FINANZIA: Computational Methods and Software for Financial Derivatives
Elias N. Houstis
Research Assistant: Kostas Pantazopoulos
Support: Purdue Research Foundation
Computational finance has already been established as a new
discipline of finance. Financial institutions
are hiring numerical analysts, engineers, and computer scientists
to apply computational techniques to a
multitude of every day problems. The option valuation problem is
among the most important applications
in need of mathematical and computational support. Option valuation
mandates the use of advanced
mathematical models. Furthermore it requires an integrated
computational support that will address the
multitude, diversity and dynamics of contemporary option products.
The difficulty with the underlying
approximate numerical models is their extrapolating nature. This
characteristic leads usually to
multidimensional free boundary value problems that are difficult to
approximate and analyze. Moreover,
their real time solution makes the study of their parallel
processing necessary. There are five technical
thrusts in the FINANZIA project:
The development and analysis of numerical methods for
non-linear parabolic initial/boundary value
problems with free boundaries and their application to the option
The development of the FINANZIA option valuation library and
its integration with existing problem
solving environments such as //ELLPACK, Excel and Matlab.
The development of parallelization methodologies to support
the mapping of option valuation models
on parallel machines including networks of workstations.
The benchmarking and performance evaluation of option
valuation algorithms in a parametric
framework following an object oriented characterization of the
option products universe.
The development of intelligent techniques for computational
finance problems including optimal
algorithm selection for the option valuation problem, neuro-fuzzy
prediction techniques for implied
parameters and secure outsourcing of financial computations.