Advances in Machine Learning for Quantitative Analysis
Authored by – Marc Eyrignoux, Specialist - Technology
Quantitative analysis is a very specific domain in Machine Learning and Artificial Intelligence, because of the nature of the data : markets are very competitive, self adaptive, irrational, and interlaced. These constraints do not exist in other domains of time-series analysis, and make them hard to predict. For example, the seasonal variations of the frequentation of Facebook are easy to detect with their open-sourced model « Prophet », and they reproduce each year, with the modularity of the « trend », ie. the current average of the curve. But the variations of the financial markets are not so easy to predict, and financial companies need specific practices and tooling that don’t exist anywhere else in AI.
Therefore, it is interesting to describe them shortly, and link them to the recent research advances of 2018, because that domain is much less open-sourced and published than the usual computer vision and language processing domains covered by the giants of the tech.