Artificial Intelligence for Asset Management and Investment

Chapter 5Model Development

AT THE HEART OF BUILDING industrial-scale enterprise machine learning (ML) is model development. I use the term “modeling” to simplify and be consistent with single words in the process (data, evaluation, performance). This is where ML learning solutions are developed. ML helps machines to learn from data. ML methods are about extracting models from data such that we can extract patterns and make better predictions. In ML, we use methods to extract models. An alternative way to think about ML is to view it as fitting the model to the data.

Please note that in this chapter I will provide an introduction to ML. If you have some background in ML, you can skip this chapter. This chapter is meant for businesspeople who are trying to get a high-level introduction to ML. The main goal is to have businesspeople learn about the methods and tools that are used by this group. This can help facilitate communications between the data science and the business teams.

WHO IS RESPONSIBLE?

In the previous chapter we focused on identifying data and features. Feature libraries helped us determine the most informative aspects of raw data. At that stage we did not know what model would fit best with what data. We did not know which model worked best for the data we have. Data function (Chapter 4) manages data but is not a model development group. Modeling group or station is where actual solutions are produced. This is where real strategies are developed and algorithms ...