Machine Learning for Asset Managers

by Marcos López de Prado

Table of Contents

Chapter 1: Introduction

• Motivation
• Theory Matters
• How Scientists Use ML
• Two Types of Overfitting
• Outline
• Audience
• Five Popular Misconceptions about Financial ML
• Frequently Asked Questions
• Conclusion

Chapter 2: Denoising and Detoning

• Motevation
• The Marcenko-Pastur Theorem
• Random Matrix with Signal
• Fitting the Marcenko-Pastur Distribution
• Denoising
• Detoning
• Experimental Results
• Conclusion

Chapter 3: Distance Metrics

• Motevation
• Correlation-Based Metrics
• Marginal and Joint Entropy
• Conditional Entropy
• Kullback-Leibler Divergence
• Cross-Entropy
• Mutual Information
• Variation of Information
• Discretization
• Distance between Two Partitions
• Experimental Results
• Conclusion

Chapter 4: Optimal Clustering

• Motevation
• Proximity Matrix
• Types of Clustering
• Number of Clusters
• Experimental Results
• Conclusions

Chapter 5: Financial Labels

• Motevation
• Fixed-Horizon Method
• Triple-Barrier Method
• Trend-Scanning Method
• Meta-Labeling
• Experimental Results
• Conclusions

Chapter 6: Feature Importance Analysis

• Motivation
• p-Values
• Feature Importance
• Probability-Weighted Accuracy
• Substitution Effects
• Experimental Results
• Conclusions*

Chapter 7: Portfolio Construction

• Motivation
• Convex Portfolio Optimization
• The Condition Number
• Markowitz's Curse
• Signal as a Source of Covariance Instability
• Nested Clustered Optimization Algorithm
• Experimental Results
• Conclusions*

Chapter 8: Testing Set Overfitting

• Motivation
• Precision and Recall
• Precision and Recall under Multiple Testing
• Sharpe Ratio
• The "False Strategy" Theorem
• Experimental Results
• The Deflated Sharpe Ratio
• Familywise Error Rate
• Conclusions