Neural Nets in Finance and Economics
Current Status (Jan 1996)
While pursuing this project and getting
some negative results on artificial timeseries data with neural networks
(the nets kept falling into local minima), I obtained quite positive
results on real data using a rule-induction algorithm, and published
these results in two conference papers (see my
publications). I am always looking for new and better sources
of data to build on the experiments described in those papers. Please
let me know if you have any suggestions.
Project Description
This is a project which began Spring quarter 1995 at Stanford. The
participants are exploring the use of nonlinear modelling methods in
economic and financial problems, broadly defined. The first and
current stage of the project is exploratory -- reading relevant
literature, obtaining relevant data, and observing the performance
of nonlinear modelling methods applied to the data.
Participants:
- George H. John
- Michael Ebstyne
- David Lin
- Mark Shaw
Resources
- Neural Networks FAQ
- Investments FAQ
Readings:
- New Brains: how "smart" computers are beating the stock market. By
Jonathan R. Laing, Barron's Feb 27, 1995, p 29-33.
- Stock Selection using Recon[TM/SM]. By George John and Peter Miller.
In Neural Nets and the Capital Markets, London, UK. 1995. World Scientific.
Back to My Home Page
© George H. John / gjohn@cs.stanford.edu