Malandreniotis, Dimitri;
(2024)
Probabilistic Forecasting Models for Multidimensional Financial Time-series With Applications to Systematic Portfolio Management.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Malandreniotis_10185435_thesis.pdf Download (14MB) | Preview |
Abstract
In the current era of financial markets, the emergence of autonomous trading systems and systematic strategies has significantly transformed the landscape of asset management. Among these developments, dynamic asset allocation is a strategy that manages a portfolio systematically, adjusting asset positions in response to ever-changing market conditions. At the heart of these strategies is a forecasting model responsible for quantifying the uncertainty of the assets’ future price movements. The quality of continuously updated probabilistic forecasts is of paramount importance for effectively managing risk, generating returns, and is crucial for success. Systematic strategies, such as these, which are designed to require frequent trading, are inherently costly; this can significantly impair the performance of a strategy, making it prohibitively expensive to operate. This thesis addresses these two limiting factors for effective dynamic asset allocation. The primary contribution of the thesis, extending over the first two research chapters, is a novel modelling methodology for forecasting the joint distributions of entire portfolios. Leveraging gradient boosting within a distributional regression framework, this methodology allows all parameters of the multivariate distribution to be time-varying conditioned on high-dimensional sets of exogenous variables, with additional structural time-variation provided by regime-switching. This framework is capable of capturing well-known statistical properties of asset returns, including time-varying covariance, higher-order co-moments, asymmetries in shape and dependence, among other stylised facts. Experiments are undertaken to evaluate the efficacy and feasibility of applying this model in systematic strategies for managing equity portfolios. The second major contribution of this thesis addresses the impact that transaction costs have on the performance of portfolio strategies. It introduces a novel method for portfolio optimisation, one that takes into consideration the potential costs of portfolio rebalancing. The performance improvements resulting from the inclusion of this new method are demonstrated to be substantial when compared to traditional portfolio optimisation techniques.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Probabilistic Forecasting Models for Multidimensional Financial Time-series With Applications to Systematic Portfolio Management |
| 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 UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10185435 |
Archive Staff Only
 |
View Item |
