Family
LightGBM
productionMITtabularboth
Transaction Categorizer.
Target (F1 convergence): hierarchical L1/L2 categorization via LightGBM + enrichment. CURRENTLY SERVED: in-memory scikit-learn TF-IDF + XGBoost flat ~27-category bundle (categorizer_v8), loaded from disk.
Foundation · LightGBM
What it does
The honest description.
Target (F1 convergence): hierarchical L1/L2 categorization via LightGBM + enrichment. CURRENTLY SERVED: in-memory scikit-learn TF-IDF + XGBoost flat ~27-category bundle (categorizer_v8), loaded from disk.
Try it (mock)
Run the model on your own input.
Demo runs in your browser against a mocked edge — no signup, no production data. The same shape as the live route.
Categorise a transaction (mock)
Loading the demo…
Architecture
How it works.
Transaction rows with date/amount/description/category/account_id_hashed/payment_method. CURRENTLY SERVED: TF-IDF over description+merchant+institution_key+statement_family + amount-bucket numerics (no embeddings, no enrichment). Target: multi-level L1/L2 with confidence. CURRENTLY SERVED: single-label flat category + confidence; below-threshold falls back to the bundled rule weak-labeler.
Training data & drift
What it learned from, and how we watch it.
Source
Inputs: Transaction rows with date/amount/description/category/account_id_hashed/payment_method. CURRENTLY SERVED: TF-IDF over description+merchant+institution_key+statement_family + amount-bucket numerics (no embeddings, no enrichment).
Drift baseline · current quarter
Drift baseline not yet documented for this model.
What it's NOT
- Transitional per the F1 convergence program: the served artifact is a flat XGBoost TF-IDF bundle (categorizer_v8), in-memory sklearn inference (not Triton, not sentence-transformers).
- The LightGBM + enrichment + hierarchical-taxonomy model described above is the convergence target, not yet served.
Who calls this model
Across the house and the API.
Internal routes
No public routes documented yet.
Version history · immutable once published
Every release, traceable.
Data Mukodi holds about you is subject to DPDP §12 erasure (use /consent in the app), and any concern about this model can be raised via [email protected] under DPDP §13 grievance, or via /security.