SUNScholar
SUNScholar is a leading digital archive for the preservation and promotion of the research output of Stellenbosch University.
Communities in DSpace
Select a community to browse its collections.
Recent Submissions
Item
BEYOND CRITICAL REALISM? REVISITING J. WENTZEL VAN HUYSSTEEN’S RATIONALITY
(Stellenbosch University, 2024-12) Lange, Hans-Christoph Thapelo; Vosloo, Robert; Stellenbosch University. Faculty of Theology. Dept. of Systematic Theology & Ecclesiology.
The apparent tension between religion and science was exacerbated by the rise of the logical positivist ideal of science. The pathos of much of the work of South African theologian J. Wentzel van Huyssteen was to develop a credible way to reestablish theology as a rational discipline within the wider academic community and culture. Motivated by contextual factors, van Huyssteen therefore proposed his critical realist approach in contrast to positivist theology via a metaphoric referential theory. In attempting to ascertain how his approach develops, an analysis of a selection of his works was undertaken. Writing from Princeton, van Huyssteen engaged the philosophical roots of modernity and postmodernity in foundationalism and nonfoundationalism respectively, finally proposing a postfoundational model for rationality as a third way. In his Gifford lectures—for which he refigured the directives with which the Gifford lectures were decreed— he sets out on the interdisciplinary quest of developing a nuanced doctrine on human uniqueness. The philosophical framework for this quest was given in van Huyssteen’s earlier work on postfoundationalism. For this reason, the disciplines of cognitive epistemology, paleoanthropology, neuroscience, and evolutionary psychology were investigated in terms of the perspectives they can provide on what makes humans unique. Their influence on the theology with which van Huyssteen participates in the interdisciplinary conversation was analysed. The salient features of van Huyssteen’s work and consequent development are thus outlined and highlighted. Moreover, his proposed theology was analysed for consistency with his own criteria of critical realism. It was found that van Huyssteen remains consistent with his intent and largely—though not fully—conforms to these criteria. A remaining positivist influence on critical realism was identified, but was nuanced by considering van Huyssteen’s chosen audience and also the potential theologies that a postfoundational rationality could host.
Item
Potenzial ästhetischer Medien für kulturbezogene Lernprozesse im DaF-Unterricht anhand der Analyse der Graphic Novel Madgermanes von Birgit Weyhe
(Stellenbosch University, 2024-12) Lange, Emily Rebecca; dos Santos, Isabel; Schier, Carmen; Stellenbosch University. Faculty of Arts and Social Sciences. Dept. of Modern Foreign Languages.
This master thesis explores the potential of aesthetic media, specifically graphic novels, to enhance culture-related learning in the context of teaching German as a foreign language. In response to recent developments in cultural studies, which have revised the concept of culture and reshaped foreign language teaching objectives, this research investigates how aesthetic media can promote a semiotic and performative understanding of culture. Central to this investigation is the concept of “discourse," which leads to a reassessment of the role of literature and aesthetic media in cultural learning practices. The primary research question examines how complex developments in cultural studies and media didactics can be integrated into foreign language through literature and aesthetic media. This thesis hypothesizes that aesthetic media, through their reflective engagement with power dynamics and their integration of diverse media forms, can meet the needs of contemporary cultural didactics. To demonstrate this, the study focuses on an analysis of the graphic novel Madgermanes by Birgit Weyhe. It argues that the novel's multimodal nature effectively fosters culture-sensitive and power-reflective learning processes, which are essential in today's globalized educational contexts. The analysis examines four key aspects of the novel, showing how its aesthetic and multimodal structure can be utilized to engage learners in discourse, enhance their symbolic competence, and promote their participation in diverse cultural narratives. The findings suggest that the graphic novel, as an aesthetic medium, supports culture-related learning processes by enabling learners to critically engage with cultural discourses, thus fostering competencies such as multiliteracies, discourse ability, and symbolic competence. The study concludes that Madgermanes is particularly effective in this role, highlighting the value of integrating aesthetic media into the foreign language curriculum to develop a more reflective and inclusive approach to cultural education.
Item
Development of an Asset Performance Management Reference Framework for Distributed Industrial Machinery
(Stellenbosch University, 2024-12) Landmann, Marius Benedikt; Louw, Louis; Palm, Daniel; Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering.
Globalisation and technological advancements have led to increased complexity in the corporate environment, affecting various sectors. This study focuses on developing a reference framework for Asset Performance Management (APM) in the context of distributed industrial machinery. The aim is to enhance asset availability and reliability, address concerns regarding third-party trust, and optimise data administration for improved overall APM. To create and improve the APM reference framework, which comprises of an APM system and a related operational concept, the study employs the Design Science Research (DSR) approach. This involves in-depth literature reviews, expert interviews, and a conducted case study. The framework integrates various technological concepts, such as cloud and edge computing, AI, Asset Administration Shell, sensors, and blockchain to accommodate a range of industrial scenarios. The reference framework encompasses processes for data collection, storage, transmission, and decision-making. A case study with a major German research institute was conducted to evaluate and refine it. The framework provides a comprehensive and flexible solution for organisations to improve their asset management practices. It shows adaptability to a range of organisational needs and offers actionable recommendations for asset management, including improving decision-making, reducing downtime, and enhancing operational efficiency. This research addresses knowledge gaps in APM for distributed machinery and contributes to the fields of industrial engineering, asset management and data quality control. It provides a foundation for future research and industry standards, bridging a gap in the current understanding and application of APM for distributed industrial machinery.
Item
A framework for intraday ensemble trading on the foreign exchange market
(Stellenbosch University, 2024-12) Koegelenberg, Dirk Johan Coetzee; van Vuuren, J. H.; Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering.
Financial trading consists of traders buying and selling financial assets in the hope of generating profit over time. These assets are traded in financial markets, an example of which is the liquid and volatile Foreign exchange (Forex) market. Generating profit when trading on the Forex market is not a trivial task. Many traders are, in fact, unsuccessful due to a variety of complicating factors such as the stochastic nature of the Forex market, market information inefficiencies, and trader cognitive biases. One might think that these problems can be conquered with enough trading experience, but research on the topic has shown that even highly skilled investment managers struggle with trading performance consistency in the long term. Modelling the behaviour of the Forex market in the light of this market complexity might, therefore, seem daunting to any novice trader. In an attempt to overcome the inefficiencies inherent to human traders, however, trading algorithms have been proposed as an alternative for automating parts of the trading process. Substantial amounts of time and resources have been committed by researchers to the design of new and innovative trading algorithms tailored to the pursuit of trading profitably and the establishment of a competitive edge over human and other algorithmic contenders. As a result, various frameworks for developing trading algorithms have been proposed in the literature, each enabling the establishment of new approaches toward the development of such a trading algorithm. These frameworks, however, typically conform to one of two extremes: They are either problem-specific (focusing on a particular portion of the trading pipeline) or lack enough depth to facilitate the inner workings of, and communication between, different framework constituent components adequately. Moreover, few frameworks emphasise the potential benefits of employing ensembling approaches during the trading process in the context of intraday trading, especially in respect of trading strategy ensembling. An intraday ensemble-based trading framework is proposed in this dissertation with the objective of addressing the shortcomings of current frameworks for this purpose in the literature. More specifically, the framework is tailored to provide a detailed (albeit holistic) road map for its users which may be used to design an ensemble-based Forex trading algorithm. Apart from the pre-processing of input data, the framework also facilitates processes such as forecasting future market behaviour and trading strategy development. Forecasting is conducted by invoking various time series forecasting methods from the realm of machine learning which are ultimately ensembled into a single forecast. This ensemble forecast is then incorporated into the trading strategies developed which are, in turn, also ensembled so as to strike a balance between risk mitigation and returns maximisation when executing Forex trades in real time. The practicality of the proposed framework is demonstrated via a computerised instantiation thereof. This framework instantiation is verified, after which it is validated by conducting two simulated real-world trading case studies.
Item
The Role of Features in Predictive Deep Learning Models for Auditory Tuberculosis Classification
(Stellenbosch University, 2024-12) Knight, Michael Sean; Niesler, T. R.; Stellenbosch University. Faculty of Engineering. Dept. of Electrical & Electronic Engineering.
The application of machine learning to cough audio for the purpose of tuberculosis (TB) detection promises to be a low-cost and easily deployable diagnostic method. Some, but little, work has been performed on this topic and while it has shown potential, a lack of consistency and scope leave a lack of clarity regarding where new research should focus. This work starts by considering previously identified model architectures and features, in conjunction with a common dataset, and methodically analyses and compares the effectiveness of several approaches. Three model architectures: logistic regression (LR), ResNets and BiLSTMs are tested, and a modification of the ResNet architecture, termed SkipNet, is proposed. Three feature types: linear filter banks (LFBs), mel filter banks (MFBs) and mel frequency cepstral coefficients (MFCCs) are applied to the architectures which themselves are combined in three different ways: ensemble models, single models and teacher-student models. Finally, forward sequential search (FSS) is explored as a means of reducing computation and improving robustness to irrelevant input. From our experiments, we conclude that ResNets provide the best-performing classifier architecture achieving an AUC of 77.48%. Among features, LFBs and MFBs usually outperformed MFCCs, with MFBs normally performing slightly better overall. The ensemble and the teacher-student configurations achieved comparable performance, typically both outperforming the single model configuration comfortably, with the teacher-student configuration generally achieving the highest overall performance. Finally, FSS successfully enhanced the performance of key models, increasing the AUC of BiLSTMs from 72.91% to 77.09% and the AUC of ResNet-18s from 74.17% to 77.49%, while reducing the input dimensionality by at least 50% and, therefore, reducing computation.