Configuring a Model
Last updated
Last updated
The Model Properties dialog is used to configure the model in terms of its target field and explanatory fields. You reach it through the pane. See for a detailed discussion of the modeling capabilities in ADVIZOR AnalystX. Review the overall Analytics to see how this dialog fits into it.
Configureing a model like this:
Model Name: Enter a unique name that identifies this new model in the "Model Name" field, or use the default name based on the table and target field.
Data Table: From the Data Table pulldown list, select a table that will be used to generate the predictive model.
Target Field: Select the Target Field from the pulldown list. This field answers the business question for the data being analyzed. Target fields may have either continuous values (e.g., real or integer numbers) or binary values (e.g., “0” or “1”). A target field cannot be a field with multiple string values.
You can create a field from the current selected field to use as the target, using the next button.
Target from Selected ...: Use this button to create a new field that contains the current selection state for use as a target.
Explanatory Fields: Explanatory fields are fields of the selected Table that will be analyzed to find those correlated with the target. Select one or more explanatory fields by clicking the check box next to the applicable field. Use the All or None buttons to turn on/off all fields. Not all fields can be explanatory fields!
The target field may not be an explanatory field.
Add data from other tables using the "copy" operation or "link()" expression in the Expression Builder.
Omit fields that are keys, where each row has a unique value.
Omit text fields containing arbitrary text that is mostly unique to each row, such as comments fields or addresses.
Date fields are omitted by default. You can translate them into an integer representing the elapsed time from a point in the past. You may want to calculate your own coding for dates, such as days from the present or days from a date in the past using the Expression Builder. You may also want to add additional fields that describe a date in terms of cycles (e.g., day of week, day of month, ...) to look for periodic patterns. These fields can be created with the Date Parser.
String fields with large numbers of categories are omitted by default. Use the Bin Categorical Field ... button (or the Expression Builder "bin()" function) to see if you can reduce the number of categories for modeling.
Choose a PValue and Training Subset to control how model building is done:
PValue: The PValue is a threshold used in determining if a relationship between an explanatory field and the target is not random and thus should be included in the model. The smaller the PValue, the less likely the relationship is to be based on random variation in the data.
Training Subset: You may train your model with a subset of the data. Use the slider to choose the percentage of the model table to randomly use. The total visible rows, rows in the training subset, and what percentage that represents is shown as well. A subset is automatically chosen if the volume of data will make the building time slow. The larger the training subset, the more accurate the model but the longer time it will take to build. You may want to use a smaller subset if you are quickly iterating models, and then use a larger subset for a final training session.
Train Model: "Training" is the process of creating the model from relationships found in the current data between the explanatory fields and the target field. This can be time consuming! The progress of model building is shown by "Accesses" in the Analytics pane.
Save: Save the model configuration without training the model.
Cancel: Close the dialog with no changes.
Help: Display assistance on using this dialog.
When processing is completed, the model is displayed in two pages in the Analyst.
Zip Codes may require special .
