Phenotype-Object Programming, Phenotype-Array Datatype, and an Evolutionary Signal-Ensemble FX Trading Model

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

@InProceedings{nacaskul:1997:pop,
  author =       "Poomjai Nacaskul",
  title =        "Phenotype-Object Programming, Phenotype-Array
                 Datatype, and an Evolutionary Signal-Ensemble FX
                 Trading Model",
  booktitle =    "ET'97 Theory and Application of Evolutionary
                 Computation",
  year =         "1997",
  editor =       "Chris Clack and Kanta Vekaria and Nadav Zin",
  pages =        "95--108",
  address =      "University College London, UK",
  month =        "15 " # dec,
  keywords =     "genetic algorithms, genetic programming",
  abstract =     "We set out to optimise a financial trading model which
                 uses an ensemble of time-series trend indicating
                 signal-models. The optimisation search is combinatorial
                 over the combination of signal-model classes as well as
                 parametric over the parameterisation of individual
                 signal-model objects. Because the optimisation
                 objective (net gain from simulated transactions against
                 a historical price series) is not differentiable w.r.t.
                 our trading model space, we look to evolutionary
                 optimisation [EO] methodologies [Fogel, 1994], e.g.
                 Genetic Algorithm [GA] [Holland, 1975/92], which rely
                 on direct solution performance evaluation. However,
                 because the multi-stage nature of our solution space
                 prohibits stochastic-evolutionary convergence, we need
                 to engineer a new EO paradigm to implement
                 combinatorial and parametric search processes
                 concurrently. This, we accomplish by exploiting the
                 inherently object-oriented [OO] [Booch, 1994;
                 Stroustrup, 1991] nature of an EO algorithm and of the
                 combinatorial-parametric solutions. We propose
                 Phenotype-Object Programming [POP] as a generalised OO
                 model and implementation of an EO algorithm and
                 Phenotype-Array Datatype [PAD] as a generalised OO
                 model and implementation of a combinatorial-parametric
                 solution [Nacaskul, 1997]. We apply this to our
                 signal-ensemble trading model and discuss experimental
                 results on DEM/JPY and USD/DEM data.",
  notes =        "http://www.cs.ucl.ac.uk/isrg/et97/",
}

Genetic Programming entries for Poomjai Nacaskul

Citations