Msc Project Defence- Department of Computer Science-Atli Egilsson
Active Portfolio Management in the Icelandic Market
Title: Active Portfolio Management in the Icelandic Market
Candidate: Atli Egilsson
Date and Time: January 7th 2020 at 9:00
Location: M102
Abstract: In the field of portfolio management, investors often rely on instincts or strategies that are yet to be captured and optimized by computers. With more and more information being digitized, we can capture an increasing amount of information that can improve asset selection and investment related strategies. Automating the fundamental steps needed for portfolio management saves time spent on manual labor and gives investors more time for focusing on strategic planning rather than looking at raw data. Besides the mean-variance model, derived in the 1950s, there is no standard way of solving the portfolio optimization problem. Based on the extended version of said model, the mean-variance cardinality constrained portfolio optimization (MVCCPO) model, we propose a robust way for individuals to manage their own investments. Using this solution, investors should be able to construct a portfolio of financial assets based on their own strategies and their preference of risk. At any point in time, investors would also know whether it is logical to reconstruct their portfolio given the market changes or in case of extracting cash for financial needs. Automation of this sort requires a way of collecting and representing available financial assets for an algorithm to take on. Given assets are compared in terms of risk and relations and a robust method for constructing a portfolio is used to factor in those measures as well as user defined inputs. Said portfolio is then comparable to new propositions at any given time for the investor to see clearly whether it is sensible to reconstruct his investments. Fundamental methods, such as expected returns and value at risk, are applied for estimating higher level information about assets and a multi-objective evolutionary algorithm (MOEA) is designed to capture investment strategies. Ideally we have a pipeline that automates the generic ways of gathering and evaluating financial assets and shows how we can include custom methods in a robust framework to manage portfolios.
