[Finanz.Inform] ISOR-Kolloquium am 30.05. 16:30

[Markus Fulmek]

Sehr geehrte Damen und Herren,

Ich leite Ihnen untenstehende Einladung zum ISOR-Kolloquium am 30.05. 16:30 weiter,
die vielleicht fuer Sie von Interesse ist.

Mit besten Gruessen,

Markus Fulmek

P.S. im Sinne des Telekommunikationsgesetzes: Sie erhalten
diese Information, weil Sie in der Mailing-Liste von INFORM
   https://lists.univie.ac.at/mailman/listinfo/finanz.inform
   (frueher:  http://www.fam.tuwien.ac.at/mailman/listinfo/mrm-l)
eingetragen sind. Wenn Sie diese Informationen NICHT (mehr)
wuenschen, senden Sie bitte einfach diese Email zurueck.


Begin forwarded message:

> From: Immanuel Bomze <immanuel.bomze at univie.ac.at>
> Date: 27. Mai 2011 12:43:53 MESZ
> To: immanuel.bomze at univie.ac.at, isds_kolloquium at lists.univie.ac.at
> Subject: [Isds_kolloquium] ISOR-Kolloquium am 30.05. 16:30
> 
> Sehr geehrte InteressentInnen des ISOR-Kolloquiums,
> 
> wieder darf ich Sie sehr herzlich zum naechsten Vortrag in unserer Reihe einladen:
> 
> ------------------------------------------------------------------------------------
> 
> Zeit: 30.05.2011, 16:30 
> 
> Ort: Leopold-Schmetterer-Seminarraum, 1010 Wien, Univ.str. 5/3.Stock
> 
> 
> J. Micheler (EZB Frankfurt): Support Vector Machines as a rating methodology in banking and finance
> 
> Support Vector Machines are currently a widely used classification technology in the field of pattern recognition and biomedicine because of its high generalization ability. Moreover, its efficient way of transforming data into a high dimensional space using different kernels and determining a decision boundary there, is unmatched. The primary goal of this thesis is to propose Support Vector Machines as a rating methodology where it did not get the deserved attention from the financial sector. I want to show how this advanced technique can be used to distinguish between solvent and insolvent corporate obligors and propose promising algorithms to estimate a point estimation of the probability of default. Further, the link between currently widely used discriminant techniques like the classical Fisher Discriminant method, the Logistic Regression and the Support Vector Machines is worked out theoretically and each of their generalization ability is empirically tested.
> 
>  The Austrian National Bank provided their ICAS (Inhouse Credit Assessment System} data, containing a huge number of yearly balance sheet positions from 2005 until 2010 of defaulted and solvent Austrian companies. Using Support Vector Machines, the ICAS data is used to find a rating model, where different variable selection methods like AUROC, VC-dimension and bounds of the generalization ability according to Structural Risk Minimization are used in a bootstrap algorithm.
> 
>  The key result of my thesis is that the SVM classification technique delivers robust and reliable ratings and probabilities of default especially if the number of usable data is limited and heavily biased. In fact, Support Vector machines can be used successfully to increase the performance of risk management in the financial system.
> ---------------------------
> 
> Wie immer finden Sie alle Details unter
> 
> http://www.univie.ac.at/statistics/isdskoll/
> 
> Mit besten Gruessen
> 
> Immanuel M. Bomze
> 
> Dept.of Statistics and Operations Research
> 
> University of Vienna
> 
> http://isor.univie.ac.at
> 
> http://homepage.univie.ac.at/immanuel.bomze/
> 
> Paper mail address:
> 
> Brünner Strasse 72
> A-1210 Wien
> Austria
> 
> Phone: voice +43-1-4277.38652
>            fax   +43-1-4277.38699
> 
> 
> 
> 
> _______________________________________________
> isds_kolloquium mailing list
> isds_kolloquium at lists.univie.ac.at
> https://lists.univie.ac.at/mailman/listinfo/isds_kolloquium

.......................................................................
Faculty of Mathematics ................ Nordbergstr. 15,  A-1090 Vienna
Phone: +43-1-427750678 . Mobile: +43-664-1031751 . Fax: +43-1-427750753
@: Markus.Fulmek at univie.ac.at ....... W3: www.mat.univie.ac.at/~mfulmek
.......................................................................

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.univie.ac.at/pipermail/finanz.inform/attachments/20110527/4914b9d9/attachment.html>