KTH Matematik |

In this thesis we put forward a possible solution for Cinnober Financial Technology AB, to be able to calculate the risk of a portfolio of interest rate swaps in real time using risk modeling. The chosen instrument is plain vanilla swaps and the time horizon is one day. The main assumptions has been that the risk of having a portfolio of plain vanilla interest rate swaps as an intermediary between various parts can be modeled based on historical data. The main approach in modeling has been a straight forward calculation based on Principal Component Analysis (PCA). Thereby daily VaR at 95% level of condence has been calculated using Monte Carlo-simulated future data. The PCA approach has been put into stress testing where the main reason has been to study the robustness of the approach. The argue is that stress testing combined with VaR gives a more comprehensive picture of risk. While VaR gives a picture of the risks in an everyday market environment, stress testing instead gives a picture of the risk in an abnormal market. When it comes to real-time risk management, the calculations should be optimized in order to reduce the latency. Two methods have been suggested in this paper that need further preparation and implementation, Parallel Vector Computing and Incremental Computing. These methods make it possible to calculate the risk as fast as possible. Thereby real-time risk management is not longer a dream but a reality that should be required from any participant in the finance sector. |

Sidansvarig: Filip Lindskog Uppdaterad: 25/02-2009 |