Stock Selection using Recon
 *George H. John        !Peter Miller             Randy Kerber
 Comp. Sci. Dept        Lockheed AI Center        1422 Flora Ave.
 Stanford Univ.         3251 Hanover Street       San Jose,  CA 95130 
 Stanford, CA 94305     Palo Alto, CA 94304
 gjohn@cs.stanford.edu  pmiller@aic.lockheed.com  kerber@best.com

* http://robotics.stanford.edu/~gjohn/
! http://hitchhiker.space.lockheed.com/~recon/

  We approach the problem of stock selection from the perspective of
  knowledge discovery in databases: given a database of
  several years of quarterly information on over a thousand
  companies, discover patterns in the data that will allow one to
  predict which stocks are likely to have exceptional returns in the
  future.  The database includes measures of trends in the stocks'
  prices as well as fundamental data on the companies.  For this
  task we employed the Recon system, which is able to induce a set
  of classification rules or a neural network to model the data it
  is given.  To evaluate Recon's performance in the stock selection
  task, we paper-traded a portfolio of the fifty stocks ranked
  highest by Recon.  When trading costs were taken into account,
  Recon's portfolio had a total return of 238% over a four-year
  period, significantly outperforming the benchmark, which returned
  93.5% over the same period.  The performance is not attributable
  to growth/value or size effects alone.  We conclude that Recon is
  a valuable tool for stock selection.

Recon is a trademark and service mark of Lockheed Missiles & Space
Company, Inc., subsidiary of Lockheed Corporation, a Lockheed Martin
Company.

Citation:

George H. John, Peter Miller, and Randy Kerber.  Stock Selection
using Recon.  In Y. Abu-Mostafa, J. Moody, P. Refenes, and A. Weigend,
editors, _Neural Networks in Financial Engineering_, pages 303--316.
World Scientific, London, 1996.