
Adaptive Strategy Selection for Concept Learning.
Authors: Spears, William M. and Diana F. Gordon (1991)
In Proceedings of the First International Workshop on Multistrategy Learning, 231246.
In this paper, we explore the use of genetic algorithms (GAs) to construct a system called GABIL that continually learns and refines concept classification rules from its interaction with the environment. The performance of this system is compared with that of two other concept learners (NEWGEM and C4.5) on a suite of target concepts. From this comparison, we identify strategies responsible for the success of these concept learners. We then implement a subset of these strategies within GABIL to produce a multistrategy concept learner. Finally, this multistrategy concept learner is further enhanced by allowing the GAs to adaptively select the appropriate strategies.
Fulltext: PostScript, PDF, or PS compressed

Learning Concept Classification Rules Using Genetic Algorithms.
Authors: De Jong, Kenneth A. and William M. Spears (1991).
In Proceedings of the Int'l Joint Conference on Artificial Intelligence, 651656.
In this paper we explore the use of an adaptive search technique (genetic algorithms) to construct a system GABIL which continually learns and refines concept classification rules from its interaction with the environment. The performance of the system is measured on a set of concept learning problems and compared with the performance of two existing systems: ID5R and C4.5. Preliminary results support that, despite minimal system bias, GABIL is an effective concept learner and is quite competitive with ID5R and C4.5 as the target concept increases in complexity.
Fulltext: PostScript, PDF, or PS compressed

Using Genetic Algorithms for Supervised Concept Learning.
Authors: Spears, William M. and Kenneth A. De Jong (1990)
In Proceedings of the IEEE AI Tools Conference, 335341.
Genetic Algorithms (GAs) have traditionally been used for nonsymbolic learning tasks. In this paper we consider the application of a GA to a symbolic learning task, supervised concept learning from examples. A GA concept learner (GABL) is implemented that learns a concept from a set of positive and negative examples. GABL is run in a batchincremental mode to facilitate comparison with an incremental concept learner, ID5R. Preliminary results support that, despite minimal system bias, GABL is an effective concept learner and is quite competitive with ID5R as the target concept increases in complexity.
Fulltext: PostScript, PDF, or PS compressed, PDF book chapter

Using Genetic Algorithms for Concept Learning.
Authors: De Jong, Kenneth A.. William M. Spears, and Diana F. Gordon (1993)
Journal: In Machine Learning, vol. 13, #2/3, 161188.
In this paper, we explore the use of genetic algorithms (GAs) as a key element in the design and implementation of robust concept learning systems. We describe and evaluate a GAbased system called GABIL that continually learns and refines concept classification rules from its interaction with the environment. The use of GAs is motivated by recent studies showing the effects of various forms of bias built into different concept learning systems, resulting in systems that perform well on certain concept classes (generally, those well matched to the biases) and poorly on others. By incorporating a GA as the underlying adaptive search mechanism, we are able to construct a concept learning system that has a simple, unified architecture with several important features. First, the system is surprisingly robust even with minimal bias. Second, the system can be easily extended to incorporate traditional forms of bias found in other concept learning systems. Finally, the architecture of the system encourages explicit representation of such biases and, as a result, provides for an important additional feature: the ability to dynamically adjust system bias. The viability of this approach is illustrated by comparing the performance of GABIL with that of four other more traditional concept learners (AQ14, C4.5, ID5R, and IACL) on a variety of target concepts. We conclude with some observations about the merits of this approach and about possible extensions.
Fulltext: PostScript, PDF, or PS compressed

Optimization of molecular clusters configurations using a Genetic Algorithm
Authors: Stoyan Pisov, A. Proykova
Comments: 4 pages, 4 figures (eps), LaTex, mbook.cls
Subjclass: Atomic and Molecular Clusters; Computational Physics
Journalref: Meeting in Physics, v.3 (2002) p.40, Heron PressSofia, ISBN 9545801204
We present a genetic algorithm developed (GA) to optimize molecular AF_6 cluster configurations with respect to their energy. The method is based on the Darvin's evolutionary theory: structures with lowest energies survive in a system of fixed number of clusters. Two existing structures from a given population are combined in a special way to produce a new structure (child) which is kept if its energy is lower than the highest energy in the ensemble. To keep the population constant we reject the structure with the highest energy. This algorithm gives a better result than the optimization techniques used previously. Using the GA we have found a new structure corresponding to the (seemingly) global minimum. The most important result is that the new structure is detected only if the molecular cluster contains more than a critical number of molecules.
Fulltext: PostScript, PDF, or Other formats

Aging, double helix and small world property in genetic algorithms
Authors: Marek W. Gutowski
Comments: Submitted to the workshop on evolutionary algorithms, Krakow (Cracow), Poland, Sept. 30, 2002, 6 pages, no figures, LaTeX 2.09 requires kaeog.sty (included)
Subjclass: Neural and Evolutionary Computing; Data Structures and Algorithms; Data Analysis, Statistics and Probability
ACMclass: F.2.1; G.1.6; I.1.2
Over a quarter of century after the invention of genetic algorithms and miriads of their modifications, as well as successful implementations, we are still lacking many essential details of thorough analysis of it's inner working. One of such fundamental questions is: how many generations do we need to solve the optimization problem? This paper tries to answer this question, albeit in a fuzzy way, making use of the double helix concept. As a byproduct we gain better understanding of the ways, in which the genetic algorithm may be fine tuned.
Fulltext: PostScript, PDF, or Other formats

Genetic Algorithms in TimeDependent Environments
Authors: Christopher Ronnewinkel, Claus O. Wilke, Thomas Martinetz
Comments: 24 pages, 14 figures, submitted to the 2nd EvoNet Summerschool
Subjclass: Biological Physics; Neural and Evolutionary Computing; Adaptation and SelfOrganizing Systems
The influence of timedependent fitnesses on the infinite population dynamics of simple genetic algorithms (without crossover) is analyzed. Based on general arguments, a schematic phase diagram is constructed that allows one to characterize the asymptotic states in dependence on the mutation rate and the time scale of changes. Furthermore, the notion of regular changes is raised for which the population can be shown to converge towards a generalized quasispecies. Based on this, error thresholds and an optimal mutation rate are approximately calculated for a generational genetic algorithm with a moving needleinthehaystack landscape. The so found phase diagram is fully consistent with our general considerations.
Fulltext: PostScript, PDF, or Other formats

Global geometry optimization of clusters using a growth strategy optimized by a genetic algorithm
Author: Bernd Hartke (Department of Theoretical Chemistry, University of Stuttgart, Germany)
Comments: accepted by Chem.Phys.Letters; 10 pages text, plus 3 pages for Title, abstract, and figure caption; figures 1a and 1b
Subjclass: Chemical Physics
A new strategy for global geometry optimization of clusters is presented. Important features are a restriction of search space to favorable nearestneighbor distance ranges, a suitable cluster growth representation with diminished correlations, and easy transferability of the results to larger clusters. The strengths and possible limitations of the method are demonstrated for Si10 using an empirical potential.
Fulltext: PostScript, PDF, or Other formats

Genetic Algorithms for Finding Polynomial Orderings
Author: J¨urgen Giesl, Fernando Esponda, and Stephanie Forrest
LuFG Informatik II, RWTH Aachen, Ahornstr. 55, 52074 Aachen, Germany. Computer Science Department, University of New Mexico, Albuquerque, NM 87131, USA
(2001)
Polynomial orderings are a wellknown method to prove termination of term rewriting systems. However, for an automation of this method, the crucial point is to nd suitable coefficients by machine. We present a novel approach for solving this problem by applying genetic algorithms.
Fulltext: PostScript, PDF

Tinkering with Genetic Algorithms: Forecasting and Data Mining in Finance and Economics
Author: George G. Szpiro
Israeli Center for Academic Studies (a liated with the University of Manchester) Kiriat Ono, Israel
Undergraduate Summer Workshop on AgentBased Computational Economics
July1 Sep 30, 2000. Taiwan
In two previous papers [13,14] genetic algorithms were presented that permit the search for dependencies among sets of data (univariate or multivariate timeseries, or crosssectional observations). These algorithms modeled after  genetic theories and Darwinian concepts, such as natural selection and survival of the fittest permit the discovery of equations, in symbolic form, that recreate or, at least, mimic the datagenerating process. This paper discusses some of the computational issues and difficulties that may arise when the genetic algorithm is applied, and suggests ways to improve the algorithm s performance.
Fulltext: PDF

Can We Believe That Genetic Algorithms Would Help without Actually Seeing them Work in Financial Data Mining?: Part 1, Theoretical Foundations
Author: ShuHeng Chen
National Chengchi University, Taipei, Taiwan 11623
Undergraduate Summer Workshop on AgentBased Computational Economics
July1 Sep 30, 2000. Taiwan
Genetic algorithms (GAs) have, time and again, shown some promising features when applied to optimization problems. The theoretical foundations of these successful applications however are rather limited, in particular, when the problem embodies a dynamic rather than a static landscape. In this paper, dynamic landscapes are treated as random variables, and we sort out a few stochastic properties which may impinge upon the performance of GAs in financial data mining. Several tests of these properties are then proposed and a priori evaluation of the potential of GAs can be made based on these proposed tests.
Fulltext: PDF

STRUCTURE LEARNING OF BAYESIAN NETWORKS FROM DATABASES BY GENETIC ALGORITHMS
Author: Jérôme HABRANT
APPLICATION TO TIME SERIES PREDICTION IN FINANCE
Ecole Nationale Supérieure des Mines de St Etienne 158, Cours Fauriel 42023 St Etienne Cedex France
Key words: Genetic algorithms, Bayesian networks, Learning from a database of cases
This paper outlines a genetic algorithm based method for constructing bayesian networks from databases. Our method permits the generation of a complete structure if there is no expert for the domain studied. Also it allows taking advantage of the knowledge about the domain by specifying connections in the network. To test our method, we applied it to time series prediction in finance with 5 shares. We experimented 3 different genetic algorithms: first, we used classical syntactical genetic operators, second we add 2 highlevel genetic operators by taking the semantic of the structures into consideration, and third, we add a last powerful operator. Furthermore, we studied 3 constraints on the structures: by assuming an ordering between the nodes, by releasing the ordering assumption and by forcing the structures to use all available information to build the forecasts. For each of the 3 genetic algorithms and the 3 constraints, we present our results concerning the genetic algorithms convergence and the predictive power of the best structures obtained. Our results are encouraging.
Fulltext: PDF

An Overview of Genetic Algorithms
Authors: D. Beasley, D.R. Bull, R.R. Martin
No part of this article may be reproduced for commercial purposes.
Genetic Algorithms (GAs) are adaptive methods which may be used to solve search and optimization problems. They are based on the genetic processes of biological organisms. Over many generations, natural populations evolve according to the principles of natural selection and "survival of the fittest", first clearly stated by Charles Darwin in The Origin of Species. By mimicking this process, genetic algorithms are able to "evolve" solutions to real world problems, if they have been suitably encoded. For example, GAs can be used to design bridge structures, for maximum strength/weight ratio, or to determine the least wasteful layout for cutting shapes from cloth. They can also be used for online process control, such as in a chemical plant, or load balancing on a multiprocessor computer system.
Fulltext: GZiped PostScript Part 1a, GZiped PostScript Part 1b

MATHEMATICS OF FINANCE  TAKING CALCULATED RISKS
Author: Peter Baring
Artificial intelligence models may be used to improve performance of information retrieval (IR) systems and the genetic algorithms (GAs) are an example of such a model. This paper presents an application of GAs as a relevance feedback method aiming to improve the document representation and indexing. In this particular form of GAs, various document descriptions compete with each other and a better collection indexing is sought through reproduction, crossover and mutation operations. In this...
Fulltext: PDF

Large Population or Many Generations for Genetic Algorithms? Implications in Information Retrieval
Author: Dana Vrajitoru
University of Neuch^atel, Computer Science Department, Pierre aMazel 7, 2000 Neuch^atel, Switzerland
Artificial intelligence models may be used to improve performance of information retrieval (IR) systems and the genetic algorithms (GAs) are an example of such a model. This paper presents an application of GAs as a relevance feedback method aiming to improve the document representation and indexing. In this particular form of GAs, various document descriptions compete with each other and a better collection indexing is sought through reproduction, crossover and mutation operations. In this paradigm, we are searching for the optimal balance between two genetic parameters: the population size and the number of generations. We try to discover the optimal parameter choice both by experiments using the CACM and CISI collections, and by a theoretical analysis providing explanation of the experimental results. The general conclusion tends to be that larger populations have better chance of significantly improving the effectiveness of retrieval.
Fulltext: PostScript, PDF, or GZiped PS

Genetic Algorithms In Finance
Lecturer: Clarence Tan
Research Project
Daphne S. L. Chou 971000318 20 November, 1997
BOND UNIVERSITY School of Information Technology COMP723 Research in Artificial Intelligence
Genetic algorithms (GAs) is one of the most fascinating areas of study in AI. Genetic algorithms are adaptive methods for solving search for solution and optimization problems. The search strategy of genetic algorithms is based on biological evolution. The algorithm simulates natural selection by means of repeatedly evolving a population of candidate solutions in order to find the optimal solution. In the last few years, genetic algorithms have been applied in a wide range of realworld fields, including machine learning, financial forecasting, medicine, and so on. Genetic algorithm procedures are said to be good and robust at searching nearoptimal or attractive solutions.
Fulltext: PDF

Statistical Properties of Genetic Learning in a Model of Exchange Rate
Authors: Jasmina Arifovic Ramazan Gencay
August 1998
Department of Economics, Simon Fraser University, Burnaby, British Columbia, V5A 1N6, Canada. We thank Cars Hommes and two anonymous referees for their comments in earlier versions of this paper. Jasmina Arifovic gratefully acknowledges ¯financial support from the Social Sciences and Humanities Research Council of Canada.
We study statistical properties of the time series of the exchange rate data generated in the environment where agents update their savings and portfolio decisions using the genetic algorithm. The genetic algorithm adaptation takes place within an overlapping generations model with two currencies and the freetrade, flexible exchange rate system. The theoretical model implies a constant exchange rate under the perfect foresight assumption. Under the genetic algorithm learning, the model's equilibrium dynamics is not constant but exhibits bounded oscillations. The time series analysis of the data indicates that the dynamics of the exchange rate returns is chaotic. Outofequilibrium inequality of rates of return on two currencies prompts the genetic algorithm agents to take advantage of the arbitrage opportunities by increasing the amount of the currency with higher rate of return in their portfolios. This pro¯t seeking results in chaotic patterns of the exchange rate series.
Fulltext: PDF

The Gene Expression Messy Genetic Algorithm For Financial Applications (1996)
Authors: Hillol Kargupta Kevin Buescher
Computational Science Division. Los Alamos National Laboratory, Los alamos, NM, USA
Fulltext: PostScript, PDF, or GZiped PS

Improving Technical Analysis Predictions: An Application of Genetic Programming
Authors: JIN LI AND EDWARD P.K. TSANG
Proceedings, Florida Artificial Intelligence Research Symposium, USA, 1999
Department of Computer Science, University of Essex, Wivenhoe Park, Colchester, United Kingdom
KEY WORDS: finance; prediction; genetic programming; investment, technical rules
Recent studies in finance domain suggest that technical analysis may have merit to predictability of stock. Technical rules are widely used for market assessment and timing. For example, moving average rules are used to make "buy" or "sell" decisions at each day. In this paper, to explore the potential prediction power of technical analysis, we present a genetic programming based system FGP (Financial Genetic Programming), which specialises in taking some well known technical rules and adapting them to prediction problems. FGP uses the power of genetic programming to generate decision trees through efficient combination of technical rules with selfadjusted thresholds. The generated rules are more suitable for the prediction problem at hand. FGP was tested extensively on historical S&P 500 data through a specific prediction problem. Preliminary results show that it outperforms commonly used, nonadaptive, individual technical rules with respect to prediction accuracy and annualized rate of return over two different outofsample test periods (three and a half year in each period).
Fulltext: PDF

A HYBRID GENETICQUANTITATIVE METHOD FOR RISKRETURN OPTIMIZATION OF CREDIT PORTFOLIOS
Authors: Frank Schlottmann Detlef Seese
Institute AIFB University Karlsruhe
Key words: Credit risk, Portfolio credit risk model, Portfolio optimization, Genetic Algorithm, CreditValueatRisk, Economic capital, Regulatory capital
This paper proposes a new combination of quantitative models and Genetic Algorithms for the task of optimizing credit portfolios. Currently, quantitative portfolio credit risk models are used to calculate portfolio risk figures, e. g. expected losses, unexpected losses and risk contributions. Usually, this information is used for optimizing the riskreturn profile of the portfolio. We show that gradientlike optimisation methods based on risk contributions can lead to inefficient portfolio structures. To avoid this local optima problem, our optimisation method combines quantitative model features with Genetic Algorithms. The hybrid approach in this paper consists of a task specific Genetic Algorithm that uses special variation operators reflecting portfolio credit risk model knowledge. The method presented here is compatible with any model providing a loss or profit/loss distribution for credit portfolios, e. g. CreditMetrics, CreditRisk+, Wilson s model, and others. As a consequence, it can be used with any risk measure based on such profitloss distributions like CreditValueatRisk, Expected Shortfall etc. We show how different additional constraints like economic and/or regulatory capital limits can be included in the optimisation process. The results of a test series with different portfolio sizes and structures in a CreditRisk+ General Sector Analysis model framework running on a standard Personal Computer are provided within this paper. They indicate that the hybrid Genetic Algorithm leads to better convergence than a standard Genetic Algorithm approach while not suffering from the local optima problem, and calculates efficient portfolio structures within reasonable time.
Fulltext: PDF