Explorations of Machine Learning in Finance
Thèse : Explorations of Machine Learning in Finance. Rechercher de 53 000+ Dissertation Gratuites et MémoiresPar Caezar • 19 Mai 2018 • Thèse • 24 594 Mots (99 Pages) • 782 Vues
[pic 1]
Department of Computer Science
Explorations of Machine Learning
in Quantitative Finance
Research Review
Lionel Cazarre-Cazalot
Supervisor: Dave Cliff
- May 2nd 2014
- Table of Contents
1 Introduction 2
1.1. Motivation 2
1.2. Refining the scope 3
1.3. Aims and objectives 3
1.4. Research methodology 4
2 Traditional financial valuation frameworks and their limitations 4
2.1 Quick review of traditional financial valuation frameworks 4
2.2 Limitations of the MPT and CAPM 5
2.3 General limitations of traditional models 5
3 Application of Machine Learning to Finance 6
3.1 Definition of Machine Learning 6
3.2 Taxonomy of ML algorithms 6
3.3 Survey of Machine Learning techniques applied to finance 7
3.4 Advantages of ML 8
3.5 Disadvantages of ML 9
3.6 Pitfalls of ML applied to stock market modelling 9
4 Evolutionary algorithms 10
4.1 Various types of Evolutionary Algorithms 10
4.2 Choosing among Evolutionary algorithms 11
5 Genetic Programming 14
5.1 The origins of GP 14
5.2 The structure and representation of GP programs 14
5.3 How GP works 16
5.3.1 The GP process in three steps 16
5.3.2 Generating the initial population 16
5.3.3 Fitness 17
5.3.4 Genetic operations 17
5.3.5 Termination criterion 18
5.4 Limitations of GP 18
6 Genetic Programmming applied to stock selection 18
6.1 Refining our objective 18
6.1.1 Stock investing 18
6.1.2 Return targets: absolute versus relative 19
6.2 Choice of the investable market 19
6.3 Candidate terminals 20
6.4 Data preprocessing 21
6.5 Candidate functions 22
6.6 Fitness function 23
6.7 Avoiding overfitting 23
6.8 Implementation 24
7 Data collection, preparation and preliminary analysis 25
7.1 Data requirements 25
7.2 Data providers 25
7.3 WRDS, Compustat and CRSP 25
7.4 Methodology used to build the dataset 26
7.4.1 Extraction of complete list of companies available from Compustat North America 26
7.4.2 Crosscheck of monthly stock prices and holding period returns against CRSP 27
7.5 The issue of missing data 27
7.6 Univariate linear regression analysis 28
7.7 Final dataset 29
7.8 Distribution of results 29
7.9 Multivariate linear regression analysis 30
8 Software implementation 31
8.1 Requirements 31
8.1.1 Business requirements 31
8.1.2 User requirements 33
8.1.3 Software requirements 34
8.2 Program design 35
8.3 Testing 36
8.4 Comment a sample run 37
8.5 Performance 40
9 Experiments 41
9.1 Objectives 41
9.2 Results 41
10 Process validation 44
10.1 Methodology 44
10.2 Results 44
10.3 Analysis 45
10.4 Caveats 46
11 Further work 46
11.1 Re-evaluating the results based on the statements publication date 46
11.2 Running the process on quarterly statements / evaluating new holding periods 47
11.3 Targeting risk-adjusted returns 47
11.4 Exploring more features 47
...