may enter an expression that identifies the rows that should not be included. These rows will be given a missing value for the predicted field.
The top box describes the rows that were excluded during model training.
To specify a subset to not model, check the "Do not predict rows from model were" check box. This will copy the excluded row expression from the first box to the second box. You may use this expression as is or edit the expression in the second box. Click the "Apply" button when the expression is what you want. it and then click "Apply".
Click the "New Model ..." button; the dialog displays.
