A fundamental objective of Artificial Intelligence (AI) is the creation of in­ telligent computer programs. In more modest terms AI is simply con­ cerned with expanding the repertoire of computer applications into new domains and to new levels of efficiency. The motivation for this effort comes from many sources. At a practical level there is always a demand for achieving things in more efficient ways. Equally, there is the technical challenge of building programs that allow a machine to do something a machine has never done before. Both of these desires are contained within AI and both provide the inspirational force behind its development. In terms of satisfying both of these desires there can be no better example than machine learning. Machines that can learn have an in-built effi­ ciency. The same software can be applied in many applications and in many circumstances. The machine can adapt its behaviour so as to meet the demands of new, or changing, environments without the need for costly re-programming. In addition, a machine that can learn can be ap­ plied in new domains with the genuine potential for innovation. In this sense a machine that can learn can be applied in areas where little is known about possible causal relationships, and even in circumstances where causal relationships are judged not to exist. This last aspect is of major significance when considering machine learning as applied to fi­ nancial forecasting.

Inhalt

1 From Learning Systems to Financial Modelling.- 1.1 Introduction.- 1.2 Adaptive Systems and Financial Modelling.- 1.2.1 Financial Modelling: The Efficient Markets Hypothesis.- 1.2.2 Learning Systems.- 1.2.3 Technical Issues.- 1.3 Time Series Analysis.- 1.3.1 Fundamentals of Time Series Forecasting and Learning.- 1.4 Brief History of Neural Networks.- 1.4.1 The Development of Neural Net Techniques.- 1.4.2 More Recent Issues.- 1.5 Book Overview.- 1.5.1 Research Objectives.- 1.5.2 Book Structure.- 1.6 Summary.- 2 Adaptive Systems and Financial Modelling.- 2.1 Financial Modelling.- 2.2 The Problems with Financial Modelling.- 2.2.1 Fuzzy Rationality and Uncertainty.- 2.2.2 Efficient Markets and Price Movement.- 2.3 Evidence Against the Efficiency Hypothesis.- 2.4 An Adaptive Systems Approach.- 2.5 Neural Nets and Financial Modelling.- 2.5.1 Comparisons Between Neural Nets and Other Time Series Methods.- 2.6 Genetic Algorithms in Finance.- 2.6.1 The Genetic Algorithm Search Technique.- 2.6.2 Applications of Genetic Algorithms.- 2.7 Summary.- 3 Feed-Forward Neural Network Modelling.- 3.1 Neural Net Search.- 3.2 MLP Training: The Model.- 3.3 MLP: Model Parameters.- 3.4 The Data.- 3.5 MLP: Training Parameters.- 3.5.1 Architecture.- 3.5.2 Activation Function.- 3.5.3 Learning Rules, Batch and On-Line Training.- 3.6 Network Performance.- 3.6.1 Convergence.- 3.6.2 Network Validation and Generalisation.- 3.6.3 Automated Validation.- 3.7 Summary.- 4 Genetic Algorithms.- 4.1 Using Genetic Algorithms.- 4.2 Search Algorithms.- 4.2.1 The GA Search Process: The Simple GA.- 4.2.2 Schema Analysis.- 4.2.3 Building Blocks Under Review.- 4.3 GA Parameters.- 4.3.1 The Shape of Space.- 4.3.2 Population Encodings.- 4.3.3 Crossover, Selection, Mutation and Populations.- 4.4 A Strategy for GA Search: Transmutation.- 4.4.1 Five New Algorithms: Morphic GAs (MGAs).- 4.5 Summary.- 5 Hypothesising Neural Nets.- 5.1 System Objectives.- 5.2 Hypothesising Neural Network Models.- 5.3 Occam's Razor and Network Architecture.- 5.3.1 Existing Regulisation and Pruning Methods.- 5.3.2 Why use Occam's Razor?.- 5.4 Testing Occam's Razor.- 5.4.1 Generating Time Series.- 5.4.2 Artificial Network Generation (ANG).- 5.4.3 ANG Results.- 5.4.4 Testing Architectures.- 5.5 Strategies using Occam's Razor.- 5.5.1 Minimally Descriptive Nets.- 5.5.2 Network Model.- 5.5.3 Network Regression Pruning (NRP).- 5.5.4 Results of NRP on ANG Series.- 5.5.5 Interpretation of the Pruning Error Profiles.- 5.5.6 Determining Topologies.- 5.6 Validation.- 5.7 GA-NN Hybrids: Representations.- 5.7.1 Fitness Measures for GA-NN Hybrids.- 5.7.2 Neural Networks and GAs: Fitness Measure for Generalisation.- 5.8 Summary.- 6 Automating Neural Net Time Series Analysis.- 6.1 System Objectives.- 6.2 ANTAS.- 6.2.1 Stage I: Primary Modelling.- 6.2.2 Stage II: Secondary Modelling.- 6.2.3 Stage III: System Modelling.- 6.3 Primary Modelling.- 6.3.1 Automating the use of Neural Nets.- 6.3.2 GA Rule-Based Modelling.- 6.4 Secondary Modelling.- 6.4.1 Generating Secondary Models.- 6.4.2 Model Integration.- 6.4.3 Model Performance Statistics.- 6.5 Validation Modules.- 6.6 Control Flow.- 6.6.1 Neural Net Control.- 6.6.2 GA Control.- 6.7 Summary.- 7 The Data: The Long Gilt Futures Contract.- 7.1 The Long Gilt Futures Contract.- 7.2 The LGFC Data.- 7.2.1 Time Series Construction.- 7.3 Secondary Data.- 7.4 Data Preparation.- 7.4.1 LGFC Data Treatment.- 7.4.2 Using Moving Averages.- 7.5 Data Treatment Modules.- 7.5.1 Moving Average Modules.- 7.6 Efficient Market Hypothesis and the LGFC.- 7.7 Summary.- 8 Experimental Results.- 8.1 Experimental Design.- 8.2 Phase I - Primary Models.- 8.2.1 NN Hypothesis Modules (Phase I).- 8.2.2 Results for GA-NN Module.- 8.2.3 In-Sample Testing and Validation of the 15-4 Neural Network.- 8.3 GA-RB Module and Combined Validation.- 8.4 Phase II - Secondary GA-RB Models.- 8.4.1 Secondary Model Control Module.- 8.5 Phase III - Validation and Simulated Live Trading.- 8.6 Controls: Analysis of ANTAS.- 8.6.1 Choosing a Network Architecture.- 8.6.2 GA Control Tests.- 8.6.3 Second Order Modelling.- 8.7 ANTAS: Conclusions.- 8.8 Summary.- 9 Summary, Conclusions and Future Work.- 9.1 Motivations.- 9.2 Objectives: Neural Networks and Learning.- 9.3 Book Outline and Results.- 9.3.1 Morphic Genetic Algorithms using Base Changes.- 9.3.2 Artificial Network Generation.- 9.3.3 Network Regression Pruning.- 9.3.4 ANTAS and the Long Gilt Futures Contract.- 9.3.5 Results.- 9.4 Conclusions.- 9.5 Future Work.- Appendices.- A Test Functions.- B ANTAS Outline Code.- C ANTAS Results.- References.