P. Perner (Ed.): MLDM 2013, LNAI 7988, pp. 27–41, 2013.
© Springer-Verlag Berlin Heidelberg 2013
Dynamic-Radius Species-Conserving Genetic Algorithm
for the Financial Forecasting of Dow Jones Index Stocks
Michael Scott Brown, Michael J. Pelosi, and Henry Dirska
University of Maryland University College, Adelphi Maryland, USA
michaels.brown@faculty.umuc.edu,
mpelosi@maui.net,
dirska@nova.edu
Abstract. This research uses a Niche Genetic Algorithm (NGA) called Dynam-
ic-radius Species-conserving Genetic Algorithm (DSGA) to select stocks
to purchase from the Dow Jones Index. DSGA uses a set of training data to
produce a set of rules. These rules are then used to predict stock prices. DSGA
is an NGA that uses a clustering algorithm enhanced by a tabu list and radial
variations. DSGA also uses a shared fitness algorithm to investigate different
areas of the domain. This research applies the DSGA algorithm to training
data which produces a set of rules. The rules are applied to a set of testing
data to obtain results. The DSGA algorithm did very well in predicting stock
movement.
Keywords: Niche Genetic Algorithm, Genetic Algorithm, stock forecasting,
financial forecasting, classification, black-box investing.
1 Introduction
Forecasting the price movements of stocks is a difficult task. The possible financial
reward of picking the correct direction that a stock will move has created much inter-
est in developing systems to predict such behavior. Early work on formal financial
forecasting began in the early 1900’s [1] and continues to this day [2]. In the last 20
years a variety of techniques have been used to predict stock movement. These
include Genetic Algorithms (GAs), Neural Networks and other artificial intelligence
techniques.
A variety of methods use GAs and Genetic Programming (GP) to predict stock and
security movements [3-5]. These methods take different approaches. Some research
uses GAs and GPs to develop classification rules, while others use GAs in hybrid
approaches [2]. Much research has been done using these evolutionary approaches to
perform black-box investing.
This paper presents a new system for financial forecasting using a Niche Genetic
Algorithm (NGA). The presented research used an NGA and a set of financial data to
derive a set of classification rules that the research later applied to another set of data.
The training and test data each comes from a full quarter of stock prices from the