Advances in Financial Machine Learning

CHAPTER 19 Microstructural Features

19.1 Motivation

Market microstructure studies “the process and outcomes of exchanging assets under explicit trading rules” (O'Hara [1995]). Microstructural datasets include primary information about the auctioning process, like order cancellations, double auction book, queues, partial fills, aggressor side, corrections, replacements, etc. The main source is Financial Information eXchange (FIX) messages, which can be purchased from exchanges. The level of detail contained in FIX messages provides researchers with the ability to understand how market participants conceal and reveal their intentions. That makes microstructural data one of the most important ingredients for building predictive ML features.

19.2 Review of the Literature

The depth and complexity of market microstructure theories has evolved over time, as a function of the amount and variety of the data available. The first generation of models used solely price information. The two foundational results from those early days are trade classification models (like the tick rule) and the Roll [1984] model. The second generation of models came after volume datasets started to become available, and researchers shifted their attention to study the impact that volume has on prices. Two examples for this generation of models are Kyle [1985] and Amihud [2002].

The third generation of models came after 1996, when Maureen O'Hara, David Easley, and others published their “probability of informed ...