A typical approach in estimating a customer’s credit risk is based on their repayment history and models such as logistic regression. This can be significantly improved by including additional data sources combined with more advanced techniques like graph machine learning and natural language processing.
Traditional credit risk scoring is limited in its scope to existing clients and tends to perform poorly on clients without a history of loan taking. We overcome these hurdles by considering not only the client’s loans and basic sociodemographic data but also their digital trace, their relationships with other entities and the detailed structure of their bank transactions. This leads to better performing risk models for existing clients and allows pre-scoring even for non-clients based on their behaviour online.
We use an ensemble of models, each focused on a particular angle of the customer’s behaviour and then combine them into a single unified model:
- As in traditional credit risk scoring, we also include bank statements, transaction histories and statistics from application forms
- We then dig deeper into the client’s transactions and determine the purpose of each transactions using methods based on natural language processing. This allows us to Understand the customer’s spending and earning structure which helps us get a more precise risk scoring.
- We Pair clients (and non-clients) with their identities in various digital sources (banking app, web browsing, loan calculator, ...) That way, we understand the client's interests and concerns - both correlate with risk
- we link clients to each other based on mutual transaction histories, device usage, etc. We then use Graph Neural Networks to transform these connections into a meaningful risk predictor for each client.
- Geo-location data are included in the risk scoring model for important predictors of client's affluence. This is even more useful for non-clients where we typically don't have a transaction history
- Improves the precision of credit risk scoring in general
- Enables credit risk scoring even for non-clients
- Provides a more complete risk profile of non-primary clients