Machine Learning Algorithms for Predictive Analysis
This study examines supervised and unsupervised machine learning algorithms and their application to predictive modelling across real-world datasets. It evaluates how different algorithmic approaches — including regression, classification, clustering, and ensemble methods — perform under varied data conditions. The research also investigates optimisation techniques for improving model accuracy and generalisation.
Background
Machine learning has transformed how organisations extract insights from data, enabling predictions in healthcare, finance, logistics, and beyond. As data volumes grow exponentially, the demand for efficient and accurate predictive models has made algorithm selection and tuning a critical research area with broad practical implications.
Research Problem
Despite the wide availability of ML algorithms, selecting the most appropriate model for a given dataset and task remains a significant challenge. Overfitting, class imbalance, feature selection, and computational constraints frequently undermine model performance in real-world deployments.
Objectives of the Study
- Evaluate the performance of key ML algorithms on benchmark datasets
- Identify factors that influence algorithm selection and model accuracy
- Apply optimisation techniques to improve predictive performance
- Propose a selection framework for practitioners based on dataset characteristics
Blockchain Technology in Supply Chain Management
This research explores the integration of blockchain technology into supply chain management to enhance transparency, traceability, and operational efficiency. It examines how distributed ledger systems enable end-to-end visibility of goods from production to delivery. The study also assesses adoption challenges faced by businesses in different industries.
Background
Global supply chains are increasingly vulnerable to fraud, counterfeiting, and inefficiency due to reliance on fragmented, opaque record-keeping systems. Blockchain offers an immutable, decentralised ledger that can address these vulnerabilities, making it a promising solution for industries where provenance and accountability are critical.
Research Problem
Traditional supply chain management systems suffer from lack of transparency, data silos, and susceptibility to manipulation. While blockchain promises to resolve these issues, its real-world adoption remains limited due to scalability concerns, cost, and integration complexity.
Objectives of the Study
- Assess current inefficiencies in traditional supply chain record-keeping
- Evaluate blockchain's effectiveness in improving traceability and transparency
- Identify adoption barriers and success factors in blockchain-based supply chains
- Propose a practical implementation framework for industry stakeholders
Cybersecurity Threats in Cloud Computing Environments
This study conducts a systematic review of cybersecurity threats targeting cloud computing environments, with a focus on multi-tenant architectures. It analyses prevalent attack vectors including data breaches, insider threats, misconfigurations, and denial-of-service attacks. The research further evaluates mitigation strategies and their effectiveness in enterprise cloud deployments.
Background
Cloud adoption has accelerated rapidly across industries, but this shift introduces new security challenges that traditional perimeter-based defences are ill-equipped to address. The shared responsibility model in cloud environments creates ambiguity around security ownership, increasing exposure to sophisticated attacks.
Research Problem
As organisations migrate critical workloads to the cloud, the frequency and severity of cloud-specific cyberattacks continue to rise. Existing security frameworks often lag behind evolving threat landscapes, leaving cloud deployments vulnerable to exploitation.
Objectives of the Study
- Identify and categorise the most prevalent cybersecurity threats in cloud environments
- Evaluate the effectiveness of existing mitigation and detection strategies
- Analyse the shared responsibility model and its security implications
- Recommend a comprehensive security framework for cloud-based enterprise systems
IoT-Based Smart Home Automation Systems
This project designs and implements an IoT-based smart home automation system using Arduino microcontrollers and the MQTT communication protocol. The system enables remote monitoring and control of household devices including lighting, temperature, and security sensors. It evaluates the system's responsiveness, energy efficiency, and user experience.
Background
Smart home technology is rapidly moving from luxury to mainstream as IoT devices become more affordable and accessible. Automation systems that enhance convenience, energy savings, and home security are increasingly important in urban living, making IoT-based solutions a key area of applied research.
Research Problem
Many existing smart home solutions are proprietary, expensive, and poorly interoperable. Low-cost, open-source implementations using platforms like Arduino remain underexplored in academic literature, particularly regarding performance benchmarking and real-world reliability.
Objectives of the Study
- Design a functional IoT smart home system using Arduino and MQTT
- Implement remote monitoring and control for multiple household devices
- Evaluate system performance in terms of latency, reliability, and energy use
- Assess user experience and propose improvements for practical deployment
Deep Learning for Medical Image Classification
This research investigates the use of deep learning, particularly convolutional neural networks (CNNs), for automated classification of medical images including X-rays, MRI scans, and histopathology slides. It benchmarks multiple CNN architectures against standard radiologist performance metrics. The study also addresses challenges such as limited labelled data and class imbalance in clinical datasets.
Background
Medical imaging diagnosis is time-intensive and subject to human error, particularly in resource-limited healthcare settings. Deep learning models have demonstrated near-expert accuracy in detecting conditions such as cancer, pneumonia, and diabetic retinopathy, positioning AI as a transformative tool in diagnostic radiology.
Research Problem
The deployment of AI-based medical image classifiers is hindered by data scarcity, interpretability concerns, and regulatory uncertainty. Without robust, explainable models that clinicians can trust, widespread clinical adoption remains limited.
Objectives of the Study
- Design and train CNN models for classification of multiple medical imaging modalities
- Compare model performance against benchmark datasets and clinical accuracy standards
- Address class imbalance and data augmentation strategies in medical datasets
- Evaluate model interpretability using explainability techniques such as Grad-CAM
Comparative Analysis of Agile Methodologies in Software Projects
This study provides a comparative analysis of three widely-used Agile methodologies — Scrum, Kanban, and Extreme Programming (XP) — in the context of software project delivery. It measures key performance indicators including sprint velocity, defect rates, team satisfaction, and time-to-release. The research draws on case studies and primary survey data from development teams.
Background
Agile has replaced Waterfall as the dominant software development paradigm in most organisations. Yet the choice between Agile frameworks is often made without empirical evidence, leading to mismatches between team needs and process structure that ultimately affect project outcomes.
Research Problem
While all three Agile methodologies aim to improve software delivery, their differing structures lead to distinct trade-offs in productivity, quality, and team dynamics. Organisations often lack the evidence needed to select the most suitable framework for their context.
Objectives of the Study
- Compare Scrum, Kanban, and XP across key software project performance metrics
- Assess the impact of each methodology on team satisfaction and communication
- Identify contextual factors that influence the suitability of each Agile framework
- Provide evidence-based guidance for Agile methodology selection
Natural Language Processing for Sentiment Analysis
This research develops transformer-based natural language processing (NLP) models for sentiment analysis of social media content. It investigates how models such as BERT and RoBERTa can be fine-tuned to classify opinions, emotions, and attitudes expressed in user-generated text. The study also evaluates performance across multiple languages and domains.
Background
Social media platforms generate billions of text posts daily, representing a rich source of public opinion data. Accurate sentiment analysis enables organisations to monitor brand perception, detect emerging issues, and understand consumer behaviour in near real-time, creating high demand for robust NLP pipelines.
Research Problem
General-purpose sentiment models often underperform on domain-specific or informal social media language, which contains slang, abbreviations, and sarcasm. Adapting large language models to these challenges requires specialised training strategies and evaluation frameworks.
Objectives of the Study
- Build and fine-tune transformer-based models for social media sentiment classification
- Evaluate model performance on multi-domain and multilingual datasets
- Address challenges of informal language, irony, and code-switching in NLP
- Compare transformer models against traditional ML baselines on accuracy and efficiency
E-Government Systems Adoption in Developing Countries
This study examines the adoption of e-government systems in developing countries, with particular focus on Sub-Saharan Africa. Using the Technology Acceptance Model (TAM) as a theoretical framework, it investigates barriers and enablers of digital governance uptake among citizens and public servants. The research combines survey data and case study analysis.
Background
E-government initiatives promise to improve public service delivery, reduce corruption, and enhance transparency in governance. However, adoption rates in developing regions remain significantly lower than in developed nations, due to infrastructure deficits, digital literacy gaps, and institutional resistance.
Research Problem
Despite significant investment in e-government platforms across Africa, many systems remain underutilised. The factors that determine successful adoption versus failure are poorly understood, limiting the effectiveness of future policy interventions.
Objectives of the Study
- Assess the current state of e-government adoption across selected Sub-Saharan African countries
- Apply the TAM framework to identify key adoption drivers and barriers
- Evaluate the role of infrastructure, digital literacy, and trust in system uptake
- Recommend strategies to improve e-government adoption and sustainability
Database Optimisation Techniques for Big Data Applications
This research evaluates database optimisation techniques for handling big data workloads, comparing indexing strategies, query planning approaches, and the performance trade-offs between NoSQL and relational database management systems (RDBMS). It applies benchmarking methodologies to assess throughput, latency, and scalability under high-volume data conditions.
Background
The explosion of digital data from IoT devices, social media, and enterprise systems has placed unprecedented demands on database infrastructure. Traditional RDBMS solutions struggle to scale efficiently, driving the adoption of NoSQL alternatives — yet the performance implications of this shift are context-dependent and not well understood.
Research Problem
Choosing the right database technology and optimisation strategy for big data applications is a complex, high-stakes decision. Poorly optimised databases lead to performance bottlenecks, increased costs, and degraded user experience in data-intensive systems.
Objectives of the Study
- Benchmark NoSQL and RDBMS performance across read-heavy, write-heavy, and mixed workloads
- Evaluate indexing strategies and their impact on query execution time
- Analyse query planning and cost estimation in modern database engines
- Propose optimisation guidelines for database architects working with big data systems
Mobile Application Development for Healthcare Delivery
This project designs and evaluates a cross-platform mobile application for healthcare delivery, focusing on remote patient monitoring and teleconsultation capabilities. It investigates the usability, security, and clinical effectiveness of mobile health (mHealth) systems in bridging healthcare access gaps. The study incorporates user testing with both patients and healthcare providers.
Background
Mobile technology has the potential to revolutionise healthcare delivery, particularly in regions with limited access to medical facilities. Smartphone penetration is outpacing fixed healthcare infrastructure in many developing countries, making mHealth applications a critical vehicle for expanding care reach.
Research Problem
Many existing healthcare applications are designed for resource-rich environments and fail to account for low-bandwidth connectivity, varied digital literacy, or the security requirements of sensitive medical data. There is a need for accessible, secure, and clinically validated mobile health solutions.
Objectives of the Study
- Design a cross-platform mHealth app for remote monitoring and teleconsultation
- Evaluate the app's usability with patients and healthcare providers
- Assess data security and privacy compliance in mobile health systems
- Measure the impact of the app on healthcare access and patient satisfaction
Quantum Computing Implications for Cryptography
This research investigates the implications of quantum computing for modern cryptographic systems, with a focus on the vulnerability of widely-used public-key algorithms such as RSA and ECC to quantum attacks. It evaluates post-quantum cryptographic algorithms proposed by NIST and assesses their practical feasibility for real-world deployment at scale.
Background
Quantum computers, once fully realised, will render many current encryption standards obsolete through algorithms like Shor's, which can factorise large integers exponentially faster than classical computers. The global shift to quantum-safe cryptography is now an urgent priority for governments, financial institutions, and technology providers.
Research Problem
The transition to post-quantum cryptography is technically complex and logistically challenging. Most organisations lack the expertise and roadmap to migrate their cryptographic infrastructure before quantum computing reaches practical capability, creating a significant security window.
Objectives of the Study
- Analyse the threat posed by quantum computing to current cryptographic standards
- Evaluate NIST post-quantum algorithm candidates on performance and security metrics
- Assess the feasibility of quantum-safe migration for enterprise cryptographic systems
- Propose a phased implementation roadmap for post-quantum cryptography adoption
Neural Network Architectures for Time Series Forecasting
This study evaluates neural network architectures — specifically LSTM, GRU, and Temporal Convolutional Networks (TCN) — for time series forecasting tasks including stock price prediction and demand planning. It compares these architectures against traditional forecasting models such as ARIMA and examines their sensitivity to hyperparameter tuning and data quality.
Background
Accurate time series forecasting is critical across sectors including finance, retail, energy, and healthcare. The rise of deep learning has introduced powerful sequence modelling capabilities that outperform classical statistical methods on complex, non-linear temporal patterns — yet model selection and interpretability remain open challenges.
Research Problem
Deep learning models for time series are notoriously difficult to tune, and their performance gains over simpler models are often inconsistent across domains. Practitioners frequently lack the guidance needed to select, configure, and validate these models for specific forecasting tasks.
Objectives of the Study
- Compare LSTM, GRU, and TCN architectures on multiple time series forecasting tasks
- Benchmark deep learning models against ARIMA and exponential smoothing baselines
- Analyse the effect of lookback window, layer depth, and regularisation on forecast accuracy
- Develop guidelines for deep learning model selection in time series forecasting
Ethical Implications of Artificial Intelligence in Decision Making
This research critically examines the ethical implications of artificial intelligence in high-stakes decision-making domains including healthcare, criminal justice, and human resources. It investigates algorithmic bias, lack of explainability, and accountability gaps in AI systems deployed in these areas. The study proposes an ethical governance framework for responsible AI deployment.
Background
AI systems are increasingly making or influencing decisions that have profound effects on people's lives — from clinical diagnoses to loan approvals and parole decisions. The opacity and potential bias of these systems raise serious ethical concerns that existing regulatory frameworks have struggled to adequately address.
Research Problem
Algorithmic bias and the "black box" nature of many AI models undermine public trust and can perpetuate systemic inequalities. Without clear accountability structures and interpretability standards, AI deployments risk causing harm at scale in critical societal domains.
Objectives of the Study
- Identify and categorise sources of algorithmic bias in AI decision-making systems
- Evaluate explainability techniques and their effectiveness in high-stakes domains
- Analyse existing AI governance frameworks for adequacy in addressing ethical risks
- Propose a comprehensive ethical framework for responsible AI development and deployment
Network Intrusion Detection Using Machine Learning
This research designs and evaluates machine learning-based network intrusion detection systems (IDS) using the CICIDS benchmark dataset. It trains and compares multiple classifiers — including Random Forest, SVM, and deep neural networks — for real-time detection of network anomalies and attack signatures. The study also addresses the challenge of imbalanced traffic class distributions.
Background
Cyber threats have evolved in sophistication and frequency, making traditional signature-based intrusion detection inadequate. Machine learning offers the ability to identify novel attack patterns from network traffic data, making adaptive, intelligent IDS a critical component of modern cybersecurity infrastructure.
Research Problem
Building effective ML-based IDS that generalise across diverse network environments remains challenging. High false positive rates, class imbalance in labelled datasets, and real-time processing constraints continue to limit the practical deployment of intelligent intrusion detection systems.
Objectives of the Study
- Train and evaluate ML classifiers on CICIDS intrusion detection datasets
- Compare model performance on accuracy, precision, recall, and F1-score
- Address class imbalance using SMOTE and other resampling techniques
- Assess the feasibility of deploying ML-based IDS in real-time network environments
Augmented Reality in Educational Technology
This study investigates the impact of augmented reality (AR) learning tools on student engagement, knowledge retention, and academic performance in secondary and higher education settings. It evaluates AR-enhanced educational experiences against traditional instruction methods using controlled experimental designs. The research also explores implementation barriers and pedagogical best practices.
Background
Immersive technologies like AR are increasingly being integrated into educational environments as institutions seek innovative ways to improve learning outcomes. AR allows learners to interact with three-dimensional content overlaid on the real world, creating engaging and contextualised learning experiences that static media cannot replicate.
Research Problem
While AR shows promise in educational contexts, rigorous empirical evidence of its impact on long-term knowledge retention and academic performance remains limited. Additionally, concerns about cost, teacher training, and equitable access present real barriers to widespread adoption.
Objectives of the Study
- Evaluate the effect of AR tools on student engagement and knowledge retention
- Compare AR-based learning outcomes against traditional instructional methods
- Identify barriers to AR adoption in educational institutions
- Recommend pedagogical strategies for effective AR integration into curricula
Cloud Migration Strategies for Enterprise Applications
This research analyses and compares three primary cloud migration strategies — lift-and-shift, refactoring, and full cloud-native rebuilding — for enterprise applications. It evaluates each strategy across dimensions of cost, risk, performance, and long-term maintainability. The study draws on real-world case studies from organisations that have undergone cloud migration projects.
Background
Organisations worldwide are accelerating their migration to cloud infrastructure to reduce costs, improve scalability, and modernise legacy systems. However, cloud migration projects frequently overrun budgets, face unexpected technical debt, and fail to deliver anticipated benefits without a well-considered strategy.
Research Problem
The selection of a cloud migration strategy is highly consequential yet often made without sufficient analysis of trade-offs. Lift-and-shift may preserve technical debt, while full rebuilds are expensive and risky — yet organisations lack structured frameworks for making this decision.
Objectives of the Study
- Define and compare lift-and-shift, refactoring, and cloud-native rebuilding strategies
- Evaluate each strategy across cost, risk, performance, and maintenance dimensions
- Analyse real-world case studies of enterprise cloud migration projects
- Develop a decision framework to guide cloud migration strategy selection
Web Accessibility Standards and Compliance in Nigerian Universities
This study audits the compliance of Nigerian university websites with Web Content Accessibility Guidelines (WCAG) 2.1, identifying specific accessibility failures and their impact on users with disabilities. It investigates the institutional, technical, and awareness barriers that prevent compliance and provides actionable recommendations for remediation.
Background
Web accessibility is both a legal obligation and a moral imperative, ensuring that people with visual, auditory, cognitive, and motor impairments can access digital content. Nigerian higher education institutions have increasingly digitised their services, yet accessibility compliance in this context remains largely unexamined.
Research Problem
A significant proportion of university websites in Nigeria fail to meet basic WCAG 2.1 accessibility standards, creating barriers for students and staff with disabilities. Without systematic audits and institutional accountability, these barriers are unlikely to be addressed.
Objectives of the Study
- Audit selected Nigerian university websites for WCAG 2.1 compliance
- Categorise and quantify accessibility violations by type and severity
- Identify institutional and technical barriers to web accessibility compliance
- Recommend a practical remediation roadmap for higher education web portals
Federated Learning for Privacy-Preserving Data Analysis
This research investigates federated learning as a privacy-preserving machine learning paradigm, enabling model training across distributed data sources without centralising sensitive data. It evaluates federated learning frameworks against centralised ML baselines on accuracy, communication efficiency, and data privacy guarantees. The study also examines differential privacy techniques as a complementary safeguard.
Background
Data privacy regulations such as GDPR and increasing public awareness of surveillance risks have made centralised ML training problematic in sensitive domains like healthcare and finance. Federated learning offers a compelling alternative by keeping data at the source and sharing only model updates, fundamentally decoupling learning from data centralisation.
Research Problem
Federated learning introduces new challenges including heterogeneous data distributions across clients, increased communication overhead, and vulnerability to adversarial model poisoning attacks. Achieving both high model accuracy and strong privacy guarantees simultaneously remains an active research problem.
Objectives of the Study
- Implement and evaluate federated learning frameworks on distributed private datasets
- Compare federated model accuracy and convergence against centralised training baselines
- Assess the impact of differential privacy on model performance and privacy guarantees
- Analyse communication costs and propose optimisation strategies for federated systems
Social Media Analytics for Business Intelligence
This study applies social media analytics techniques to extract business intelligence from Twitter and Instagram data, focusing on consumer trend forecasting and brand sentiment analysis. It develops and evaluates data mining pipelines that process large volumes of user-generated content to generate actionable insights for marketing and product strategy.
Background
Social media has become one of the richest sources of real-time consumer opinion data available to businesses. Companies that effectively harness this data gain competitive advantages in brand management, product development, and marketing campaign targeting, making social media analytics a high-value capability.
Research Problem
Extracting reliable, actionable insights from noisy, unstructured social media data is technically challenging. Existing analytics tools often lack contextual accuracy, multilingual capability, and the ability to distinguish genuine consumer sentiment from bots and spam.
Objectives of the Study
- Develop a social media data mining pipeline for Twitter and Instagram
- Apply NLP techniques for sentiment classification and trend detection
- Evaluate the accuracy and reliability of extracted business intelligence insights
- Propose a social media analytics framework for business intelligence applications
Autonomous Vehicle Navigation Algorithms
This research investigates navigation algorithms for autonomous vehicles, including Simultaneous Localisation and Mapping (SLAM), path planning, and obstacle avoidance techniques. It evaluates the performance of these algorithms using LIDAR and camera sensor fusion in simulated and real-world driving environments. The study also addresses edge cases in adverse weather and complex urban scenarios.
Background
Autonomous vehicle technology promises to radically reduce road accidents, improve transport efficiency, and expand mobility access. Navigation and perception remain the most technically complex challenges in autonomous driving, with current systems still falling short of the robustness required for fully unsupervised operation.
Research Problem
Existing navigation algorithms perform reliably in controlled conditions but degrade significantly in dynamic, unpredictable environments. Achieving the level of safety and reliability required for commercial autonomous vehicle deployment requires significant advances in sensor fusion, real-time processing, and algorithmic robustness.
Objectives of the Study
- Implement and evaluate SLAM and path planning algorithms in simulated AV environments
- Assess obstacle detection accuracy using LIDAR-camera sensor fusion
- Test algorithm performance under adverse conditions including low visibility and dynamic obstacles
- Propose improvements to navigation robustness for real-world autonomous vehicle deployment
Microservices Architecture vs Monolithic Systems
This research conducts a comparative evaluation of microservices architecture and monolithic system design for enterprise applications, analysing trade-offs in scalability, resilience, deployment complexity, and team autonomy. It draws on case studies from technology organisations that have undergone architecture transitions and applies quantitative performance benchmarking.
Background
The shift from monolithic to microservices architecture has been widely adopted by technology companies seeking greater deployment flexibility and independent scalability. However, this transition introduces significant complexity in service orchestration, distributed data management, and inter-service communication.
Research Problem
Many organisations undertake microservices migrations without a clear understanding of the associated complexity and cost. Conversely, others retain monolithic architectures that limit their scalability and agility. Evidence-based guidance on architecture selection remains limited.
Objectives of the Study
- Benchmark microservices and monolithic architectures on performance and scalability metrics
- Evaluate deployment complexity, failure modes, and resilience in both approaches
- Analyse the organisational and team dynamics implications of each architecture
- Develop an evidence-based framework for enterprise architecture decision-making
Digital Forensics and Incident Response Frameworks
This study examines digital forensics methodologies and incident response frameworks used to investigate cyber incidents. It covers chain of custody procedures, digital evidence acquisition, memory and disk forensics, and post-breach analysis techniques. The research evaluates the effectiveness of established frameworks including NIST and SANS in guiding real-world investigations.
Background
Cybercrime is growing in scale and sophistication, creating urgent demand for qualified digital forensics professionals and robust incident response capabilities. Organisations that lack structured forensic and response frameworks face extended breach dwell times, regulatory penalties, and reputational damage.
Research Problem
Digital forensic investigations are often hindered by jurisdictional complexities, encrypted evidence, anti-forensic techniques employed by attackers, and the sheer volume of data generated by modern IT environments. Existing frameworks require continuous updating to address emerging threat actors and technologies.
Objectives of the Study
- Review and compare established digital forensics and incident response frameworks
- Evaluate chain of custody and evidence integrity procedures in cyber investigations
- Analyse real-world case studies of digital forensic investigations
- Recommend best practices for building organisational digital forensics capabilities
5G Network Deployment Challenges in Rural Areas
This research investigates the technical, economic, and regulatory challenges impeding 5G network deployment in rural and underserved areas. It analyses spectrum allocation policies, infrastructure cost models, and the role of government and private sector partnerships in closing the digital divide through next-generation connectivity.
Background
5G technology promises transformative improvements in network speed, latency, and device density — but its benefits are predominantly being realised in urban centres. Rural communities risk being left further behind the digital frontier without targeted strategies to address the structural barriers to 5G rollout in low-density areas.
Research Problem
The economics of 5G deployment in rural areas are challenging due to high infrastructure costs relative to low population density, limited spectrum availability, and regulatory fragmentation. Without intervention, the 5G divide will exacerbate existing socioeconomic inequalities.
Objectives of the Study
- Assess the current state of 5G deployment and coverage gaps in rural areas
- Analyse spectrum allocation policies and their impact on rural connectivity
- Evaluate public-private partnership models for rural 5G infrastructure investment
- Propose a policy framework to accelerate equitable 5G deployment
Open Source Software Adoption in Public Sector Organisations
This study investigates the adoption of open source software (OSS) in public sector organisations, examining the cost savings, security considerations, and transition barriers associated with replacing proprietary systems in government IT environments. It draws on comparative case studies from multiple countries to identify success factors and lessons learnt.
Background
Governments increasingly recognise the fiscal and strategic value of open source software, including lower licensing costs, vendor independence, and transparent security auditing. Several countries have established formal OSS adoption policies, yet implementation challenges continue to limit full realisation of these benefits.
Research Problem
Public sector OSS adoption is frequently undermined by organisational inertia, procurement practices designed for proprietary vendors, skills gaps, and concerns about long-term support and security. The gap between policy intent and operational reality remains significant.
Objectives of the Study
- Assess the drivers and barriers of OSS adoption in government organisations
- Evaluate cost savings, security outcomes, and service continuity in OSS transitions
- Analyse case studies of successful and failed public sector OSS implementations
- Recommend a framework for sustainable open source adoption in government IT
Data Privacy Regulations and Compliance (GDPR / NDPA)
This research conducts a comparative analysis of data privacy regulatory frameworks, with a focus on the EU's General Data Protection Regulation (GDPR) and Nigeria's Nigeria Data Protection Act (NDPA). It examines their impact on technology businesses, data processing practices, and individual rights, and evaluates the effectiveness of enforcement mechanisms.
Background
Data privacy has emerged as one of the most consequential areas of technology regulation globally. With increasing data collection by businesses and governments, strong regulatory frameworks are essential to protect citizens' rights and ensure responsible data stewardship. Nigeria's recent NDPA marks a significant step towards continental data protection harmonisation.
Research Problem
Many organisations, particularly SMEs and startups, struggle to achieve full compliance with data protection regulations due to complexity, cost, and lack of guidance. Cross-border data flows create additional compliance challenges when multiple frameworks apply simultaneously.
Objectives of the Study
- Compare the scope, principles, and enforcement mechanisms of GDPR and NDPA
- Assess the compliance burden on technology businesses operating under both frameworks
- Evaluate the effectiveness of data protection authorities in enforcing regulations
- Propose practical compliance strategies for organisations operating under multiple data privacy regimes
Human-Computer Interaction Design for Elderly Users
This research applies accessibility-first design principles to the development of digital products intended for elderly users aged 65 and above. It investigates cognitive, visual, and motor challenges that affect older adults' interaction with digital interfaces and evaluates design interventions that improve usability and adoption among this demographic.
Background
As populations age globally, digital inclusion of elderly users is becoming an important social and economic priority. Older adults represent a rapidly growing segment of internet users, yet mainstream digital product design consistently fails to accommodate the specific interaction needs of this group, creating barriers to access and participation.
Research Problem
Existing HCI guidelines and accessibility standards do not fully address the nuanced needs of elderly users, particularly those with age-related cognitive and sensory decline. Most digital products are designed with younger users as the primary audience, resulting in interfaces that are difficult, frustrating, or impossible for older adults to use.
Objectives of the Study
- Review HCI design frameworks and accessibility standards relevant to elderly users
- Conduct usability testing of existing digital products with elderly participants
- Identify specific design failures and their impact on elderly user experience
- Develop and evaluate accessibility-first design guidelines for elderly-centred digital products
Predictive Maintenance Using Industrial IoT Sensors
This study develops and evaluates machine learning models for predictive maintenance in industrial environments, using data from IoT sensors monitoring machinery health indicators such as vibration, temperature, and pressure. It compares anomaly detection and classification approaches for predicting equipment failure before it occurs, reducing unplanned downtime.
Background
Unplanned equipment failures cost manufacturing industries billions of dollars annually through downtime, emergency repairs, and lost productivity. Predictive maintenance, enabled by IoT sensors and AI, allows organisations to shift from reactive or schedule-based maintenance to condition-based interventions precisely when needed.
Research Problem
Implementing effective predictive maintenance systems requires high-quality labelled fault data, which is often scarce in operational environments. Additionally, integrating ML models with existing industrial control systems presents significant technical and organisational challenges.
Objectives of the Study
- Collect and preprocess IoT sensor data from industrial equipment
- Develop and compare ML models for anomaly detection and failure prediction
- Evaluate model performance against real-world maintenance records
- Propose an implementation framework for deploying predictive maintenance in manufacturing
Real-Time Object Detection with YOLO Architecture
This research implements and evaluates the YOLOv8 (You Only Look Once) architecture for real-time object detection in high-speed surveillance and autonomous systems applications. It benchmarks detection accuracy, inference speed, and robustness across multiple deployment environments including edge devices and cloud GPU instances.
Background
Real-time object detection is a foundational computer vision task powering applications from autonomous driving to security surveillance and industrial quality control. YOLO architectures have consistently achieved state-of-the-art results by framing object detection as a single regression problem, enabling much faster inference than two-stage detectors.
Research Problem
Deploying high-accuracy real-time object detection on resource-constrained edge hardware is challenging due to the computational demands of deep convolutional models. Balancing detection accuracy with inference speed and power consumption remains an active optimisation problem.
Objectives of the Study
- Implement YOLOv8 for multi-class object detection on benchmark datasets
- Benchmark detection accuracy (mAP) and inference speed across deployment environments
- Evaluate model performance on edge hardware with varying computational constraints
- Optimise the model for real-time deployment using quantisation and pruning techniques
Recommender Systems for E-Commerce Platforms
This study designs and evaluates recommender systems for e-commerce platforms, comparing collaborative filtering, content-based, and hybrid approaches. It investigates how recommendation quality affects user engagement and conversion rates, and examines the cold-start problem and scalability challenges in large-scale product catalogues.
Background
Recommender systems are a cornerstone of modern e-commerce, driving a significant proportion of sales on major platforms by surfacing relevant products to users at the right moment. As product catalogues grow and user behaviour becomes more complex, the accuracy and efficiency of recommendation engines has a direct impact on revenue and customer satisfaction.
Research Problem
Many e-commerce recommender systems struggle with the cold-start problem for new users or products, sparsity in user-item interaction matrices, and the trade-off between recommendation accuracy and diversity. Existing evaluation metrics also do not always align well with business objectives.
Objectives of the Study
- Implement collaborative filtering, content-based, and hybrid recommender algorithms
- Evaluate recommendation quality using precision, recall, NDCG, and diversity metrics
- Address the cold-start problem using contextual and demographic features
- Assess the business impact of recommendation quality on user engagement and conversion
Virtual Reality Applications in Military Training Simulations
This research evaluates the effectiveness of virtual reality (VR) simulations in military training contexts, examining how immersion fidelity, scenario realism, and motion sickness mitigation affect training outcomes. It compares VR-based training to conventional simulation methods across measures of skill transfer, situational awareness, and trainee stress responses.
Background
Military training demands high-fidelity, repeatable scenarios that are safe, cost-effective, and measurable. VR technology offers the potential to create immersive combat training environments that can be deployed at scale without the logistical and safety costs of live exercises, making it a growing area of defence research and investment.
Research Problem
Despite its promise, VR military training faces limitations including cybersickness in extended sessions, the cost of high-fidelity hardware, and questions about the degree to which VR-acquired skills transfer to real operational environments. Rigorous empirical evaluation is needed to guide investment decisions.
Objectives of the Study
- Evaluate VR training effectiveness on military skill acquisition and knowledge transfer
- Assess the impact of immersion fidelity and scenario realism on training outcomes
- Measure cybersickness incidence and evaluate mitigation strategies in extended VR sessions
- Compare VR and conventional simulation training on cost-effectiveness and performance metrics