Theses and Dissertations

Permanent URI for this collection

http://196.216.167.205/handle/10628/948

Browse

Now showing 1 - 20 of 56

Predictive Modelling of Taxpayer Compliance Behaviour Using Machine Learning At NamRA
(Namibia University of Science and Technology, 2025-07-31) Sifani, Edwin Sifani
This study explored the application of machine learning techniques for predictive modelling of taxpayer compliance behaviour at the Namibia Revenue Agency (NamRA). Multiple classification algorithms were systematically optimised using 5-fold cross-validated Grid Search and Randomised Search, as implemented in the scikit-learn library (v1.2), to enhance predictive accuracy. The hyperparameter search spaces were tailored to each model’s architecture; for instance, Random Forest optimisation included the number of estimators and maximum depth, while Gradient Boosting models emphasised learning rate and structural parameters. The optimisation process yielded notable improvements, with cross-validated accuracy scores ranging from 64% to 68%. The best-performing model, an optimised Random Forest classifier, achieved an accuracy of 68%. These findings demonstrated the efficacy of hyperparameter tuning in improving model performance and underscore the potential of machine learning to support data-driven compliance management at NamRA. The use of SHAP and LIME in this study provided valuable interpretability of taxpayer compliance predictions, highlighting key factors such as income group, the COVID-19 period, taxpayer registration office, and marital status. These insights align with existing research and reveal how financial capacity, macroeconomic disruptions, and administrative or demographic variables influence compliance. SHAP offers a global view of feature importance, while LIME provides personalised explanations, enhancing trust and communication. Despite modest predictive accuracy, the interpretability benefits support targeted policy interventions and suggest future improvements through richer data and fairness assessments.
Developing A Data-Driven Financial Model for Decision Support in Evaluating Investment Portfolio Performance
(Namibia University of Science and Technology, 2025-09-12) Nghilundwa, Pendapala
Contemporary financial markets demonstrate heightened complexity and volatility, necessitating sophisticated instruments for the precise assessment of investment portfolios. This research explores the application of machine learning (ML) models to predict Month-to-Date (MTD) returns, aiming to enhance financial decision-making. Conventional models frequently exhibit insufficient flexibility to fluctuating market conditions, highlighting the necessity for data-driven approaches that prioritise portfolio-specific metrics, such as Market Value Dirty and Year-to-Date (YTD) returns. Meanwhile, macroeconomic variables such as gross domestic product (GDP) growth, inflation, and interest rates played a secondary role. The study employed a quantitative method, using secondary data from January 2017 to August 2024, which comprised financial measures and macroeconomic variables. Four machine learning models were developed, namely Random Forest, Gradient Boosting, XGBoost and Long Short-Term Memory (LSTM). Data preprocessing and feature engineering played a critical role in model development. Feature engineering involved creating cumulative MTD returns, moving averages (5-day and 10-day), and volatility metrics to capture market trends and risk dynamics. These features were derived from daily returns, grouped by year and month, and calculated using rolling operations. Data normalisation was applied to standardise input variables, and missing values resulting from rolling operations were filled to ensure dataset completeness. The dataset was then divided into training and testing datasets using a 1:1 ratio. Model performance was evaluated using mean squared error (MSE) and R-squared (R²) metrics, with cross-validation assuring robustness. Among the models, Gradient Boosting attained the lowest mean squared error (MSE: 2.39 × 10⁻⁶) and the highest R² (0.922), outperforming Random Forest (MSE: 2.72×10⁻⁶, R²: 0.911), XGBoost (MSE: 3.13×10⁻⁶, R²: 0.898), and LSTM (MSE: 3.98×10⁻⁶, R²: 0.870). Feature importance analysis highlighted Market Value Dirty, YTD Return, and Benchmark (BM) Size as the most influential predictors. At the same time, macroeconomic variables such as interest rates and inflation contributed minimally to short-term forecasting. This demonstrates the dominance of portfolio-specific metrics in predicting MTD returns. While LSTM excelled in capturing temporal relationships, its predictive accuracy lagged due to volatility during high-risk periods. The results affirm the effectiveness of machine learning models in enhancing financial decision-making. Gradient Boosting and Random Forest models offer accurate predictions and valuable insights into key portfolio-specific factors, underscoring their utility for risk management and strategic planning. The dissertation recommends further exploration of hybrid models, the inclusion of additional macroeconomic variables, and the integration of real-time data to enhance predictive accuracy and robustness. These improvements establish data-driven approaches as essential instruments for financial firms operating in unpredictable markets.
A Centralised Hadoop-Based Framework for Big Data Analytics in Prime Ministerial Offices: A Namibian Case Study
(Namibia University of Science and Technology, 2025-09-30) Haitamba, Pombili I.P.
Big Data is reshaping the way governments operate, influencing how decisions are made and services delivered. However, most governments, particularly in developing countries, face challenges in managing the growing volume of data generated across public institutions. These challenges can be attributed to a lack of expertise in handling complex and diverse datasets generated in high volume which are becoming difficult for traditional databases to manage. In Namibia, the Office of the Prime Minister (OPM) plays a significant role in coordinating governance and public service administration. The OPM has a mandate to oversee Information and Communication Technology (ICT) in the Public Service and promote e-governance initiatives, which involve the usage and handling of data. However, despite its central role, data from different government ministries remains fragmented, building up silos. This fragmentation limits the government's ability to extract value from the data, hinders coordination across ministries, and ultimately leads to poor service delivery. This study addressed the challenges of data fragmentation within the OPM. It explored the feasibility of designing a centralised database system using Hadoop to integrate and analyse big data across ministries, with the goal of improved administrative efficiency, enhanced public service delivery, and promoting e-governance. Synthetic datasets were generated in Mockaroo, to represent datasets from government ministries. A Hadoop-based setup was undertaken to simulate a centralised database framework integrating all government ministries’ data in Hadoop Distributed File System (HDFS) for storage. A MapReduce job was run in Hadoop, using Java code for analyses across ministries, from Ministry of Home Affairs, Immigration, Safety and Security (MHAISS), Ministry of Labour (MoL), and Ministry of Education, Arts and Culture (MEAC). The job compared poverty indicators based on birth year, educational achievement, and employment status across the three ministries and aggregated the results for regional poverty analysis. The findings of this research show that Hadoop is a cost-effective, open-source framework that has the capabilities to store versatile datasets that currently exists within the Namibian public service into one centralised database that supports big data analysis. The processing layer of Hadoop, MapReduce was able to process a job in minutes that would normally take five to seven working days to complete in the OPMs current administration. Adopting this framework would enable the OPM to make informed decisions backed by evidence, eliminate inefficiencies in public service delivery, and enhance public trust through improved service delivery.
Design of a Model for Augmenting Digital Forensics into Information System Audit in The Financial Sector
(Namibia University of Science and Technology, 2024-09) Iipumbu, Ericky
Audits of information systems demonstrate whether IT controls are effective in protecting company property, ensuring data integrity, and aligning with the organisation's overarching objectives. Information System auditors evaluate all components of financial and business controls, including information technology systems and physical and logical security protocols. As long as cyberattacks continue to occur in the financial sector, it is increasingly important to take robust measures to ensure the integrity and security of information systems. New models that incorporate digital forensics methods into the audit process have emerged as a result of the fact that traditional audit processes frequently fail to meet evolving objectives. Based on the findings of a systematic literature review that demonstrated that the evidence-gathering techniques employed in information system audits are inadequate for assessing the efficacy of internal controls, this study proposes a model that integrates digital forensics into information system auditing that was designed using Design Science Research techniques. The model is envisaged to improve the information system auditing process by augmenting digital forensics processes into information system auditing. The model incorporates key digital forensic components embedded into audit procedures to enhance the accuracy and reliability of evidence collection, ensure the integrity and authenticity of digital evidence, and facilitate a more detailed analysis of audit data. The inclusion of these digital forensic techniques is essential for addressing complex cyber threats and fraud within the financial sector, providing auditors with robust tools to conduct more thorough and defensible investigations, and ensuring compliance with industry standards and regulatory requirements. In summary, the proposed approach offers a structured approach to the integration of digital forensic methodologies into the auditing process and establishes guidelines, in a manner that ensures audit opinions for information systems auditors are unqualified rather than qualified and disclaimer.
Examining The Influence of Facebook and Whatsapp as Sources of News Among Youth in Katima Mulilo, Namibia
(Namibia University of Science and Technology, 2024-07-09) MUSHAUKWA, ARON
The influence and impact of social media as sources of news can no longer be ignored. None should turn a blind eye to how it has captured the imagination of young people, who are the majority users of social media. It is crucial to understand that social media has transformed the landscape of information dissemination. Offering unprecedented access, speed, and interaction on a global scale. It has also democratised access to knowledge, empowering individuals to seek and share information instantly across borders and culture. This study explored the influence of Facebook & WhatsApp as sources of news among youth in Katima Mulilo in the Zambezi Region of Namibia. The study found out that social media evokes the feelings of young people and opens their mind to wider understanding of things. Facebook in particular, allows them to participate in deliberations and discussions, and to share their opinions on various news issues. The data collected and analysed indicate that young people in Katima Mulilo see Facebook and WhatsApp as vital sources of news, connecting, networking and communication, and breaks barriers as it is faster and provides instant news regardless of where one is. The study recommends that media organisations should actively engage with young people on their social media pages, by proactively proving access to news at all times. The study further recommends to local news outlets to open WhatsApp channels in order provide unhindered link to their Facebook pages and websites.
Investigating NBC television coverage of Low-Income Housing: A case study of the Shack Dwellers Federation of Namibia
(Namibia University of Science and Technology, 2021-12) Shuunyuni,Hendrina
The coverage of low-income housing in television news is a critical area influencing public perception and policymaking. This thesis examines the television (TV) coverage of low-income housing, focusing on the Shack Dwellers Federation of Namibia (SDFN). Previous research highlights that media, particularly TV, plays a vital role in shaping social and political discourse. However, there is a noticeable gap in the representation and accurate reportage of low-income housing issues, which remains underexplored and often biased towards wealthier narratives. One of the main challenges identified is the inadequate and sometimes biased media coverage of low-income housing issues, which leads to misinformed public perception and ineffective policymaking. This lack of accurate representation fails to address the SDFN's and its constituents' systemic challenges, perpetuating social and economic inequalities. To address these challenges, the study adopts a qualitative research methodology involving in-depth interviews, focus group discussions, and content analysis. The research involves participants from the SDFN, TV news reporters from Namibia Broadcasting Corporation (NBC), and members of low-income communities. The study uses agenda-setting and framing theories to analyse how television news reports on low-income housing and the impact of these reports on public perception and policy. The findings reveal that TV news coverage significantly impacts the livelihoods of the SDFN members by shaping public opinion and policy decisions. The study uncovers instances of media bias, where the plight of low-income housing is either underreported or misrepresented. The results highlight the need for more balanced and inclusive reporting practices. Furthermore, the research identifies strategies to improve TV news reportage, including training programs for journalists, strategic communications partnerships between government agencies and media, and fostering collaborative efforts between NGOs and media outlets. The academic impact of this research lies in its contribution to media studies and housing policy research, particularly in the context of developing countries like Namibia. Socio-economically, the findings could lead to enhanced public awareness, better-informed policy decisions, and improved living conditions for low-income communities. The research calls for immediate action to address media biases and enhance the representation of low-income housing issues in television news. This will benefit the shack dwellers and contribute to developing more equitable housing policies and practices.
Predicting Lapse Rate in Life Insurance Using Machine Learning Algorithms: A Case Study of Sanlam Namibia
(Namibia University of Science and Technology, 2025-10-07) Kamati, Nelson Mitchell.
The prediction of lapse rate in the life insurance industry holds significant importance for insurers aiming to maintain profitability and retain a strong customer base. Accurate lapse prediction will enable insurers to deploy personalised retention efforts, effective risk management, and financial planning depending on the model’s results. The study explored the use of Machine Learning (ML) techniques to predict lapse rates in life insurance. The dissertation's main contribution was to empirically compare and benchmark 4 machine learning classifier models (Logistic Regression, Gradient Boost, Random Forest, Support Vector Machine). The study utilised secondary data that was obtained from Sanlam Namibia’s database, including policyholder demographics, policy features, premium amounts, payment frequency, lapse status, and other relevant information that could influence the lapse rate. Different types of machine learning algorithms were used on the acquired data to assess their performance in forecasting lapse rate. The study's findings showed that machine learning methods can accurately predict the lapse rates in life insurance. The findings reveal the ensemble model’s high predictive capacity. Gradient Boosting is the best overall classifier, with a 94% and F1-Scor of 94% respectively. Furthermore, some key policy and clients’ characteristics were found to have substantial predictive potential for the lapse rate. However, the limitations of the study must be considered. Finally, the study recommends using ensemble models rather than single model classifiers because they are more effective at predicting life insurance lapses.
Examining the effects of ChatGPT on Television News Production at the Namibian Broadcasting Corporation
(Namibia University of Science and Technology, 2025-01-28) Kadhila, Anna Tashiya
There is limited research on the specific effects and challenges of using Artificial Intelligence (AI) technology, particularly ChatGPT, in newsrooms in Africa, and Namibia in particular. From preliminary research, the Namibian Broadcasting Corporation (NBC) makes use of ChatGPT and other chatbots in the production of television news as an AI tool, but it has been impossible to determine the extent thereto from casual observations. Through qualitative research through conducting face-to-face interviews with fifteen (15) journalists as the population of the study, this thesis aimed to examine the challenges presented by ChatGPT and potentially other chatbots on television news production at NBC, while evaluating their potential to enhance news-gathering processes and their effects from the African context. The study highlights the dual impact of ChatGPT in Namibian television newsrooms, where it enhances efficiency but struggles with cultural and linguistic representation. While AI streamlines news production, its limitations in processing indigenous languages and reflecting local contexts raise concerns about audience alienation and content authenticity. Ethical challenges, including job security and editorial oversight, further emphasize the need for responsible AI integration. Ultimately, the findings call for policies that balance AI’s benefits with Namibia’s socio-cultural realities, ensuring its use strengthens rather than diminishes journalistic integrity.
Assessing The Efficacy of Artificial Intelligence Chatbots for Delivering Public Information: A Case Study of the “Being A Public Servant in Namibia - The Pocket Guide 2.0 Pocket Bot”
(Namibia University of Science and Technology, 2025-01-31) Nassauw, C.C.
The research evaluated the Namibian government's "Being a Public Servant in Namibia – The Pocket Guide 2.0 Pocket Bot" artificial intelligence chatbot which serves as a digital self-service tool for civil servants. The research aimed to determine whether the chatbot fulfils its intended purpose to deliver precise and easily accessible public information to its users. The research employed an exploratory qualitative case study methodology. Twenty civil servants including administrative staff, technical experts, and management personnel participated in semi-structured interviews to gather data. The interview data analysis through thematic methods showed that the users appreciated the chatbot because it provided quick access at all times. The users also highlighted three main concerns with the chatbot system which included its user interface design ,its natural language processing capabilities, and the accuracy of its policy-aligned information. Recommendations were drawn from the research findings for the purpose of improving the chatbot interface, expanding its subject matter coverage, enhancing its response speed, and accuracy to fulfil civil servants’ needs. The recommendations provide guidance to policymakers and government agencies for enhancing chatbot services, enabling them to deliver public information more efficiently while improving user satisfaction.
ASSESSING THE ADAPTIVE CAPACITY OF SELECTED NAMIBIAN PRINT AND BROADCAST MEDIA IN RESPONSE TO CHANGES IN PRACTICE IMPOSED BY COVID-19
(Namibia University of Science and Technology, 2025-01) Lamyaa, Linus
The COVID-19 pandemic significantly disrupted the global media landscape, posing unprecedented challenges for journalists and media organisations. This study examines how media and journalists in Namibia responded to changes in journalism practices imposed by COVID-19. Using an exploratory research design and adopting a qualitative research approach, this study assessed changes to journalism practice as a result of the Covid-19 pandemic, and how journalists and media organisations responded to the changes. Data was collected through in-depth interviews with journalists and media practitioners, from selected print and broadcast media, to understand their experiences and adaptations during the pandemic. Findings reveal that the Namibian media industry underwent substantial transformations, including a shift from traditional newsroom operations to remote and digital reporting. The study also found that misinformation was a major challenge, necessitating rigorous fact-checking and verification processes to maintain credibility and public trust. Financial constraints further exacerbated the difficulties, with many media organisations experiencing significant revenue losses, leading to salary cuts, job redundancies, and shifts in business models. Additionally, the pandemic took a toll on journalists' mental health, with many reporting stress and burnout due to increased workloads, job insecurity, and exposure to distressing news. Despite these challenges, journalists demonstrated resilience by embracing digital transformation, collaborative reporting, and innovative storytelling techniques. Alternative revenue streams, including digital subscriptions and external funding sustained media operations. The study concludes that while COVID-19 accelerated digital transformation and innovation in journalism, it also exposed vulnerabilities in financial sustainability and journalists’ well-being. Strengthening institutional support, investing in digital infrastructure, and implementing mental health initiatives are recommended to enhance media resilience in future crises. These findings contribute to the broader discourse on media adaptation in crisis situations.
AN EXAMINATION OF THE SAFETY AND WELFARE STRATEGIES OF PRINT JOURNALISTS REPORTING ON COVID-19 IN NAMIBIA
(Namibia University of Science and Technology, 2025-01) PAHEJA, SIRIRIKA.
The role of the media during a crisis cannot be over emphasised. During the outbreak of the Covid-19 pandemic, the media assumed a central position in sharing health information during this unprecedented time. In recognition of the significance of journalism, the Namibian government classified journalists as essential workers, allowing them to collaborate freely with healthcare workers during lockdowns. However, being frontline workers raised pertinent issues about the safety and welfare of journalists as they faced challenges, including job losses, mental health problems arising from poor psychosocial care and challenges of working in far-flung regions with meagre resources. Using a qualitative approach and the Protection Motivation Theory, this study examined the safety and welfare strategies employed by selected print journalists in Namibia while reporting on Covid-19. The study assessed selected Namibian print journalists’ perception of their vulnerability to Covid-19, the safety and welfare strategies employed during the COVID-19 pandemic and the effectiveness of the safety and welfare strategies employed when reporting during the COVID-19 pandemic. The Namibian, New Era and Namibian Sun newspapers participated in the study. The study found that journalists felt vulnerable, and experienced emotional and psychological challenges while reporting on Covid-19. The study further observed that various safety and welfare strategies were used by journalists. These included, among others, self-talk and introspection, innovation, escapism, institutional support and adapting and adhering to established health protocols. The study found that while basic support for journalists was provided, mental health support was notably absent despite its crucial role in ensuring that journalists work in a conducive environment. The study concluded that the lack of proper and adequate institutional support for journalists, particularly covering COVID-19 pandemic, compromised both their physical and emotional safety. This undermined their ability to perform their professional duties effectively.
A MACHINE LEARNING-DRIVEN APPROACH FOR ACCIDENT PREDICTION AND TRAFFIC SAFETY ANALYSIS IN NAMIBIA
(Namibia University of Science and Technology, 2024-12) Ngolo, Maria
Globally, road traffic accidents contribute a large portion of injuries, fatalities, and significant economic losses and ongoing research has projected that by 2030, car crashes would be the 5th top reason for loss of life around the world. The key cause of traffic accidents is hard to determine nowadays because of a complex mix of factors, such as road conditions, weather conditions, and the mental condition of the drivers, to list a few. Without a thorough understanding of the characteristics and causes, intelligence-led countermeasures to decrease crashes cannot be created or implemented. Therefore, if traffic accident characteristics can be better understood, it might be easier to take some mitigative action. Nowadays, the utility of machine learning methods in the field of road traffic crashes is gaining traction. The objective of this dissertation is to analyse historical data for a five-year period (2018-2023) and to understand the patterns in accident occurrences by making use of machine learning methods. Machine learning models as such Random Forest, Support Vector Machine, K-Nearest Neighbours, Association Rule Algorithm (AARA), and k-Clustering were employed on the dataset. The Apriori Association Rule algorithm explored the rules with high lift and high support, respectively. The research shows that the Random Forest model is the reliable model in predicting crash severity, reaching an accuracy of approximately 81%. Factors such as junction type, poor road sign conditions, uncontrolled traffic, weather, lighting, road surface, vehicle type, and driver behaviours were identified as the significant variables influencing road accidents. Pedestrian, rollovers, and collision are the leading crash causes of the road accidents, and they are associated with uncontrolled traffic and daylight. Additionally, the research shows notable differences in accident rates by region, month, year, day of the week, and hour of the day, underscoring the impact of geographical features, seasonal trends, and commuting habits on accident rates. The findings indicate that high traffic volumes and urban congestion are the main causes of the greatest accident rates in metropolitan areas, especially in the Khomas Region of Namibia. Further, more accidents are happening more toward the weekend as compared to weekdays and during night hours.
AUTOMATED FRAUD DETECTION IN NAMIBIA’S PUBLIC INSTITUTIONS' FINANCIAL TRANSACTIONS USING MACHINE LEARNING: A DEEP LEARNING APPROACH
(Namibia University of Science and Technology, 2024-12) Johannes, Pandeni Paavo
Financial fraud continues to be a significant concern in public-sector financial operations, undermining the credibility of financial statements and eroding public trust. Traditional methods used by financial experts, such as auditing, are frequently ineffective in addressing the growing complexity of fraudulent activities and effectively mitigating associated risks. This study aimed to tackle this issue by creating an automated fraud detection system based on deep learning designed for Namibia's public sector financial transactions. The Ministry of Finance provided the primary data for the study through the Office of the Auditor-General, which included accounts payable records from public entities with large transaction volumes for the fiscal years 2021/2022 and 2022/2023. The task of fraud detection is framed as a classification problem. The study explored three common deep learning models: Autoencoders, Generative Adversarial Networks (GAN) and Convolutional Neural Networks (CNN). These models' performance was evaluated using historical and simulated financial data, focusing on accuracy, inference time, and resource utilisation. A comparative analysis revealed that the CNN model performed exceptionally well, with the highest accuracy (0.95), F1-score (0.98), and lowest false positive rate (0.038). In contrast, the GAN model excelled in inference time (7.17 ms per transaction) and precision (0.99). This study proposes a scalable, data-driven approach to improving fraud detection in large public-sector financial datasets, thereby increasing accountability in Namibia's public financial systems.
Leveraging Machine Learning for Enhanced Efficiency in Mineral Processing Through Silica Estimation at Rosh Pinah Zinc Mine
(Namibia University of Science and Technology, 2025-04-18) Angula, Tuhafeni Kiiga
In mineral processing, real-time silica estimation in ore samples remains a significant challenge which influences operational inefficiencies, increased costs and recovery rate. Traditional methods depend on periodic laboratory sample tests, which, while accurate, introduce delays that exacerbate issues related to delays on changes of the blending ratio to stabilise the silica entering the plant circuit. These conventional methods provide valuable insights into mineral content patterns but fail to timely deliver the information needed for optimal mineral processing efficiency. This study addresses these challenges by introducing machine learning techniques to estimate silica content in the ore feeds. Four commonly used Machine Learning (ML) models were considered, namely Multiple Linear Regression, Support Vector Regression, Random Forest, and Extreme Gradient Boosting. A dataset consisting of 5967 entries and 10 features collected from mining operations at Rosh Pinah Zinc Mine in Namibia for the past five years (2019-2023) was used to train and test the ML models. The study identifies the key features that significantly influence silica content estimation. Pearson correlation analysis was applied, which shows elements such as Sulphur (-0.54), Zinc (-0.49) and Manganese (-0.49) exhibiting strong negative correlation. Among the four ML models, Random Forest Regression emerged as the highly performing algorithm due to its ability to capture complex and non-linear trends, hence superior performance. The evaluation was obtained using standard machine learning performance metrics, namely Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R²) scores. The Random Forest model demonstrated 85.4% estimation accuracy, with low error rates of MSE at 0.3821 and MAE at 0.2848. This study demonstrates the capability of machine learning to offer a timely data-driven approach to enhance ore blending and improve overall mineral processing efficiency. In addition, this enhances decision-making processes in mineral processing, reducing operational costs and increasing recovery rates.
CO-DESIGNING A CYBERSECURITY PRACTICES FRAMEWORK FOR UNDERSERVED RURAL COMMUNITIES
(Namibia University of Science and Technology, 2024-06-30) Nhinda, Gabriel Tuhafeni
The increasing ubiquity of digital technologies and the Internet has resulted in societies becoming heavily dependent on them for communication, healthcare, education, business operations, and social interaction. While digital transformation promises many benefits, it also presents significant challenges, particularly in ensuring safe and secure online participation. Cybersecurity, as a critical enabler of digital trust, is essential for inclusive and resilient digital societies. However, in underserved rural communities, especially in the Global South, cybersecurity practices remain largely unfamiliar due to limited telecommunications infrastructure, socio-economic inequality, and low digital literacy. This study addresses this gap by co-designing a cybersecurity practices framework with and for underserved rural communities. Rooted in a multidisciplinary approach, the research integrates elements of human-computer interaction (HCI), community engagement, and behavioural insights. A qualitative methodology was employed, using co-design sessions, focus group discussions, and interviews, to understand community-specific perceptions and practices of cybersecurity. The study draws on indigenous philosophies of Ubuntu and Uushiindaism to contextualise security as a communal, ethical, and relational construct. The framework was evaluated through a two-pronged process: expert focus groups (ex-post) and community-based co-design sessions (exante and ex-post). Evaluation centred on ecological utility—ensuring contextual relevance, cultural fidelity, sustainability, comprehensibility, and local ownership. The study further maps key actors and relationships influencing cybersecurity practices in rural contexts, including interactions with national agencies, community policing units, local radio, and ministries responsible for awareness and legislation. This research contributes to a more inclusive understanding of cybersecurity by recognising the socio-cultural dynamics that shape digital safety in underserved communities. It bridges the gap between globally dominant cybersecurity paradigms and the lived realities of rural African users. The framework empowers communities to engage confidently with technology, facilitates digital access to essential services, promotes skills transfer through seasonal resident engagement, and supports the preservation of indigenous languages. For policymakers, developers, and educators, it offers actionable insights into designing secure, contextually appropriate technologies for the margins while ensuring broader applicability across user groups.
Predicting International Tourist Arrivals in Namibia Using Machine Learning
(Namibia University of Science and Technology, 2024-08) Shivute, Selma.
The lack of accurate and timely predictions for international tourist arrivals in Namibia remains an open problem, leading to inefficiencies in tourism planning and resource allocation. Traditional methods, primarily based on seasonal trends and historical data, dominate the forecasting landscape. Although traditional approaches are good at capturing seasonal patterns, there is always a lack of accounting for more dynamic and non-linear interactions between the predictive variables and more reliable results. In recent years, machine learning models like Seasonal Autoregressive Integrated Moving Average (SARIMA), Random Forest, and Prophet have gained increasing support for their ability to handle complex, non-linear data and provide more accurate tourist arrival forecasts. SARIMA is particularly good at modelling seasonal time series data, Random Forest excels in capturing non linear relationships, and Prophet is designed to handle time series data as well as irregular and missing data. However, attempts to implement these three models in predicting international tourist arrivals in the Namibian context have exposed limitations such as a constant prediction with Random Forest and the need for extensive tuning in SARIMA and Prophet, which may result in their prediction accuracy showing little improvements in Namibia’s tourism sector. This study aimed to develop and test the three models: SARIMA, Random Forest, and Prophet, to predict international tourist arrivals in Namibia more accurately. Accurate forecasts can improve decision-making within the tourism sector, infrastructure planning and resource allocation. The methodology used involved data preparation and data exploratory strategies to determine the relationship between exploratory variables and dependent variables. It also included training and validating these models on historical data obtained from the Ministry of Home Affairs, Immigration, Safety and Security. The models were hyper-tuned to overcome the limitations of accuracy by improving accurate predictions. It is expected that these models will overcome the limitations of non-accuracy predictions of tourist arrivals. The effectiveness of these models' accuracy was evaluated using Root Mean Square Error and Mean Absolute Error and comparing their performance against each other to determine the preferred model for Namibia. The results indicated that SARIMA achieved the most accurate prediction, followed by Random Forest, and Prophet performed poorly in predicting international visitor arrivals in Namibia. The two models are anticipated to contribute significantly to more efficient tourism management and planning in Namibia.
Exploring Machine Learning on Geochemistry Data For Efficient Prediction of Metal Concentrations in Copper Deposits
(Namibia University of Science and Technology, 2024-01-25) Joel, Lydia.
Naturally occurring ore bodies like Copper often occur in compound form with other useful metals such as Silver, Lead and Zinc. Due to the cost, mining companies find it difficult to pay for analysis of various metals in their samples and end up focusing on analysing one metal or a few, leaving out a bunch of other associated metal concentrations in the deposit. Additionally, analysing different metals in samples can take time, and this increased turnaround time of receiving results from the laboratory can negatively affect production. The research used a geochemistry dataset comprising of 3,282 samples from the Kombat Copper deposit area in Namibia to predict copper (Cu) concentrations from zinc (Zn) and lead (Pb) concentrations. In addition to the metal concentrations, the dataset had sample coordinates and grid names features. The four machine learning algorithms used were Random Forest (RF), K-Nearest Neighbour (KNN), Decision Tree (DT), and Support Vector Machine (SVM). These models were used because they were the commonly employed models for similar purposes, in the literature reviewed. The learning task was a regression problem, therefore, the primary metric utilised to assess the machine learning model and draw performance conclusions was the regression score (R-squared), which quantifies how well the model explains the variance in the data. The R squared score represents the percentage of variance in the dependent variable (target) that can be predicted from the independent variables (features). It ranges on a scale of 0 to 1, where 1 indicates a perfect fit. In addition Mean Squared Error (MSE), Root means squared error (RMSE), mean absolute error (MAE), Adjusted R-squared, and explained variance metrices were also looked at. Based on the R-squared metric, the KNN model outperformed the other three models, predicting 57% of the relationship between the dependent and independent variables. K-NN was followed by RF with 0.55 score, DT with a 0.49 score and the SVM with a 0.44 score. KNN model appeared to be the best choice among the four models for making predictions for the dataset. Further optimisation of the models improved their prediction accuracy, with the KNN model still with a superior performance of R-squared at 70% (0.70) with n-estimators set at 4 and the test size set to 10%. Predicting metal contents from geochemistry data with machine learning can iv help mining companies reduce costs by supplementing lab-based analyses with model-based predictions in determining grades.
Recommending a Machine Learning Model to Detect the Fatigue State for Employees at Namdeb
(Namibia University of Science and Technology, 2024-04) Nakale, S.N.
Workplace fatigue is one of the major risk factors across different industries and it negatively impacts productivity and workplace safety. Thus, fatigue detection and monitoring is essential to promote occupational health and safety. The advancements in data collection technologies have made it possible for industries to develop data driven solutions by developing Artificial Intelligence (AI) and Machine Learning (ML) based fatigue monitoring and detection systems. The Advanced Driver Assistance Systems (ADAS) is one of the technologies that has been adopted in various industries to improve safety for drivers. Namdeb implemented the ADAS in recent years and they have identified the need for a system to detect and classify employees’ fatigue state using data from the ADAS. In this study, an ML based fatigue detection system was proposed. Facial behavioural fatigue features were used to detect fatigue. The proposed system deployed some of the commonly used ML classification algorithms and it was evaluated on a simulated dataset, the Yawning Detection Dataset (YawDD), and a real-world dataset, data from the Namdeb ADAS. The results showed that most of the supervised ML classifiers achieved a fatigue prediction accuracy above 90% for both datasets. The Random Forest (RF) based fatigue detection-based model was found to be the best model. The k-Means which is an unsupervised ML classifier exhibited the worst performance. However, the reliability and generalisability of the results based on the real-world dataset can be improved by using a larger dataset. The major challenge to developing behavioural based fatigue detection systems for real world setting like the mining environment is face detection accuracy which is affected by factors such as low image resolution due to poor and variable lighting conditions, face orientation to camera and proximity of face to the camera. The significant contribution of this study is the use of real-world dataset to test the proposed fatigue detection system. Overall, the study contributes to the promotion of the eighth Sustainable Development Goal (SDG) of promoting safe working environments.
ASSESSING THE STRATEGIC ROLE OF INTERNAL COMMUNICATION AT THE MINISTRY OF HOME AFFAIRS, IMMIGRATION, SAFETY AND SECURITY IN NAMIBIA
(Namibia University of Science and Technology, 2023-05-04) KADHIKWA, SAKEUS IITA.
Internal communication has long been the lifeblood of successfully and effectively managed organizations, especially when such communication is strategically managed. Though literature has established the benefits of internal communication, especially when it comes to employee engagement, commitment, and improved corporate reputation, research assessing the strategic role of internal communication in government departments is limited. The aim of the study was to assess the strategic role of internal communication at the Ministry of Home Affairs, Immigration, Safety and Security in Namibia. To achieve this objective, four secondary objectives were pursued. The first secondary objective focused on examining the challenges faced by the Department of Home Affairs and Immigration in implementing the internal communication strategy. The second objective focused on analyzing the impact of internal communication on service delivery, while the third focused on internal communication and employee morale. The last objective examined the alignment between internal communication and the department’s strategy. Data for the study were collected using a structured interview guide. Face-to-face online were conducted with a conveniently selected sample of 25 participants. The methodology that was employed was a qualitative research approach that followed a case study research design. Several challenges affecting the effective implementation of the internal communication strategy were identified. Furthermore, it was noted that participants had mixed views about the role of internal communication and service delivery, as well as its strategic role in the organization. Based on these findings, the study makes recommendations and suggestions for future research.
DESIGNING A BRING YOUR OWN DEVICE SECURITY AWARENESS MODEL FOR MOBILE DEVICE USERS IN NAMIBIAN ENTERPRISES
(NAMIBIA UNIVERSITY OF SCIENCE AND TECHNOLOGY, 2023-05-30) Shihepo, Ester
The phrase Bring Your Own Device (BYOD) also known as Dual-Use Devices is a mutual practice which has increased employees’ access to new mobile technologies and a rising trend within many organisations. The concept refers to employers allowing their employees to bring their personal mobile devices to workplaces and use them as their workstations. Enterprises are enjoying the benefits of BYOD, which allows them to cut operational costs as they do not need to purchase computers for their employees. Employees are enjoying the comfort and convenience offered by BYOD; however, this exposes organisations to security breaches. There is currently a lack of security awareness among mobile device users within enterprises against BYOD cyber threats. The situation has made it difficult for organisations to monitor the usage of resources among the mobile users towards protecting the confidentiality, integrity, and availability of corporate data. Moreover, cyber attackers see more potential with mobile devices as company and personal data get mixed up on such devices. Although the BYOD concept has not been formally implemented within the enterprise, it does not mean that the enterprise data is not prone to attackers. This study presents a BYOD Security Awareness Model designed following Design Science Research methods based on findings of a single case study conducted in one of the enterprises in Windhoek, Namibia. Qualitative research following the interpretivism philosophy was used. To select participants, purposeful random sampling method was used for this study. Data was collected using interviews, a questionnaire and through literature review. Furthermore, the study implemented the qualitative content analysis as the data analysis technique. The study identified malware and network spoofing as some of the BYOD related threats affecting the case enterprise. The researcher observed lack of end user awareness on BYOD security as a cause of BYOD related security threats. The study identified four main components of the model namely: BYOD threats, security awareness, policy and access control. The BYOD security awareness model will be a guideline to Namibian enterprises in creating BYOD security awareness among their mobile devices with the aim to safeguard the organisational data. Furthermore, the findings will also contribute to the new technology horizon of Namibia’s future BYOD security awareness by motivating enterprises to implement mechanisms that will protect the enterprise confidential information. Since Namibia is reported as one of the least ranked countries in Africa in terms of cyber security, the model is a guideline on how enterprises can create BYOD security awareness among users within their enterprises and improve their security posture as well as that of the nation. Additionally, the model will also contribute to the BYOD security awareness knowledge to researchers and practitioners through conference papers and thesis publication.

Browse

Recent Submissions