Predicting FX with Deep Learning

A Hybrid Multi-Model Approach

Mikhail Mew
Geek Culture

--

Photo by Aleksandr Popov on Unsplash

At a basic level the act of investing, requires the purchase of an asset with a view of price appreciation at some time horizon in the future. Although simple in nature, the complexity involved can quickly compound when successive assets are added to a portfolio increasing in size. This complexity stems from the fact that financial assets have fundamental and empirical relationships that underpin the co-movement of their returns as function of their exposure to market, economic conditions, and trading behavior. Furthermore, as monetary capital is a limiting factor in the real-world, investors are only able to participate in a subset of mutually exclusive opportunities from the investible universe. This process, aptly named portfolio construction, is a predicament faced by all professional money managers and retail traders alike.

In order to build robust and scalable portfolios, a deep understanding of the relationship between assets is required in addition to a view on which assets will offer superior risk adjusted return opportunities. To achieve these objectives, investors have often look looked at patterns in the past price of the asset to help predict the future return in a practice called technical analysis. Attempts to predict financial markets using technical analysis date as far back as the 17th century…

--

--

Mikhail Mew
Geek Culture

Researcher | Investor | Data Scientist | Curious Observer. Thoughts and insights from the confluence of investing and machine learning.