2 options

Using genetic algorithms to find technical trading rules in financial markets.

Online

Available online

Connect to full text

Dissertations & Theses @ University of Pennsylvania Available online

Dissertations & Theses @ University of Pennsylvania
Format:
Author/Creator:
Contributor:
Language:
English
Subjects (All):
Local Subjects:
Physical Description:
336 pages
Contained In:
Dissertation Abstracts International 55-09A.
System Details:
  • Mode of access: World Wide Web.
  • text file
Summary:
There is a long practical tradition of "technical analysis" in many organized financial markets to forecast future price changes. As technical trading rules rely on publicly available historical data such as past prices, they should not lead to risk-adjusted profits above the equilibrium return in informationally efficient markets. In the previous trading rule tests, the rules have been taken as given. In this dissertation, a genetic algorithm based on the ideas of natural evolution is used to infer technical trading rules from the past prices. The algorithm was applied to a composite stock index (S&P 500) and separately to the corresponding futures. For the S&P 500 index, the rules found by the algorithm led to statistically significant excess returns above the buy-and-hold strategy in the out-of-sample test period of 1970-89. Rules had significant market timing ability, and the results could not be explained by models of stock returns tested through bootstrapping simulations. For S&P 500 futures, it was found that the trading rule signals were triggered by large price changes, which were later partially reversed. There were also consistent patterns in the option trading activity coinciding with the trading rule signals. While the observed patterns may be due to rational hedging behaviour, the findings are also consistent with an explanation that prices are affected by irrational changes in the investor sentiment, arising from overreaction to recent and salient information. The results suggest that the trading rules found by the genetic algorithm exploit the cognitive biases of market participants.
Notes:
  • Thesis (Ph.D. in Decision Sciences) -- Graduate School of Arts and Sciences, University of Pennsylvania, 1994.
  • Source: Dissertation Abstracts International, Volume: 55-09, Section: A, page: 2931.
  • Supervisor: Franklin Allen.
Local Notes:
School code: 0175.
Access Restriction:
Restricted for use by site license.

The Penn Libraries is committed to describing library materials using current, accurate, and responsible language. If you discover outdated or inaccurate language, please fill out this feedback form to report it and suggest alternative language.