Ian Dempsey

An Introduction to Evolutionary Computation in Finance

The world of finance is an exciting and challenging environment. Recent years have seen an explosion in the application of computational intelligence methodologies in finance. In this article we provide an overview of some of these applications concentrating on those employing an evolutionary computation approach.

Adaptive Trading With Grammatical Evolution

This study reports on the performance of an on-line evolutionary automatic programming methodology for uncovering technical trading rules for the S&P 500 and Nikkei 225 indices. The system adopts a variable sized investment strategy based on the strength of the signals produced by the trading rules. Two approaches are explored, one using a single population of rules which is adapted over the lifetime of the data and another whereby a new population is created for each step across the time series. The results show profitable performance for the trading periods explored with clear advantages for an adaptive population of rules.

Constant creation in grammatical evolution

We present an investigation into constant creation in Grammatical Evolution (GE), a form of grammar-based Genetic Programming (GP). The methods for constant creation evaluated include digit Concatenation (Cat) and a grammatical version of ephemeral random constants called persistent random constants. Experiments conducted on a diverse range of benchmark problems uncover clear advantages for a digit Cat approach.

Introducing Grammar Based Extensions for Grammatical Evolution

This paper presents a series of extensions to standard grammatical evolution. These grammar-based extensions facilitate the exchange of knowledge between genotype and phenotype strings, thus establishing a better correlation between the search and solution spaces, typically separated in grammatical evolution. The results obtained illustrate the practical advantages of these extensions, both in terms of convenience and potential increase in performance.

Analysis of a digit concatenation approach to constant creation

This study examines the utility of employing digit concatenation, as distinct from the traditional expression based approach, for the purpose of evolving constants in Grammatical Evolution. Digit concatenation involves creating constants (either whole or real numbers) by concatenating digits to form a single value. The two methods are compared using three different problems, which are finding a static real constant, finding dynamic real constants, and a quadratic map, which on iteration generates a chaotic time-series. The results indicate that the digit concatenation approach results in a significant improvement in the best fitness obtained across all problems analysed here.

Meta-grammar constant creation with grammatical evolution by grammatical evolution

This study examines the utility of meta-grammar constant generation on a series of benchmark problems. The performance of the meta-grammar approach is compared to a grammar which incorporates grammatical ephemeral random constants, digit concatenation, and an expression based approach. It is found that the meta-grammar approach to constant creation is particularly beneficial on the dynamic problem instances in terms of the best fitness values achieved.

Grammatical Constant Creation

This study examines the utility of grammatical ephemeral random constants, and conducts an analysis of the preferences of evolutionary search when a number of different grammar based constant generation methods are provided with Grammatical Evolution. Three constant generation techniques are supplied, namely, grammatical ephemeral random constants, digit concatenation, and an expression based approach. A number of constant generation problems are tackled to analyse this approach, with results indicating a preference for both the digit concatenation and grammatical ephemeral random constants. The provision of different constant generation strategies allowing the evolutionary process to automatically determine which technique to adopt would, therefore, appear to be advantageous.

Investigations into Market Index Trading Models Using Evolutionary Automatic Programming

This study examines the potential of an evolutionary automatic programming methodology to uncover a series of useful technical trading rules for the US S&P stock index. Index values for the period 01/01/1991 to 01/10/1997 are used to train and test the evolved rules. A number of replacement strategies, and a novel approach to constant evolution are investigated. The findings indicate that the automatic programming methodology has much potential with the evolved rules making gains of approximately 13% over a 6 year test period.

Constant generation for the financial domain using grammatical evolution

This study reports the work to date on the analysis of different methodologies for constant creation with the aim of applying the most advantageous method to the dynamic real world problem of a live trading system. The methodologies explored here are Digit Concatenation and Grammatical Ephemeral Random Constants with clear advantages identified for a digit concatenation approach in combination with the ability to form new constants through their recombination within expressions.

Survey of EC in Dynamic Environments

The domain of EC in dynamic environments can be broken up into four main areas. This chapter is therefore divided up into four sections so as to comprehensively detail the prior art in the field as it stands. Section 3.2 focuses on the definition of dynamic environments. It examines the different types of changes that can occur as well as the features of each type of change that differentiate it from the other. Section 3.3 identifies the various approaches researchers have adopted in attempting to tune the EC paradigm to dynamic environments. Following from this, Section 3.4 describes the difficulties encountered when trying to measure the performance of an evolutionary algorithm placed in a dynamic environment. It also covers the types of metrics adopted to date. Section 3.5 looks at the different benchmark problems explored in the literature. The chapter then continues with a review of the findings of this survey in Section 3.6.