View model accuracy

View how well your model is performs over time. For a more complete picture, the graph also shows model accuracy when trained and feature coverage.

Related documentation:

• How Predictive Routing works

The Model Accuracy tab

The Model Accuracy tab provides a view into how well your model performs over time.

This tab is on the Predictors window:

Prerequisites: To use this tab, your account must include one or more Predictors and have scoring data available.

Configure the view

The Model Accuracy tab enables you to select the following information:

• The name and ID of the predictor. Use the selector to specify which predictor to view.
• The name and ID of the model. Use the selector to specify which model to view.
• The displayed date range. Use the year and month selectors to specify the month to view.
• The Plot By selector enables you to set the granularity with which the graph displays data. You can choose to view data by Day, Hour, or 15-Minute interval.

Note: This page does not display real-time updates. GPR refreshes the data displayed once per day, at 3:00 am UTC.

How to read the graph

The Model Accuracy graph shows the mean absolute error value, indicating the difference over time between the results the model predicts and the actual outcomes. This value, and how to interpret it, varies depending on the type of KPI your model optimizes.

The graph also includes a horizontal line marking the accuracy level when the model was trained, enabling you to compare the training result with real-world performance. The gray bars show the feature coverage level, since feature coverage issues are often the reason for degradations in model accuracy.

Note: The Feature Coverage tab can help you troubleshoot issues if feature coverage does seem to be the source of a model's drift away from accuracy. To view the five features having the strongest impact on the KPI value, see the Top Features horizontal bar graph on the Simulator tab.

Example

When you view the graph, keep in mind that the right-hand vertical axis shows the mean absolute error rate, not the actual KPI value. The example image on this page shows the graph for a model designed to optimize CSAT. The horizontal red training line shows that when you created the model, it had an error rate of 1.26. That is, the average difference between the predicted CSAT and the actual CSAT was 1.26 points. As you have used this model over time, this difference between the predicted CSAT and the actual CSAT has ranged from 1.40 to 1.56.

To export the model accuracy data as a CSV file, click the Export to CSV button.