Briola, Antonio;
(2024)
Deep Complex Networks: Applications in Financial Systems Modeling.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Briola_10201193_thesis.pdf Download (6MB) | Preview |
Abstract
Financial markets are highly stochastic environments with a generally low signal-to-noise ratio, characterized by the interplay of a heterogeneous group of actors competing at different timescales, possessing asymmetric levels of information and diverse technical abilities to trade financial securities. Higher-order networks and automated learning models have emerged as valuable tools for handling the related complexity. In this thesis, I aim to provide the academics and practitioners community with original insights involving concepts from three main research fields: (i) market microstructure; (ii) network science; and (iii) deep learning modeling. Such a research effort can be summarized into four main contributions: - I study the practicability of the Triangulated Maximally Filtered Graph (a theoretical concept from Information Filtering Networks theory) in modeling the evolutionary process of the time-dependent hierarchical organization of financial markets. For this scope, I analyze the cryptocurrency market and compare the obtained results with those achieved in the past twenty years of similar research in the stock market, uncovering comparable behaviors between the two systems. - Moving the analysis to finer-grained timescales (i.e., orders' submission level), I study how the microstructural properties of traditional financial assets (i.e., equities) affect the effectiveness of a state-of-the-art deep learning model in Limit Order Book forecasting, and I propose a novel evaluation metric to quantify the so-called 'simulation-to-reality' gap. - I discuss an innovative methodology for efficiently mapping the informational content of an arbitrary Information Filtering Network into a novel class of deep learning models called 'Homological Convolutional Neural Networks'. I test their effectiveness in forecasting problems from the so-called 'unconquered castle' of deep learning (i.e., tabular learning), demonstrating that this new approach achieves state-of-the-art performances in a controlled yet challenging environment. - I show how to adapt 'Homological Convolutional Neural Networks' for multivariate time-series forecasting tasks. Experimental results shed new light on deeper spatial dynamics in high-frequency financial systems, narrowing the gap between microstructural modeling and deep learning-based forecasting of Limit Order Books.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Deep Complex Networks: Applications in Financial Systems Modeling |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Copyright © The Author 2024. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request. |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10201193 |
Loading...
Loading...Loading...Loading...Archive Staff Only
 |
View Item |
