Applications of Computational Intelligence in Data-Driven Trading
1. Edition December 2019
304 Pages, Hardcover
Wiley & Sons Ltd
ISBN:
978-1-119-55050-1
John Wiley & Sons
Further versions
"Life on earth is filled with many mysteries, but perhaps the most challenging of these is the nature of Intelligence."
- Prof. Terrence J. Sejnowski, Computational Neurobiologist
The main objective of this book is to create awareness about both the promises and the formidable challenges that the era of Data-Driven Decision-Making and Machine Learning are confronted with, and especially about how these new developments may influence the future of the financial industry.
The subject of Financial Machine Learning has attracted a lot of interest recently, specifically because it represents one of the most challenging problem spaces for the applicability of Machine Learning. The author has used a novel approach to introduce the reader to this topic:
* The first half of the book is a readable and coherent introduction to two modern topics that are not generally considered together: the data-driven paradigm and Computational Intelligence.
* The second half of the book illustrates a set of Case Studies that are contemporarily relevant to quantitative trading practitioners who are dealing with problems such as trade execution optimization, price dynamics forecast, portfolio management, market making, derivatives valuation, risk, and compliance.
The main purpose of this book is pedagogical in nature, and it is specifically aimed at defining an adequate level of engineering and scientific clarity when it comes to the usage of the term "Artificial Intelligence," especially as it relates to the financial industry.
The message conveyed by this book is one of confidence in the possibilities offered by this new era of Data-Intensive Computation. This message is not grounded on the current hype surrounding the latest technologies, but on a deep analysis of their effectiveness and also on the author's two decades of professional experience as a technologist, quant and academic.
About the Author xvii
Acknowledgments xix
About the Website xxi
Introduction xxiii
Motivation xxiv
Target Audience xxvi
Book Structure xxvii
1 The Evolution of Trading Paradigms 1
1.1 Infrastructure-Related Paradigms in Trading 1
1.1.1 Open Outcry Trading 2
1.1.2 Advances in Communication Technology 2
1.1.3 The Digital Revolution in the Financial Markets 3
1.1.4 The High-Frequency Trading Paradigm 5
1.1.5 Blockchain and the Decentralization of Markets 6
1.2 Decision-Making Paradigms in Trading 7
1.2.1 Discretionary Trading 8
1.2.2 Systematic Trading 8
1.2.3 Algorithmic Trading 9
1.3 The New Paradigm of Data-Driven Trading 11
References 14
2 The Role of Data in Trading and Investing 15
2.1 The Data-Driven Decision-Making Paradigm 15
2.2 The Data Economy is Fueling the Future 17
2.2.1 The Value of Data - Data as an Asset 18
2.3 Defining Data and Its Utility 20
2.4 The Journey from Data to Intelligence 24
2.5 The Utility of Data in Trading and Investing 30
2.6 The Alternative Data and Its Use in Trading and Investing 34
References 36
3 Artificial Intelligence - Between Myth and Reality 39
3.1 Introduction 39
3.2 The Evolution of AI 41
3.2.1 Early History 41
3.2.2 The Modern AI Era 43
3.2.3 Important Milestones in the Development of AI 44
3.2.4 Projections for the Immediate Future 48
3.2.5 Meta-Learning - An Exciting New Development 49
3.3 The Meaning of AI - A Critical View 51
3.4 On the Applicability of AI to Finance 54
3.4.1 Data Stationarity 57
3.4.2 Data Quality 58
3.4.3 Data Dimensionality 59
3.5 Perspectives and Future Directions 60
References 62
4 Computational Intelligence - A Principled Approach for the Era of Data Exploration 63
4.1 Introduction to Computational Intelligence 63
4.1.1 Defining Intelligence 63
4.1.2 What is Computational Intelligence? 64
4.1.3 Mapping the Field of Study 66
4.1.4 Problems vs. Tools 68
4.1.5 Current Challenges 69
4.1.6 The Future of Computational Intelligence 70
4.1.7 Examples in Finance 71
4.2 The PAC Theory 72
4.2.1 The Probably Approximately Correct Framework 73
4.2.2 Why AI is a Very Lofty Goal to Achieve 75
4.2.3 Examples of Ecorithms in Finance 78
4.3 Technology Drivers Behind the ML Surge 81
4.3.1 Data 82
4.3.2 Algorithms 82
4.3.3 Hardware Accelerators 82
References 84
5 How to Apply the Principles of Computational Intelligence in Quantitative Finance 87
5.1 The Viability of Computational Intelligence 87
5.2 On the Applicability of CI to Quantitative Finance 91
5.3 A Brief Introduction to Reinforcement Learning 94
5.3.1 Defining the Agent 96
5.3.2 Model-Based Markov Decision Process 98
5.3.3 Model-Free Reinforcement Learning 101
5.4 Conclusions 104
References 104
6 Case Study 1: Optimizing Trade Execution 107
6.1 Introduction to the Problem 107
6.1.1 On Limit Orders and Market Microstructure 109
6.1.2 Formulation of Base-Line Strategies 111
6.1.3 A Reinforcement Learning Formulation for the Optimized Execution Problem 112
6.2 Current State-of-the-Art in Optimized Trade Execution 114
6.3 Implementation Methodology 116
6.3.1 Simulating the Interaction with the Market Microstructure 116
6.3.2 Using Dynamic Programming to Optimize Trade Execution 118
6.3.3 Using Reinforcement Learning to Optimize Trade Execution 119
6.4 Empirical Results 122
6.4.1 Application to Equities 122
6.4.2 Using Private Variables Only 123
6.4.3 Using Both Private and Market Variables 123
6.4.4 Application to Futures 124
6.4.5 Another Example 126
6.5 Conclusions and Future Directions 127
6.5.1 Further Research 127
References 128
7 Case Study 2: The Dynamics of the Limit Order Book 129
7.1 Introduction to the Problem 129
7.1.1 The New Era of Prediction 130
7.1.2 New Challenges 131
7.1.3 High-Frequency Data 132
7.2 Current State-of-the-Art in the Prediction of Directional Price Movement in the LOB 133
7.2.1 The Contrarians 136
7.3 Using Support Vector Machines and Random Forest Classifiers for Directional Price Forecast 138
7.3.1 Empirical Results 139
7.4 Studying the Dynamics of the LOB with Reinforcement Learning 141
7.4.1 Empirical Results 142
7.4.2 Conclusions 144
7.5 Studying the Dynamics of the LOB with Deep Neural Networks 145
7.5.1 Results 148
7.6 Studying the Dynamics of the Limit Order Book with Long Short-Term Memory Networks 149
7.6.1 Empirical Results 152
7.6.2 Conclusions 153
7.7 Studying the Dynamics of the LOB with Convolutional Neural Networks 153
7.7.1 Empirical Results 155
7.7.2 Conclusions 156
References 157
8 Case Study 3: Applying Machine Learning to Portfolio Management 159
8.1 Introduction to the Problem 159
8.1.1 The Problem of Portfolio Diversification 160
8.2 Current State-of-the-Art in Portfolio Modeling 161
8.2.1 The Classic Approach 161
8.2.2 The ML Approach 162
8.3 A Deep Portfolio Approach to Portfolio Optimization 163
8.3.1 Autoencoders 164
8.3.2 Methodology - The Four-Step Algorithm 166
8.3.3 Results 167
8.4 A Q-Learning Approach to the Problem of Portfolio Optimization 167
8.4.1 Problem Statement 168
8.4.2 Methodology 169
8.4.3 The Deep Q-Learning Algorithm 169
8.4.4 Results 170
8.5 A Deep Reinforcement Learning Approach to Portfolio Management 170
8.5.1 Methodology 170
8.5.2 Data 171
8.5.3 The RL Setting: Agent, Environment, and
Policy 172
8.5.4 The CNN Implementation 172
8.5.5 The RNN and LSTM Implementations 172
8.5.6 Results 173
References 174
9 Case Study 4: Applying Machine Learning to Market Making 175
9.1 Introduction to the Problem 175
9.2 Current State-of-the-Art in Market Making 177
9.3 Applications of Temporal-Difference RL in Market Making 180
9.3.1 Methodology 180
9.3.2 The Simulator 181
9.3.3 Market Making Agent Specification 182
9.3.4 Empirical Results 185
9.4 Market Making in High-Frequency Trading Using RL 189
9.4.1 Methodology 190
9.4.2 Experimental Setting 191
9.4.3 Results and Conclusions 192
9.5 Other Research Studies 192
References 193
10 Case Study 5: Applications of Machine Learning to Derivatives Valuation 197
10.1 Introduction to the Problem 197
10.1.1 Problem Statement and Research Questions 199
10.2 Current State-of-the-Art in Derivatives Valuation by Applying ML 200
10.2.1 The Beginnings: 1992-2004 201
10.2.2 The Last Decade 202
10.3 Using Deep Learning for Valuation of Derivatives 204
10.3.1 Implementation Methodology 205
10.3.2 Empirical Results 207
10.3.3 Conclusions and Future Directions 208
10.3.4 Other Research Studies 208
10.4 Using RL for Valuation of Derivatives 210
10.4.1 Using a Simple Markov Decision Process 210
10.4.2 The Q-Learning Black-Scholes Model (QLBS) 212
References 214
11 Case Study 6: Using Machine Learning for Risk Management and Compliance 217
11.1 Introduction to the Problem 217
11.1.1 Challenges 218
11.1.2 The Problem 219
11.2 Current State-of-the-Art for Applications of ML to Risk Management and Compliance 219
11.2.1 Credit Risk 219
11.2.2 Market Risk 220
11.2.3 Operational Risk 221
11.2.4 Regulatory Compliance Risk and RegTech 222
11.2.5 Current Challenges and Future Directions 223
11.3 Machine Learning in Credit Risk Modeling 224
11.3.1 Data 225
11.3.2 Models 225
11.3.3 Results 226
11.4 Using Deep Learning for Credit Scoring 227
11.4.1 Introduction 227
11.4.2 Deep Belief Networks and Restricted Boltzmann Machines 228
11.4.3 Empirical Results 230
11.5 Using ML in Operational Risk and Market Surveillance 230
11.5.1 Introduction 230
11.5.2 An ML Approach to Market Surveillance 232
11.5.3 Conclusions 233
References 233
12 Conclusions and Future Directions 237
12.1 Concluding Remarks 237
12.2 The Paradigm Shift 239
12.2.1 Mathematical Models vs. Data Inference 240
12.3 De-Noising the AI Hype 243
12.3.1 Why Intellectual Honesty Should Not Be Abandoned 244
12.4 An Emerging Engineering Discipline 245
12.4.1 The Problem 246
12.4.2 The Market 246
12.4.3 A Possible Solution 246
12.5 Future Directions 247
References 248
Index 249
Acknowledgments xix
About the Website xxi
Introduction xxiii
Motivation xxiv
Target Audience xxvi
Book Structure xxvii
1 The Evolution of Trading Paradigms 1
1.1 Infrastructure-Related Paradigms in Trading 1
1.1.1 Open Outcry Trading 2
1.1.2 Advances in Communication Technology 2
1.1.3 The Digital Revolution in the Financial Markets 3
1.1.4 The High-Frequency Trading Paradigm 5
1.1.5 Blockchain and the Decentralization of Markets 6
1.2 Decision-Making Paradigms in Trading 7
1.2.1 Discretionary Trading 8
1.2.2 Systematic Trading 8
1.2.3 Algorithmic Trading 9
1.3 The New Paradigm of Data-Driven Trading 11
References 14
2 The Role of Data in Trading and Investing 15
2.1 The Data-Driven Decision-Making Paradigm 15
2.2 The Data Economy is Fueling the Future 17
2.2.1 The Value of Data - Data as an Asset 18
2.3 Defining Data and Its Utility 20
2.4 The Journey from Data to Intelligence 24
2.5 The Utility of Data in Trading and Investing 30
2.6 The Alternative Data and Its Use in Trading and Investing 34
References 36
3 Artificial Intelligence - Between Myth and Reality 39
3.1 Introduction 39
3.2 The Evolution of AI 41
3.2.1 Early History 41
3.2.2 The Modern AI Era 43
3.2.3 Important Milestones in the Development of AI 44
3.2.4 Projections for the Immediate Future 48
3.2.5 Meta-Learning - An Exciting New Development 49
3.3 The Meaning of AI - A Critical View 51
3.4 On the Applicability of AI to Finance 54
3.4.1 Data Stationarity 57
3.4.2 Data Quality 58
3.4.3 Data Dimensionality 59
3.5 Perspectives and Future Directions 60
References 62
4 Computational Intelligence - A Principled Approach for the Era of Data Exploration 63
4.1 Introduction to Computational Intelligence 63
4.1.1 Defining Intelligence 63
4.1.2 What is Computational Intelligence? 64
4.1.3 Mapping the Field of Study 66
4.1.4 Problems vs. Tools 68
4.1.5 Current Challenges 69
4.1.6 The Future of Computational Intelligence 70
4.1.7 Examples in Finance 71
4.2 The PAC Theory 72
4.2.1 The Probably Approximately Correct Framework 73
4.2.2 Why AI is a Very Lofty Goal to Achieve 75
4.2.3 Examples of Ecorithms in Finance 78
4.3 Technology Drivers Behind the ML Surge 81
4.3.1 Data 82
4.3.2 Algorithms 82
4.3.3 Hardware Accelerators 82
References 84
5 How to Apply the Principles of Computational Intelligence in Quantitative Finance 87
5.1 The Viability of Computational Intelligence 87
5.2 On the Applicability of CI to Quantitative Finance 91
5.3 A Brief Introduction to Reinforcement Learning 94
5.3.1 Defining the Agent 96
5.3.2 Model-Based Markov Decision Process 98
5.3.3 Model-Free Reinforcement Learning 101
5.4 Conclusions 104
References 104
6 Case Study 1: Optimizing Trade Execution 107
6.1 Introduction to the Problem 107
6.1.1 On Limit Orders and Market Microstructure 109
6.1.2 Formulation of Base-Line Strategies 111
6.1.3 A Reinforcement Learning Formulation for the Optimized Execution Problem 112
6.2 Current State-of-the-Art in Optimized Trade Execution 114
6.3 Implementation Methodology 116
6.3.1 Simulating the Interaction with the Market Microstructure 116
6.3.2 Using Dynamic Programming to Optimize Trade Execution 118
6.3.3 Using Reinforcement Learning to Optimize Trade Execution 119
6.4 Empirical Results 122
6.4.1 Application to Equities 122
6.4.2 Using Private Variables Only 123
6.4.3 Using Both Private and Market Variables 123
6.4.4 Application to Futures 124
6.4.5 Another Example 126
6.5 Conclusions and Future Directions 127
6.5.1 Further Research 127
References 128
7 Case Study 2: The Dynamics of the Limit Order Book 129
7.1 Introduction to the Problem 129
7.1.1 The New Era of Prediction 130
7.1.2 New Challenges 131
7.1.3 High-Frequency Data 132
7.2 Current State-of-the-Art in the Prediction of Directional Price Movement in the LOB 133
7.2.1 The Contrarians 136
7.3 Using Support Vector Machines and Random Forest Classifiers for Directional Price Forecast 138
7.3.1 Empirical Results 139
7.4 Studying the Dynamics of the LOB with Reinforcement Learning 141
7.4.1 Empirical Results 142
7.4.2 Conclusions 144
7.5 Studying the Dynamics of the LOB with Deep Neural Networks 145
7.5.1 Results 148
7.6 Studying the Dynamics of the Limit Order Book with Long Short-Term Memory Networks 149
7.6.1 Empirical Results 152
7.6.2 Conclusions 153
7.7 Studying the Dynamics of the LOB with Convolutional Neural Networks 153
7.7.1 Empirical Results 155
7.7.2 Conclusions 156
References 157
8 Case Study 3: Applying Machine Learning to Portfolio Management 159
8.1 Introduction to the Problem 159
8.1.1 The Problem of Portfolio Diversification 160
8.2 Current State-of-the-Art in Portfolio Modeling 161
8.2.1 The Classic Approach 161
8.2.2 The ML Approach 162
8.3 A Deep Portfolio Approach to Portfolio Optimization 163
8.3.1 Autoencoders 164
8.3.2 Methodology - The Four-Step Algorithm 166
8.3.3 Results 167
8.4 A Q-Learning Approach to the Problem of Portfolio Optimization 167
8.4.1 Problem Statement 168
8.4.2 Methodology 169
8.4.3 The Deep Q-Learning Algorithm 169
8.4.4 Results 170
8.5 A Deep Reinforcement Learning Approach to Portfolio Management 170
8.5.1 Methodology 170
8.5.2 Data 171
8.5.3 The RL Setting: Agent, Environment, and
Policy 172
8.5.4 The CNN Implementation 172
8.5.5 The RNN and LSTM Implementations 172
8.5.6 Results 173
References 174
9 Case Study 4: Applying Machine Learning to Market Making 175
9.1 Introduction to the Problem 175
9.2 Current State-of-the-Art in Market Making 177
9.3 Applications of Temporal-Difference RL in Market Making 180
9.3.1 Methodology 180
9.3.2 The Simulator 181
9.3.3 Market Making Agent Specification 182
9.3.4 Empirical Results 185
9.4 Market Making in High-Frequency Trading Using RL 189
9.4.1 Methodology 190
9.4.2 Experimental Setting 191
9.4.3 Results and Conclusions 192
9.5 Other Research Studies 192
References 193
10 Case Study 5: Applications of Machine Learning to Derivatives Valuation 197
10.1 Introduction to the Problem 197
10.1.1 Problem Statement and Research Questions 199
10.2 Current State-of-the-Art in Derivatives Valuation by Applying ML 200
10.2.1 The Beginnings: 1992-2004 201
10.2.2 The Last Decade 202
10.3 Using Deep Learning for Valuation of Derivatives 204
10.3.1 Implementation Methodology 205
10.3.2 Empirical Results 207
10.3.3 Conclusions and Future Directions 208
10.3.4 Other Research Studies 208
10.4 Using RL for Valuation of Derivatives 210
10.4.1 Using a Simple Markov Decision Process 210
10.4.2 The Q-Learning Black-Scholes Model (QLBS) 212
References 214
11 Case Study 6: Using Machine Learning for Risk Management and Compliance 217
11.1 Introduction to the Problem 217
11.1.1 Challenges 218
11.1.2 The Problem 219
11.2 Current State-of-the-Art for Applications of ML to Risk Management and Compliance 219
11.2.1 Credit Risk 219
11.2.2 Market Risk 220
11.2.3 Operational Risk 221
11.2.4 Regulatory Compliance Risk and RegTech 222
11.2.5 Current Challenges and Future Directions 223
11.3 Machine Learning in Credit Risk Modeling 224
11.3.1 Data 225
11.3.2 Models 225
11.3.3 Results 226
11.4 Using Deep Learning for Credit Scoring 227
11.4.1 Introduction 227
11.4.2 Deep Belief Networks and Restricted Boltzmann Machines 228
11.4.3 Empirical Results 230
11.5 Using ML in Operational Risk and Market Surveillance 230
11.5.1 Introduction 230
11.5.2 An ML Approach to Market Surveillance 232
11.5.3 Conclusions 233
References 233
12 Conclusions and Future Directions 237
12.1 Concluding Remarks 237
12.2 The Paradigm Shift 239
12.2.1 Mathematical Models vs. Data Inference 240
12.3 De-Noising the AI Hype 243
12.3.1 Why Intellectual Honesty Should Not Be Abandoned 244
12.4 An Emerging Engineering Discipline 245
12.4.1 The Problem 246
12.4.2 The Market 246
12.4.3 A Possible Solution 246
12.5 Future Directions 247
References 248
Index 249
CRIS DOLOC is a leading computational scientist with more than 25 years of experience in quantitative finance. He holds a PhD in Computational Physics and is currently teaching at the University of Chicago in the Financial Mathematics program. Cris is also the founder of FintelligeX, a technology platform designed to promote data-driven education, and he is very passionate about the opportunities that recent developments in Cognitive Computing and Computational Intelligence could bring to the field of Quant education.
