• Login
    View Item 
    •   SU+ Home
    • Research and Publications
    • Faculty of Information Technology (FIT)
    • FIT Projects, Theses and Dissertations
    • MSc. CIS Theses and Dissertations
    • MSc. CIS Theses and Dissertations (2020)
    • View Item
    •   SU+ Home
    • Research and Publications
    • Faculty of Information Technology (FIT)
    • FIT Projects, Theses and Dissertations
    • MSc. CIS Theses and Dissertations
    • MSc. CIS Theses and Dissertations (2020)
    • View Item
    JavaScript is disabled for your browser. Some features of this site may not work without it.

    A hybrid predictive prototype for portfolio selection using probability based quadratic programming and neural networks

    Thumbnail
    View/Open
    Fulltext Thesis (4.325Mb)
    Date
    2020
    Author
    Muganda, Brian Wesley
    Metadata
    Show full item record
    Abstract
    A portfolio is a collection of investments held by an investment company, hedge fund, financial institution or individual. This collection of investment features a combination of financial assets such as stocks, bonds or options. The designing of a portfolio (fund allocation to each asset and selection of the assets) is done according to the investor’s risk tolerance, investment time frame and investment objectives. Robo-advisors, which are automated algorithm-driven investment platforms that use quantitative algorithms to manage investors’ portfolios, are at times used to perform portfolio allocation for investors having defined their risk preferences and investment time frames. However a majority of these robo-advisors still rely on classical mean variance allocation techniques of modern portfolio theory. This research therefore, developed a robo-financial advisor prototype on a hybrid programming architecture by using artificial neural networks to predict portfolio returns and variances based on underlying multi-asset Uhlenbeck process (OU) and geometric Brownian motion (GBM) processes’ estimates. These results were subsequently used in the probability-based quadratic optimization algorithm to provide an optimal portfolio allocation strategy. This probability-based quadratic programming approach is novel and is based on return certainty probability and value-at-risk constraints acting as proxies for investor’s risk tolerance. The results showed that neural network algorithm performed averagely well forecasting being able to predict the correct level of 2 of 5 assets and to predict correctly the trends of the remaining 3 assets, it however yielded low standard deviations compared to the OU and GBM models. The quadratic optimization algorithm supported investment in shorter time horizons since portfolio risk was lowest. Diversified allocation was achieved in the shorter time horizons. Longer horizons allocations were biased towards asset with lower standard deviations. Lowest risk portfolio was the ones with a lower certainty probability of target return and vice versa. Also, it shows a hybrid programming paradigm is an effective approach to leverage on strengths, speed and functionality of different programming languages; an elixir for multifaceted dissociable programming problems.
    URI
    http://hdl.handle.net/11071/12066
    Collections
    • MSc. CIS Theses and Dissertations (2020) [2]

    DSpace software copyright © 2002-2016  DuraSpace
    Contact Us | Send Feedback
    Theme by 
    Atmire NV
     

     

    Browse

    All of SU+Communities & CollectionsBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

    My Account

    Login

    DSpace software copyright © 2002-2016  DuraSpace
    Contact Us | Send Feedback
    Theme by 
    Atmire NV