Bank categorization

Transaction categorization consists of assigning a nature to each entry on a statement: salary, rent, loan instalment, subscription, maintenance payment, refund. Without it, a list of transactions is only a string of heterogeneous labels, often truncated and specific to each bank. With it, income, fixed expenses and commitments can be reconstructed. This is the step that makes bank data usable for assessing creditworthiness. The difficulty lies in the labels themselves: the same retailer is written ten different ways, and a transfer between two accounts held by the same person must not be counted as income, at the risk of overstating their situation.