Credit Scoring

Optimize credit portfolios and verify proofs of income by automating credit decisions in milliseconds, radically outperforming traditional scorecards in both consumer and business lending

Help underwriters evaluate creditworthiness using alternate data sources

Predict credit potential of those with no to little credit history data

Real time updates to scores as spending, employment status, etc change

Reduced model building from 6 mos to a few days

Better predictions with networks

Increase accuracy with graphs that add “highly predictive features” and allows for capturing unique features about customer nodes

H2O.ai saved $20M per year in credit underwriting for single customer

More accurate predictions of bad credit

Expanded channels to underserved, also providing financial inclusion

Use datasets with meta-features  in a networked context to generate accurate credit scores of individuals by training machine learning models to generate credit scoring models. 

Capture Network Properties of the Customers

Predictive Modelling to give better credit scores.
 

H2O AI App, built with Wave: Credit Scoring with Graphs

To interact with the Credit Scoring with Graphs app, you must log in with your H2O Cloud Account. If you do not have an account, you can request a demo here. This app demonstrates the use of machine learning as part of an overall plan to minimize employee attrition. Users can view predictions of employee departure, forecast churn rates, and identify relevant factors contained in employee data.

• Data Preparation (Customer Metadata, Telecom Data Calls made by the customers)

• Nodes are the Customers, if they called each other in a given time span, there exists an edge between them.

• Tabular Data converted into a Network/Graph Data.

• Driverless AI is used to generate credit scoring models, Wave is used to display the outputs.

• Credit Score displayed on Wave

• Unsupervised Machine Learning, Graph Neural Network, Predictive Modelling

• Interactive Component: users can tune multiple parameters (Graph Complexity, Sampling Ratio, Vector Dimension, Node2Vec epochs, etc.

• Dashboard to display insights