About Predicted Attributes
A predicted attribute is an attribute that can predict the value of a specified attribute by using Machine Learning algorithms that are trained on your historical data.
For each predicted attribute, a separate attribute is automatically created to store its Confidence score. Predicted attributes and their corresponding confidence percentages can be included as columns in the Reconciliation list and Profile list. Predicted attributes can also used in formats, but they are read-only. The values of predicted attributes can be copied into standard attributes or custom attributes.
The accuracy of predicted attributes can be determined using the following parameters:
- Expected Accuracy
After a predicted attribute is created and trained, it is assigned an Expected Accuracy. This accuracy is for the prediction model and it indicates the overall accuracy of the model, with the specified parameters. This value is displayed in the Edit Predicted Attribute dialog and can be added as a column in the Predicted tab under Attributes tab.
- Confidence
For each attribute value that is predicted using the predicted attribute, a confidence percentage is generated. This value indicates the accuracy of the individual prediction. A value of 100 represents the highest confidence and zero represents the lowest. Both the predicted attribute value and its confidence can be used wherever attributes appear.
For example, you create a predicted attribute named PredictRiskRating.
After this attribute is saved, Account Reconciliation automatically creates a
corresponding attribute PredictRiskRating (Confidence) to store the
confidence percentage for this predicted attribute values. After the predicted attribute
is successfully trained, the Edit Predicted Attribute dialog
shows a value for Expected Accuracy. This value represents the
overall accuracy of the prediction model. Next, you use the
PredictRiskRating predicted attribute to predict attribute values
for the November 2025 period. The trained predicted attribute predicts attribute values
for all reconciliations in the November 2025 period. Assume that 30 attribute values are
predicted. For each of these values, a percentage is assigned for
PredictRiskRating (Confidence). So, PredictRiskRating
(Confidence) could be 95 for the one reconciliation and 89 for the second.
These values represent the confidence level of the prediction for those reconciliations.
Both the predicted attribute and its corresponding confidence attribute can be displayed
in the Reconciliation list and Profile list. Use Select Columns
to select the attributes that you want to display.
Target and Parameters Allowed for Predicted Attributes
The following table lists the types of attributes that can be used as Target and Parameter attributes for predicted attributes.
Table 4-3 Predicted Attributes
| Target of Predicted Attribute | Parameters of Predicted Attribute | |
|---|---|---|
| Standard Attributes (Text, Number, Integer, Yes/No, True/False, List, User) | Yes | Yes |
| System Attributes (Text, Number, Integer, Yes/No, True/False, List, User) | Yes | Yes |
| System Attributes (Multi-Line Text, Date, Date/Time) | No | No |
| Standard Attributes (Multi-Line Text, Date, Date/Time) | No | No |
| Predicted Attribute | No | No |
| Predicted Attribute Confidence | No | No |
| Balance Attribute | No | Yes |
| Group Attribute | No | No |
| Group Attribute Members | No | Yes |
| Calculated Attribute | No | Yes |
| Transactional User Attributes, Action Plan Attributes | No | No |
| Account Segments Attribute | No | Yes |