Monitoring new values of categorical fields helps to locate and examine changes in the model's input.
For example, setting the monitor for a feature named
statewill help us discover a new region for which the model is asked to predict results.
For this monitor, the following comparison methods are available:
Configuration may slightly vary depending on the comparison method you choose.
You may select as many fields as you want 😊
Note that the monitor will run on each field chosen separately.
The monitor supports both
arrayfield types(for categorical arrays).
arrays, it will monitor all the categories in the arrays together and not per dimension. Please note that by default, the monitor will run only over the first 500 categories seen within the arrays and ignore the rest, if your use case requires monitoring arrays with more unique values, please reach out so we can update your specific configuration.
categoricalfields, the monitor will support up to 256 unique values.
For the fields you chose in the previous step, the monitor will raise an alert if the number of new values in the inspection period compared to the baseline values exceeds your threshold.
This step is essential to ensure you have the right amount of alerts that fit your needs. You can always readjust it later if needed.