Module ml 

ML-based transaction categorization.

Trains a Multinomial Naive Bayes classifier on existing ledger transactions to predict the expense/income account for new transactions based on their payee and narration text.

Uses TF-IDF vectorization plus a small, self-contained Multinomial Naive Bayes classifier (see MultinomialNB) implemented in pure std — no external ML or linear-algebra crates. Earlier versions delegated to linfa-bayes and then ferrolearn-bayes, but both dragged heavy, occasionally wasm-incompatible dependencies in for an algorithm that is ~80 lines of textbook arithmetic.

Example

let model = CategorizationModel::train(&existing_directives)?;
let predictions = model.predict("WHOLE FOODS", Some("groceries"));
// → [("Expenses:Groceries", 0.92), ("Expenses:Dining", 0.05), ...]

Structs

CategorizationModel

A trained categorization model.

Enums

MlError

Error type for ML operations.