Finlens' AI transaction classification system utilizes a three-tier technology architecture:
- initial recognition layer: Parsing the transaction description text (e.g. "AMZN*MKTPLACE" is recognized as e-commerce procurement) by NLP, and matching the preset rules with the amount, counterparty, and other features.
- machine learning layer: Train classification models based on historical multi-million transaction data, and automatically recognize 300+ business scenario labels such as "office supplies" and "travel expenses".
- User feedback layer: Record each manual adjustment behavior and continuously optimize the model through reinforcement learning
Three ways of user-optimized classification:
- Instant Fix: Click on the Errors tab on the Transactions page and select the correct category from the drop-down menu
- batch file: Shift+Select multiple similar transactions and use the "Recategorize" function to unify the adjustments.
- Rule setup: Setting permanent categorization rules for specific merchants (e.g. "UBER*RIDE")
This answer comes from the articleFinlens: AI-powered financial automation tools for startups and accountantsThe
May not be reproduced without permission:AI productivity tools " How does Finlens' AI automated deal categorization feature work? How can users optimize the classification results?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil