Skip to main content

Auto-Categorization

The auto-categorization feature uses machine learning to predict the category of a transaction based on previously categorized transactions.

Setting Up​

tip

The auto-categorization feature is trained on previously categorized transactions, so it is recommended to categorize a good number of transactions manually first to provide a strong training dataset for the model.

To set up auto-categorization:

  1. Navigate to the Settings page and find the Auto-Categorization settings under the Advanced Settings section.
  2. Train the machine learning model by clicking the Train Model button. This may take a few minutes depending on the amount of transaction data you have and the performance of your machine.
  3. Once the model training is complete, you can enable auto-categorization by clicking the toggle button.
  4. Auto-categorization is now enabled, and new transactions will be automatically categorized.
note

Auto-categorization is only applied to transactions imported via external provider sync or CSV import.

Configuration​

The auto-categorization feature can be configured with the following settings.

Confidence Threshold​

When making a prediction, the machine learning model will have a confidence level associated with that prediction. The confidence threshold setting allows you to specify a minimum confidence level for auto-categorization. If the model's confidence in its prediction is below the threshold, the transaction will not be auto-categorized and will instead be left uncategorized for manual review.

To configure this threshold, navigate to the Settings page and find the Auto-Categorization settings under the Advanced Settings section. You can adjust the confidence level with the Auto-Categorizer Minimum Probability setting.