ETD Collection

Permanent URI for this collectionhttps://wiredspace.wits.ac.za/handle/10539/104


Please note: Digitised content is made available at the best possible quality range, taking into consideration file size and the condition of the original item. These restrictions may sometimes affect the quality of the final published item. For queries regarding content of ETD collection please contact IR specialists by email : IR specialists or Tel : 011 717 4652 / 1954

Follow the link below for important information about Electronic Theses and Dissertations (ETD)

Library Guide about ETD

Browse

Has files: YesSubject: search.filters.subject.Simulation methods

Search Results

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    The optimally diversified equity portfolio in South Africa: an artificial intelligence approach
    (2017) Block, Aaron Eliyahu
    Diversification has remained a central tenet in investment theory over multiple decades due to its demonstrated value as a risk mitigation technique. Increasing the number assets in a portfolio, where the magnitude of correlation is relatively slim, increases the amount of diversification while also encountering increased costs in the form of transaction costs, taxes and the like. Thus, it is imperative to solve for the optimal point of diversification to ensure an investor does not encounter unnecessary costs. This study aims to solve for the point of optimal diversification in an equity portfolio, focusing on the South African environment. This is achieved by employing a framework using both the traditional simulation method as well as more advanced mathematical techniques, namely: genetic programming and particle swarm optimisation. Marked improvements are realised in this study with regards to the methodology and results through the application of advanced mathematical approaches in addition to removing the restriction of equal weightings being applied to each share in the portfolio. The results revealed that an optimal portfolio can be constructed using up to only 15 shares. Secondly, the genetic programming approach demonstrated increased strength compared to the traditional simulation and particle swarm optimisation approaches, obtaining a greater level of diversification with fewer shares. Finally, although the aim of the study is focused on modelling the relationship between the number of shares in a portfolio and the achievable diversification benefits, it is also established that the portfolios indicated as being optimally diversified achieved market beating returns.
  • No Thumbnail Available
    Item
    Application of stochastic orebody simulation techniques to assess geological volume and grade uncertainty for gold reef deposits
    (2017) Chanderman, Lisa
    This dissertation discusses the use of stochastic orebody modelling techniques for assessing geological uncertainty associated with gold mineralisation at Geita Gold Mine in Tanzania, and proposes a practical methodology that can be applied to similar studies. As part of the pre-feasibility stage studies for underground mining at Geita, stochastic simulations were required to assess the geological uncertainty associated with isolating (modelled) high grade lenses that occur within the known low grade mineralisation currently targeted for underground mining. Two different simulation techniques are applied in this research: Sequential Indicator Simulation to generate lithofacies realisations from which to assess ore category boundaries and shapes for use in quantifying volumetric uncertainty; and Direct Block Simulations to simulate gold grade realisations from which to assess grade uncertainty. This study identified potential upside and downside mine planning scenarios for volumes and total metal content from the ore category and grade simulations respectively. The findings of the results demonstrated that the high grade zones are much more broken up and discontinuous than the currently modelled high grade shape. The current business case uses a probabilistic high grade shape based on a single grade indicator and a probability choice of 50 percent as the threshold for high grade. The results of the study consider a simulation of possible outcomes based on the same threshold grade indicator and hence quantify the uncertainty or total geological risk. This geological risk may be introduced to mine designs, production schedules and NPV predictions The stochastic workflow developed can be applied to analogous deposit types to assess the risk related to geological uncertainty. The work includes a description of practical considerations to be accounted for when applying the techniques.
  • No Thumbnail Available
    Item
    The effect of simulation bias on action selection in Monte Carlo Tree Search
    (2016) James, Steven Doron
    Monte Carlo Tree Search (MCTS) is a family of directed search algorithms that has gained widespread attention in recent years. It combines a traditional tree-search approach with Monte Carlo simulations, using the outcome of these simulations (also known as playouts or rollouts) to evaluate states in a look-ahead tree. That MCTS does not require an evaluation function makes it particularly well-suited to the game of Go — seen by many to be chess’s successor as a grand challenge of artificial intelligence — with MCTS-based agents recently able to achieve expert-level play on 19×19 boards. Furthermore, its domain-independent nature also makes it a focus in a variety of other fields, such as Bayesian reinforcement learning and general game-playing. Despite the vast amount of research into MCTS, the dynamics of the algorithm are still not yet fully understood. In particular, the effect of using knowledge-heavy or biased simulations in MCTS still remains unknown, with interesting results indicating that better-informed rollouts do not necessarily result in stronger agents. This research provides support for the notion that MCTS is well-suited to a class of domain possessing a smoothness property. In these domains, biased rollouts are more likely to produce strong agents. Conversely, any error due to incorrect bias is compounded in non-smooth domains, and in particular for low-variance simulations. This is demonstrated empirically in a number of single-agent domains.