Created by W.Langdon from gp-bibliography.bib Revision:1.3872

@PhdThesis{Miles:thesis, author = "Stanley Miles", title = "Adaptive efficiency of futures and stock markets : analysis and tests using a genetic programming", school = "York University", year = "2006", address = "TORONTO, ONTARIO, Canada", month = mar, keywords = "genetic algorithms, genetic programming", isbn13 = "978-0-494-19800-1", URL = "http://search.proquest.com/docview/304985250/F090C35B735040DEPQ/1?accountid=14511", URL = "http://phdtree.org/pdf/25522336-adaptive-efficiency-of-futures-and-stock-markets-analysis-and-tests-using-a-genetic-programming-approach/", URL = "https://www.library.yorku.ca/find/Search/Results?lookfor0[]=Miles&lookfor0[]=&type0[]=Author&join1=AND&lookfor1[]=Genetic+Programming&lookfor1[]=&type1[]=AllFields&mylang=en", size = "328 pages", abstract = "We propose a nonparametric method for finding approximate solutions to dynamic portfolio choice models: the use of genetic programming to directly estimate the optimal trading strategy. After we validate our methodology by conducting a simulation exercise to demonstrate that genetic programming can recover the true analytic solution to two models, we apply it to the path-dependent problem of a futures investor who is subject to initial and maintenance margin constraints, a problem that is difficult to solve using analytic methods. The resulting approximate solution in functional form can be used to complement the Monte Carlo numerical solution to this problem. We proceed to evaluate the performance of our nonparametric approach in the presence of estimation risk and model risk. We apply the algorithm to evolve trading strategies for 10 futures markets and 24 stock markets. We extend the results of recent studies that tested the efficient market hypothesis; these studies investigated whether market participants can find trading rules that use historical data as input that consistently produce abnormally high out-of-sample risk-adjusted returns (indicating that the markets are not efficient). Previous studies were limited to trading rules that returned simple buy/sell signals. Our approach is broader, allowing the study of trading strategies developed under a framework consistent with the standard financial economics model, with a trading strategy defined as the proportion of an investor's total wealth invested into the risky asset (that is, a strategy is a proportion rather than a simple buy/sell signal). The trading strategies evolved by our methodology demonstrate high out-of-sample risk-adjusted fitness for most futures markets, but strategies were produced for only a small fraction of periods because strategies were accepted only if they met criteria for in-sample fitness. Conversely, when our methodology was applied to the stock markets, it produced rules meeting the in-sample fitness criteria for most periods, but the rules were in general characterized by low out-of-sample risk-adjusted fitness. Because of the difficulty of evolving trading strategies that outperformed simple strategies, we conclude that the 10 futures markets and the 24 stock markets examined were adaptively efficient during the 1990's and the late 1980's.", notes = "Adaptive efficiency of futures and stock markets: Analysis and tests using a genetic programming approach NR19800", }

Genetic Programming entries for Stan Miles