Data-Driven SME Credit Scoring
A client-side risk engine that uses behavioral signals to generate instant credit ratings and probability of default (PD).
The Challenge
SME lending is often blocked by a lack of structured data. Traditional banks rely on outdated annual reports, missing the real-time signals hidden in daily transaction ledgers and cash flow patterns.
The Solution
This system ingests raw transaction data to engineer behavioral features—like cash conversion cycles and late-payment flags—feeding a calibrated PD model that outputs defensible, explainable risk ratings.
Tech Stack
Feature Engineering
The core of the model lies in transforming raw logs into signals. We look at liquidity ratios, revenue volatility, and "red flag" behaviors like increasing chargeback density to build a holistic view of borrower health.
Scoring Engine
Adjust risk factors to see how defaulting probability changesCredit Score
71/100
Risk Rating
B
Est. Default Prob.
29%
*This is a simplified interactive demo. For the full probabilistic model and Python implementation, please refer to the GitHub repository.