Portfolio optimisation with couplas and evolutionary algorithms

Project description 

Modern portfolio theory was popularised by Harry Markowitz in 1952 and gives predictions of expected return and risk measures of portfolios held by investors. There have been solvers for varying types of portfolio optimisation problems of assets with varying constraints. Many such solvers assume the Markowitz model, using the covariance matrix and mean returns of the assets to find the optimal distribution of ideal weights. However, these approaches assume the underlying joint distribution of assets is Gaussian and also ignores the fact that investors may desire other characteristics of the asset distribution than mean and standard deviation. Asset distributions of real-world financial data are rarely normal and exhibit so-called “heavy tails” behaviour (e.g., financial crashes).

In statistics, copulae model the marginal of each asset separately (which may be any distribution) and also the interdependencies between assets (this could be using Monte-Carlo methods or GARCH models, for example). The project is to research copula models which give novel real-world characterisations of risk, using them in conjunction with a meta-heuristic solver  implemented on an HPC cluster to perform computations on large real-world datasets in order to predict the joint probability distributions of stocks over time.

Publications

  • Tahani Alotaibi, Panagiotis Tziogkidis & Matthew Craven, 2019 "Constructing Class-Based Portfolios on Gulf Markets with Metaheuristics", OR-61
  • Tahani Alotaibi & Matthew Craven, 2019 "Efficient Frontiers in Portfolio Optimisation with Minimum Proportion Constraints" Proc. Genetic and Evolutionary Computation Conference

    Popular posts from this blog

    New PhD Students: Pawel Manikowski and Rahul Soni

    We're very pleased to welcome two new PhD students to the group this week. Pawel Manikowski will be investigating the optimisation of floating offshore wind farms, and is supported by an EPSRC DTP fellowship. His work will be looking at modelling the movement of these floating structures, as well as optimising their placement of individual turbines within a floating wind farm. Floating wind farms are a nascent technology that holds the promise of greatly expanding renewable energy generation. Rahul Soni will be working on a project investigating the use of hyper-heuristics, combining machine learning with optimisation, to solve multi-objective portfolio optimisation problems and exploring novel algorithms in this area. We are looking forward to sharing the results of their research with you soon!
    Read more

    New PhD Student: Mathew Walter

    This week we welcomed our latest PhD student to the group. Mathew Walter is going to be working on visualising the processes that are used to generate solutions in evolutionary algorithms with the aim of making optimisation processes more explainable. Mathew joins us having graduated from Plymouth with a degree in Mathematics, and is supervised by David Walker and Matthew Craven. We look forward to sharing the outputs from his research with you soon.
    Read more