Speaker:	Oleksandr Romanko
	Department of Computing and Software, McMaster University
	Quantitative Research Group, Algorithmics Inc.

Title: Credit Risk Optimization



In this presentation we investigate credit risk optimization problems appearing in managing portfolios of credit positions. Credit risk modeling is a challenging problem due to the fact that the distribution of portfolio losses is not normal and exhibits fat tails. In that context, Mean-Variance approach may not be appropriate when minimizing Value-at-Risk or Expected Shortfall are the manager�s targets. We evaluate and compare a number of models for minimizing credit risk with Value-at-Risk, Expected Shortfall and Mean-Variance trade-off as the risk measures. Those models include minimizing VaR and Expected Shortfall with CLT-based nonlinear model, LLN-based linear model, Monte-Carlo sampling-based linear model and Worst-Case optimization linear model as well as minimizing variance with Mean-Variance quadratic model. Theoretical results as well as numerous computational tests on large datasets are reported. The analysis of comparative performance of the models is provided. The optimization solvers that we use for numerical computations are CPLEX, MOSEK and IPOPT. Based on our evaluation results, we draw conclusions and make practical recommendations for embedding credit risk optimization into the risk management framework.

Joint work with Ian Iscoe, Alex Kreinin and Helmut Mausser, Algorithmics Inc.