Conclusions and Next Steps

Our goal for this book was to enable you to apply machine learning (ML) to a variety of data sources and extract signals that add value to a trading strategy. To this end, we took a more comprehensive view of the investment process, from idea generation to strategy evaluation, and introduced ML as an important element of this process in the form of the ML4T workflow.

While demonstrating the use of a broad range of ML algorithms, from the fundamental to the advanced, we saw how ML can add value at multiple steps in the process of designing, testing, and executing a strategy. For the most part, however, we focused on the core ML value proposition, which consists of the ability to extract actionable information from much larger amounts of data more systematically than human experts would ever be able to.

This value proposition has really gained currency with the explosion of digital data that made it both more promising and necessary...

Key takeaways and lessons learned

A central goal of the book was to demonstrate the workflow of extracting signals from data using ML to inform a trading strategy. Figure 23.1 outlines this ML-for-trading workflow. The key takeaways summarized in this section relate to specific challenges we encounter when building sophisticated predictive models for large datasets in the context of financial markets:

Figure 23.1: Key elements of using ML for trading

Important insights to keep in mind as you proceed to the practice of ML for trading include the following:

• Data is the single most important ingredient that requires careful sourcing and handling.
• Domain expertise is key to realizing the value contained in data and avoiding some of the pitfalls of using ML.
• ML offers tools that you can adapt and combine to create solutions for your use case.
• The choices of model objectives and performance diagnostics are key to productive iterations toward...

ML for trading in practice

As you proceed to integrate the numerous tools and techniques into your investment and trading process, there are numerous things you can focus your efforts on. If your goal is to make better decisions, you should select projects that are realistic yet ambitious given your current skill set. This will help you to develop an efficient workflow underpinned by productive tools and gain practical experience.

We will briefly list some of the tools that are useful to expand on the Python ecosystem covered in this book. They include big data technologies that will eventually be necessary to implement ML-driven trading strategies at scale. We will also list some of the platforms that allow you to implement trading strategies using Python, possibly with access to data sources, and ML algorithms and libraries. Finally, we will point out good practices for adopting ML as an organization.

Data management technologies

The central role of data in the ML4T...

Conclusion

We started by highlighting the explosion of digital data and the emergence of ML as a strategic capability for investment and trading strategies. This dynamic reflects global business and technology trends beyond finance and is much more likely to continue than to stall or reverse. Many investment firms are just getting started to leverage the range of artificial intelligence tools, just as individuals are acquiring the relevant skills and business processes are adapting to these new opportunities for value creation, as outlined in the introductory chapter.

There are also numerous exciting developments for the application of ML to trading on the horizon that are likely to propel the current momentum. They are likely to become relevant in the coming years and include the automation of the ML process, the generation of synthetic training data, and the emergence of quantum computing. The extraordinary vibrancy of the field implies that this alone could fill a book and...