Predicting Currency Crises
Presented: Tuesday, April 10, 2007
Speaker: Kris Kumar, Citigroup
View the on-demand web cast. Download instructions on how to view the web cast.
Download the web cast presentation.
Discover how statistical learning is being applied to predict currency crises. This web cast will focus on a model based on statistical learning that is a substantial improvement over existing approaches. The currency risk predicted from the model can then be hedged using Forwards, Options and other existing market instruments. Topics include:
- Currency Crises and its economic impact for participants in the FX market place
- Application of statistical learning algorithms to emerging market currencies
- Performance comparison of the naïve bayes classifier to parametric classifiers Fisher’s Linear Discriminate Classifier and Quadratic Discriminate classifier and the non-parametric k-nearest neighbour, Classification Trees, and Support Vector Machines.
- In-sample and out-of-sample performance of the classifiers
- Feature selection and pruning the error rate
- Results from combining the classifiers using ensemble learning and "bundling"
Kris Kumar, Citigroup
|
Krishna Kumar is a Vice President and Quantitative Analyst at Citigroup's Foreign Exchange Value Added Services and Products. From 2003 until early 2006, Mr. Kumar was a Quantitative Risk Analyst in Model Validation (Risk Architecture) at Citigroup, focused on model risk in Fixed Income, FX and Credit pricing models. Prior to that he was a Quantitative Associate with GlobeOp Risk Services, providing Hedge Fund and Fund-of-Funds risk measurement. Mr. Kumar holds a M.S. in Computational Finance and a B.S. (Hons) in Mathematics. |
|
