Faculty of Science 1

Outstanding Doctoral Thesis Award

YANG, Jing

Thesis Title

Discovering Organisational Models from Event Logs for Workforce Analytics

Supervisors

• Associate Professor Chun Ouyang  (Principal Supervisor)
• Dr Wil Van Der Aalst  (External Supervisor)
• Professor Arthur Harry Maria ter Hofstede  (Associate Supervisor)

Citation

This research investigates knowledge discovery from business process execution data to support workforce analytics and improve business process management. It presents a set of novel data-driven methods to systematically construct and analyse organisational models, which can be utilised for guiding organisational structure design and staff deployment toward process improvement. The usefulness of the proposed methods was evaluated through experiments using public datasets from five real-world organisations across three business domains.

Doctor of Philosophy

BURKE, Adam Trent

Thesis Title

Process Mining with Labelled Stochastic Nets

Supervisors

• Dr Sander Jacobus Johannes Leemans  (External Supervisor)
• Professor Moe Kyaw Wynn  (Principal Supervisor)
• Professor Arthur Harry Maria ter Hofstede  (Associate Supervisor)

Citation

This research investigates how organisations behave by analysing the data they produce. Discovering the hidden structure of organisational processes is a step towards understanding and improving them. This thesis presents novel algorithms for discovering and analysing organisational models, particularly those that explicitly include probability. Techniques are investigated formally and experimentally, with new software tools and real-life data. This includes data from modern IT systems and a novel dataset of Qing dynasty civil service records.

DING, Zhe

Thesis Title

Enhanced Genetic Algorithms Towards Virtual Resource Management for Energy Efficiency in Data Centres

Supervisors

• Dr Maolin Tang  (Associate Supervisor)
• Professor Yuchu Tian  (Principal Supervisor)
• You‑Gan Wang  (External Supervisor)

Citation

This thesis deploys advanced virtual-resource-management strategies to reveal the increasing energy consumption of data centres. Due to the weak balance between energy efficiency and computational efficiency of existing strategies, a few enhanced genetic algorithms are designed and implemented for energy-aware virtual resource management, via data structure, fitness function, progressive fidelity, and population control, respectively. Experimental results proves the effectiveness of these implemented genetic algorithms, improving their performance while maintaining well balance between energy efficiency and computational efficiency.

GOSWAMI, Bhargavi

Thesis Title

Novel Communication Strategy for Low Latency and Scalable Neighbourhood Area Networks in Smart Grids

Supervisors

• Professor Raja Jurdak  (Principal Supervisor)
• Dr Ghavameddin Nourbakhsh  (Associate Supervisor)

Citation

The smart grid continues to expand with rising applications and evolving power grid. There is a need for low-latency communication to connect large-scale smart meters for serving real-time and non-real-time applications of smart grids. The novel contribution of this thesis to fulfil these research gaps are a) HCS - Hybrid Communication Strategy for addressing the needs of non-real-time applications of smart grids, b) SDRTN - Software Defined Real-Time Networks, the communication strategy for real-time applications of the smart grid, and c) NAS - Node Allocation Strategy to solve the large-scale low-latency communication in NAN of smart grid.

KANE, Luke Elijah

Thesis Title

Architectures for Performance and Security in IoT Smart Grids

Supervisors

• Dr Yin‑Miao Liu  (Principal Supervisor)
• Dr Matthew Edmund McKague  (Associate Supervisor)
• Associate Professor Geoffrey Ralph Walker  (Associate Supervisor)

Citation

This research examined the relationship between performance and security of Internet of Things (IoT) devices and networks from the perspective of the smart grid. In addition, it presented architectures for evaluating IoT performance, and implementing appropriate security measures. This work was critical as IoT enables many commonly used devices and systems. These can range from smart appliances, all the way through to our national critical infrastructure assets. Through practical experimentation, this study demonstrated that with a balanced approach, IoT networks can be implemented securely and meet application performance requirements.

KNIGHT, Jennine Arlette V

Thesis Title

Transforming Academic Libraries through Strategic Partnerships: The Caribbean Case of the UWI, Cave Hill Campus

Supervisors

• Professor Peter David Bruza  (Associate Supervisor)
• Adjunct Professor Sylvia Lauretta Edwards  (Principal Supervisor)
• Adjunct Professor Bill Fisher  (External Supervisor)

Citation

Challenged by being under-resourced, with an unsupportive institutional and library culture, and inadequate and outdated library spaces, this case study explores opportunities for transformation of Campus Libraries, University of the West Indies, Cave Hill Campus.
70 participants comprising students and staff responded to the LibQUAL+™ survey. The resultant data were analysed using descriptive statistics. Nine participants comprising administrators and academic librarians participated in semi-structured interviews. Their responses were analysed using inductive qualitative content analysis. This thesis offers recommendations for transforming Campus Libraries from warehouses of knowledge to places and spaces for wider stakeholder engagement and strategic partnerships.

ROWE, Benjamin Alan

Thesis Title

Analyzing Long-Duration Recordings of the Environment Using Autoencoder-Extracted Features

Supervisors

• Dr Philip Eichinski  (External Supervisor)
• Professor Paul Roe  (Associate Supervisor)
• Dr Jinglan Zhang  (Principal Supervisor)

Citation

This research addresses a critical challenge in ecological monitoring: efficiently analysing vast amounts of audio data for biodiversity assessment. It recognises the limitations of traditional methods reliant on extensive human labelling of data and explores the promising potential of autoencoders, a deep learning technique. An innovative framework for automatically extracting meaningful representations from unlabeled ecological audio is proposed, which can then be used for analysis.

SILVA, Mapulage Don Rangika Lilukshi

Thesis Title

A Parallel Approach to Centralised Multi-Vehicle Dynamic Route Guidance with Predictive Traffic Modelling

Supervisors

• Dr Marc Miska  (Associate Supervisor)
• Dr Maolin Tang  (Principal Supervisor)
• Professor Yuchu Tian (Associate Supervisor)

Citation

This thesis proposes an innovative centralised multi-vehicle dynamic route guidance framework, addressing urban travel challenges. Incorporating proactive dynamic routing and a traffic prediction model, it optimises average travel time while maintaining individual efficiency. Scalable parallelisation ensures effectiveness in large networks, reducing execution times. Simulations showcase superior performance over benchmarks, demonstrating real-world applicability. This comprehensive approach contributes significantly to urban transportation optimisation.

TENNAKOON MUDIYANSELAGE, Gayani Tennakoon

Thesis Title

Knowledge Discovery from Social Networks Using Interaction Frequency and User Hierarchy

Supervisors

• Adjunct Professor Shlomo Geva  (Associate Supervisor)
• Professor Richi Nayak  (Principal Supervisor)

Citation

The immense popularity of social media creates a great opportunity for businesses, citizens, politicians, and other services to reach society. Graph mining techniques are commonly used for knowledge discovery in social networks to unveil intricate structural patterns among users. However, the increasing size and complexity of social networks demand effective computational methods that can mine novel patterns. This thesis aims to incorporate two inherent properties of social networks, powerlaw interaction frequency distribution, and hierarchical organisation of users for discovering the knowledge of prominent users and communities in the network. This thesis proposes effective computational methods to extract this knowledge.

VELMURUGAN, Mythreyi

Thesis Title

Evaluating Post Hoc Explanations Using Tabular Data: A Functionally-Grounded Perspective

Supervisors

• Associate Professor Chun Ouyang  (Principal Supervisor)
• Adjunct Associate Professor Catarina Alexandra Pinto Moreira  (External Supervisor)
• Dr Renuka Sindhgatta Rajan  (External Supervisor)
• Associate Professor Yue Xu  (Associate Supervisor)

Citation

The complex models generated by modern machine learning techniques lack transparency, and require specialised interpretation algorithms to ensure that they are used safely and responsibly. This thesis investigates post-hoc algorithms used to explain the decision-making of machine learning models, and how the characteristics of the datasets used affect explanation quality. Although datasets in tabular format are ubiquitous in data science, the unique characteristics of tabular data and their effect on machine learning explainability are underexplored. This thesis explores what makes a good explanation, how explanations can be assessed and evaluates how well popular XAI techniques perform on tabular data.

WHEELER, Helvi

Thesis Title

Exploring and Co-Designing Emerging Technologies to Enhance the Reading Experience of Namibian Children

Supervisors

• Professor Margot Brereton  (Principal Supervisor)
• Associate Professor Alessandro Soro  (Associate Supervisor)
• Heike Winschiers‑Theophilus  (External Supervisor)

Citation

Reading plays a vital role in our daily lives. The aim of this research is to address the lack of reading culture among primary school learners in Namibia and identify suitable co-design techniques for reading technologies that enhance the reading experiences for children. Different reading technologies were created such as: Augmented Book, Collaborative Reading Games, and Scape Tale. Through exploring various co-designs, approaches and evaluation of different reading technologies, this study uncovers important features that should be integrated into reading technology games. These features foster embodiment, agency, collaboration, and consideration of emotions.

YEH, Yi‑Ting

Thesis Title

The Value of Big Data Analytics (BDA) Use: A Conceptualisation of the Multilevel BDA-Use-Value Framework

Supervisors

• Dr Rebekah Eden  (External Supervisor)
• Dr Erwin Johannes Fielt  (Principal Supervisor)
• Dr Rehan Abbas Syed  (Associate Supervisor)

Citation

While big data analytics (BDA) has gained traction from health organisations to leverage its analytical capabilities to enhance performance, many BDA investments fail to realise their expected value. Therefore, this thesis aims to address how the use of BDA results in business value in a digital hospital setting. The research used a systematic literature review to develop a conceptual framework and further explored the framework using two detailed case studies. Subsequently, through further theorising the findings, a Use-Learning-Outcomes Theory for BDA was developed that explains the challenges associated with creating value from the use of BDA in health organisations.