Title: Credit Risk Analytics Modeling
Presented by:
Michael Libman
Managing Director
BancLab LLC
New York, NY
Description: BancLab provides customized credit risk analytics solutions for lenders and government agencies. Our core engine for analysis is a large database that contains 25 years of credit performance on over 1 million loans. There are several analytics problems that we have to deal with when working with our clients:
1. Sampling Problem: Our clients generally have portfolios that consist of less than 1000 loans and require a much larger sample from which to analyze historical performance. Alternatively, some of our clients want to expand into new markets (ex. New York bank that wants to consider expansion into Florida) and these clients require information on how a portfolio similar to theirs would perform in these new markets. The goal of this project is to develop a general strategy for identifying the composition of the representative sample of K loans from our database of 1 million loans that matches the
composition of a client's portfolio based on N characteristics and establish new ways to assess an accuracy in matching.
2. Portfolio Problem: Our clients have a historical portfolio of a fixed number of loans and want us to help them improve their credit performance. The goal of this project is to develop a mathematical model that identifies loan types that when added to the client's portfolio could improve the "adjusted" portfolio credit performance.