Abstract
The ongoing global economic turmoil has compelled the asset management industry to explore innovative approaches to financial risk management. Portfolio optimization and risk budgeting stand at the forefront of computational finance studies, captivating both academics and practitioners alike. Analyses conducted by practitioners investigating sub-optimal performance during the 2008 financial crisis, despite the widespread reliance on Markowitz’s portfolio optimization for diversification, unveiled a deficiency in existing investment techniques.In today’s dynamic global financial markets, algorithmic trading is integral to investment strategies for attaining financial objectives. The evolution of technology and novel algorithmic trading methods has ushered in more efficient trade execution, leading to lower transaction costs, enhanced portfolio performance, and heightened transparency. However, algorithmic trading grapples with notable shortcomings, such as users often becoming complacent in persistently applying the same algorithm without continuous evaluation and adaptation to changing market conditions. These drawbacks result in inconsistent performance, depriving investors of the ability to determine optimal levels of returns and volatility. Consequently, existing trading algorithms fall short of meeting all trading needs.
This research project delves into existing financial models and introduces innovative portfolio management methods grounded in ensemble modeling of machine learning techniques. The approach involves constructing an investment portfolio through the decomposition of marginal asset risk contributions within a given universe. A diversification strategy is then formulated to address systematic and unsystematic risks, ultimately defining an optimization model. The study seeks to apply machine learning to enhance algorithmic trading strategies by identifying minimum risk contributors from an asset universe and maximizing diversification among them, utilizing techniques akin to the Maximum Diversification ratio. The optimal portfolio construction involves employing an ensemble of computational financial models, including Risk Budgeting and Maximum Diversification portfolio strategies.
The primary objective of this research is to illustrate an autonomous portfolio management model and benchmark its performance against industry standards such as the global major indices (S&P 500, S&P 100, and Dow Jones Industrial). Factors such as investment horizon, returns, and risks will be considered for comparison, showcasing the outcomes of our portfolio optimization and diversification strategies. This research project aims to enhance current portfolio construction techniques and outperform global major indices and industry-standard risk parity portfolios across various investment horizons and strategies.
| Date of Award | 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Dongmo Zhang (Supervisor) & Yi Guo (Supervisor) |