How to read this book

If you are reading this, then you are probably aware that ML has become a strategic capability in many industries, including the investment industry. The explosion of digital data that drives much of the rise of ML is having a particularly powerful impact on investing, which already has a long history of using sophisticated models to process information. The scope of trading across asset classes implies that a vast range of new, alternative data may be relevant in addition to the market and fundamental data that used to be the focus of the analytical efforts.

You may have also come across the insight that the successful application of ML or data science requires the integration of statistical knowledge, computational skills, and domain expertise at the individual or team level. In other words, it is essential to ask the right questions, identify and understand the data that may provide the answers, deploy a broad range of tools to obtain results, and interpret them in a way that leads to the right decisions.

Consequently, this book takes an integrated perspective on the application of ML to the domain of investment and trading. In this section, we will lay out what to expect, how it goes about achieving its objectives, and what you need to both meet your goals and have fun in the process.

What to expect

This book aims to equip you with the strategic perspective, conceptual understanding, and practical tools to add value from applying ML to the trading and investment process. To this end, it covers ML as an important element in a process rather than a standalone exercise.

First and foremost, it covers a broad range of supervised, unsupervised, and reinforcement learning algorithms useful for extracting signals from the diverse data sources relevant to different asset classes. It introduces a ML workflow and focuses on practical use cases with relevant data and numerous code examples. However, it also develops the mathematical and statistical background to facilitate the tuning of an algorithm or the interpretation of the results.

The book recognizes that investors can extract value from third-party data more than other industries. As a consequence, it covers not only how to work with market and fundamental data but also how to source, evaluate, process, and model alternative data sources such as unstructured text and image data.

It relates the use of ML to research and evaluate alpha factors to quantitative and factor-based strategies and introduces portfolio management as the context for the deployment of strategies that combine multiple alpha factors. It also highlights that ML can add value beyond predictions relevant to individual asset prices, for example to asset allocation and addresses the risks of false discoveries from using ML with large datasets to develop a trading strategy.

It should not be a surprise that this book does not provide investment advice or ready-made trading algorithms. Instead, present building blocks required to identify, evaluate, and combine datasets that suitable for any given investment objective, select and apply ML algorithms to this data, and develop and test algorithmic trading strategies based on the results.

Who should read this book

You should find the book informative if you are an analyst, data scientist, or ML engineer with an understanding of financial markets and interest in trading strategies. You should also find value as an investment professional who aims to leverage ML to make better decisions.

If your background is software and ML, you may be able to just skim or skip some introductory material on ML. Similarly, if your expertise is in investment, you will likely be familiar with some or all of the financial context. You will likely find the book most useful as a survey of key algorithms, building blocks and use cases than for specialized coverage of a particular algorithm or strategy. However, the book assumes you are interested in continuing to learn about this very dynamic area. To this end, it references numerous resources to support your journey towards customized trading strategies that leverage and build on the fundamental methods and tools it covers.

You should be comfortable using Python 3 and various scientific computing libraries like numpy, pandas, or scipy and be interested in picking up numerous others along the way. Some experience with ML and scikit-learn would be helpful, but we briefly cover the basic workflow and reference various resources to fill gaps or dive deeper.

How the book is organized

The book provides a comprehensive introduction to how ML can add value to the design and execution of trading strategies. It is organized in four parts that cover different aspects of the data sourcing and strategy development process, as well as different solutions to various ML challenges.

Part I – the framework – from data to strategy design

The first part provides a framework for the development of algorithmic trading strategies. It focuses on the data that power the ML algorithms and strategies discussed in this book, outlines how ML can be used to derive trading signals, and how to deploy and evaluate strategies as part of a portfolio.

The remainder of this chapter summarizes how and why ML became central to investment, describes the trading process and outlines how ML can add value. Chapter 2, Market and Fundamental Data, covers sources and working with original exchange-provided tick and financial reporting data, as well as how to access numerous open-source data providers that we will rely on throughout this book.

Chapter 3, Alternative Data for Finance, provides categories and criteria to assess the exploding number of sources and providers. It also demonstrates how to create alternative data sets by scraping websites, for example to collect earnings call transcripts for use with natural language processing (NLP) and sentiment analysis algorithms in the second part of the book.

Chapter 4, Alpha Factor Research, provides a framework for understanding how factors work and how to measure their performance, for example using the information coefficient (IC). It demonstrates how to engineer alpha factors from data using Python libraries offline and on the Quantopian platform. It also introduces the zipline library to backtest factors and the alphalens library to evaluate their predictive power.

Chapter 5, Strategy Evaluation, introduces how to build, test and evaluate trading strategies using historical data with zipline offline and on the Quantopian platform. It presents and demonstrates how to compute portfolio performance and risk metrics using the pyfolio library. It also addresses how to manage methodological challenges of strategy backtests and introduce methods to optimize a strategy from a portfolio risk perspective.

Part 2 – ML fundamentals

The second part covers the fundamental supervised and unsupervised learning algorithms and illustrates their application to trading strategies. It also introduces the Quantopian platform where you can leverage and combine the data and ML techniques developed in this book to implement algorithmic strategies that execute trades in live markets.

Chapter 6, The Machine Learning Process, sets the stage by outlining how to formulate, train, tune and evaluate the predictive performance of ML models as a systematic workflow.

Chapter 7, Linear Models, it shows how to use linear and logistic regression for inference and prediction and how to use regularization to manage the risk of overfitting. It presents the Quantopian trading platform and demonstrates how to build factor models and predict asset prices.

Chapter 8, Time Series Models, covers univariate and multivariate time series, including vector autoregressive models and cointegration tests, and how they can be applied to pairs trading strategies. Chapter 9, Bayesian Machine Learning, presents how to formulate probabilistic models and how Markov Chain Monte Carlo (MCMC) sampling and Variational Bayes facilitate approximate inference. It also illustrates how to use PyMC3 for probabilistic programming to gain deeper insights into parameter and model uncertainty.

Chapter 10, Decision Trees and Random Forests, shows how to build, train and tune non-linear tree-based models for insight and prediction. It introduces tree-based ensemble models and shows how random forests use bootstrap aggregation to overcome some of the weaknesses of decision trees. Chapter 11, Gradient Boosting Machines ensemble models and demonstrates how to use the libraries xgboost, lightgbm, and catboost for high-performance training and prediction, and reviews in depth how to tune the numerous hyperparameters.

Chapter 12, Unsupervised Learning introduces how to use dimensionality reduction and clustering for algorithmic trading. It uses principal and independent component analysis to extract data-driven risk factors. It presents several clustering techniques and demonstrates the use of hierarchical clustering for asset allocation.

Part 3 – natural language processing

Part three focuses on text data and introduces state-of-the-art unsupervised learning techniques to extract high-quality signals from this key source of alternative data.

Chapter 13, Working with Text Data, demonstrates how to convert text data into a numerical format and applies the classification algorithms from part two for sentiment analysis to large datasets. Chapter 14, Topic Modeling, applies Bayesian unsupervised learning to extract latent topics that can summarize a large number of documents and offer more effective ways to explore text data or use topics as features for a classification model. It demonstrates how to apply this technique to earnings call transcripts sourced in Chapter 3, Alternative Data for Finance, and to annual reports filed with the Securities and Exchange Commission (SEC).

Chapter 15, Word Embeddings, uses neural networks to learn state-of-the-art language features in the form of word vectors that capture semantic context much better than traditional text features and represent a very promising avenue for extracting trading signals from text data.

Part 4 – deep and reinforcement learning

Part 4 introduces deep learning and reinforcement learning.

• Chapter 16, Deep Learning, introduces Keras, TensorFlow and PyTorch, the most popular deep learning frameworks and illustrates how to train and tune various architectures.
• Chapter 17, Convolutional Neural Networks, illustrates how to use CNNs with image and text data
• Chapter 18, Recurrent Neural Networks, presents RNNs for time series data
• Chapter 19, Autoencoders and Generative Adversarial Nets, shows how to use deep neural networks for unsupervised learning with autoencoders and presents GANs that produce synthetic data
• Chapter 20, Reinforcement Learning, demonstrates the use of reinforcement learning to build dynamic agents that learn a policy function based on rewards using the OpenAI gym platform

What you need to succeed

The book content revolves around the application of ML algorithms to different datasets. Significant additional content is hosted on GitHub to facilitate review and experiments with the examples discussed in the book. It contains additional detail and instructions as well as numerous references.

Data sources

We will use freely available historical data from market, fundamental and alternative sources. Chapter 2, Market and Fundamental Data and Chapter 3, Alternative Data for Finance cover characteristics and access to these data sources and introduce key providers that we will use throughout the book. The companion GitHub repository (see beneath) contains instructions on how to obtain or create some of the datasets that we will use throughout and includes some smaller datasets.

A few sample data sources that we will source and work with include, but are not limited to:

• NASDAQ ITCH order book data
• Electronic Data Gathering, Analysis, and Retrieval (EDGAR) SEC filings
• Earnings call transcripts from Seeking Alpha
• Quandl daily prices and other data points for over 3,000 US stocks
• Various macro fundamental data from the Federal Reserve and others
• Large Yelp business reviews and Twitter datasets
• Image data on oil tankers

Some of the data is several GB large (for example the NASDAQ and SEC filings). The notebooks indicate when that is the case.

GitHub repository

The GitHub repository contains Jupyter Notebooks that illustrate many of the concepts and models in more detail. The Notebooks are referenced throughout the book where used. Each chapter has its own directory with separate instructions where needed, as well as reference specific to the chapter's content.

Jupyter Notebooks is a great tool for creating reproducible computational narratives, and it enables users to create and share documents that combine live code with narrative text, mathematical equations, visualizations, interactive controls, and other rich output. It also provides building blocks for interactive computing with data, such as a file browser, terminals, and a text editor.

You can find the code files placed at:
https://github.com/PacktPublishing/Hands-On-Machine-Learning-for-Algorithmic-Trading.

Python libraries

The book uses Python 3.7, and recommends miniconda to install the conda package manager and to create a conda environment to install the requisite libraries. To this end, the GitHub repo contains an environment.yml file. Please refer to the installation instructions referenced in the GitHub repo's README file.