School of Computing and Informatics Technology (CIT) Collection

Permanent URI for this collection

http://10.29.1.53:4000/handle/10570/88

Browse

Recent Submissions

Now showing 1 - 5 of 631
  • Item
    Interpretable multi-strategy learning for predictive therapeutic adherence to osteoporosis treatment among chronic patients
    (Makerere University, 2025) Busingye, Caroline
    This study presents an interpretable multi-strategy learning framework to predict thera- peutic adherence to osteoporosis treatment among chronic patients, advancing the use of Artificial Intelligence (AI) in adherence prediction. Traditional, ensemble, deep learning, and hybrid models were developed and evaluated using Explainable AI (XAI) techniques LIME, SHAP, and Permutation Feature Importance to identify key adherence factors. The Extreme Gradient Boosting (XGBoost) model achieved the best overall valida- tion performance with 68.3% accuracy, 66.3% F1 score, and 73.4% AUC-ROC and was deployed as a web-based prediction tool on Render for real-time adherence prediction. XAI results showed that lifestyle factors like smoking, low physical activity, and inad- equate calcium intake negatively affected adherence, while regular exercise and sufficient vitamin D intake improved it. These findings highlight an interpretable, data-driven pathway for clinicians and researchers to support personalized interventions and enhance adherence outcomes in osteoporosis management. Deep Learning, Explainable Artifi- cial Intelligence (XAI), Feature Importance, LIME, Machine Learning Interpretability, Osteoporosis, Postmenopausal, SHAP, Therapeutic Adherence, XGBoost
  • Item
    A framework to improve electronic viral load result distribution to lower facilities : a case study of Kayunga Hospital, Uganda
    (Makerere University, 2026) Nakigozi, Vivian
    This study was conducted to investigate the challenges that lead to long viral load results turnaround time at the lower facilities in Kayunga District. The study objectives were: to determine the requirements for a framework, structured system designed to improve the distribution of electronic viral load results to low level health facilities, and to design that framework to support improved data flow and accessibility, and to evaluate how well the framework, system developed to improve the efficiency and reliability of electronic viral load results distribution, enhance the delivery of these results to the intended facilities. The study used a mixed–methods approach, which combines qualitative and quantitative methods to gain a comprehensive understanding of the issues. Qualitative research helped to explore the experiences and perceptions of healthcare providers and patients. It used methods like interviews to gather in-depth insights into the challenges in electronic viral load result distribution, while Quantitative research provided statistical analysis and trends. It employed surveys and statistical analysis to identify patterns and trends in viral load data. The study used a case study and survey research as a research strategy to conduct an in-depth analysis of a specific healthcare facility to understand the challenges and context of electronic viral load results distribution, questionnaires to collect data from a large sample of healthcare providers and patients, providing a broader understanding of the issue. A sample size of 279 respondents was used from a population of 882 respondents. Data was collected using the questionnaire and interview guide. The response rate of 100% was obtained from which key findings suggested a positive response and provided valuable insights for health facilities to identify critical areas for improvement in the testing processes. From the study, it was learned that to determine the requirements for the framework, we had to first address the identified challenges that lead to long TAT to identify the requirements. It was also learnt that to design the framework, we had to first identify the design choices to fulfil the identified requirements for framework design. Therefore, it was concluded that to overcome the challenges, there is a need to enhance standard operating procedures governing sample handling and processing to address the issue of the longtime taken to process samples and reduce delays in testing processes. This requirement can be achieved if the Ministry of Health (MOH) strengthens primary healthcare delivery systems by introducing data quality policies and practices for the institution, and also ensures that standard operating procedures are integrated into automatic processes at all health facility levels. The study recommendations include the need for continuous improvement activities such as continuing medical education, quality improvement projects, root cause analysis, and monthly staff rotations. For results that are misplaced, developing a root cause analysis, phone call follow-ups, and increasing the frequency of reviewing hub riders from monthly to twice a month.
  • Item
    Public participatory GIS tool for supporting citizen involvement in strategic planning of local government service delivery: a case of Kira Municipality in Uganda
    (Makerere University, 2025) Nabwami, Doreen
    Citizen participation is a cornerstone of effective and democratic local governance. In Uganda, the decentralization framework provides for citizen involvement in planning and decision-making; however, in practice, participation in the strategic planning of Local Government (LG) service delivery remains limited. This study aimed to develop a Public Participatory Geographical Information System (PPGIS) tool for supporting citizen involvement in the strategic planning of LG service delivery. Specifically, the research sought to: (1) examine existing challenges and requirements for citizen participation; (2) design a PPGIS tool to support citizen involvement; (3) implement the tool; and (4) test and validate its effectiveness. An exploratory survey and a case study design were employed, focusing on Kira Municipality. Data was collected through questionnaires, interviews, and document reviews from both citizens and LG officials. Quantitative data was analyzed using descriptive statistics, while qualitative data was thematically analyzed. The findings revealed persistent barriers to citizen participation, including limited GIS capacity, inadequate institutional support, and insufficient access to information. In response, a PPGIS tool (CISP) was designed, implemented, and validated to facilitate interactive and spatially enabled participation in strategic planning. CISP demonstrated the potential to improve transparency, collaboration, and real-time feedback between citizens and LG officials. The study concludes that integrating PPGIS into local governance processes can strengthen participatory decision-making and accountability. It recommends further adoption of such tools across Ugandan LGs to promote inclusive, data-driven, and citizen-centered service delivery.
  • Item
    Machine learning models for short-term rainfall prediction using Uganda’s Lake Victoria Basin Weather Dataset
    (Makerere University, 2025) Tumusiime, Andrew Gahwera
    As climate change intensifies, accurate short-term rainfall forecasting has become an urgent research priority. Numerical Weather Prediction models often struggle with precipitation due to high computational requirements and large error margins. This dissertation addresses these challenges by introducing a curated multi-station dataset for the Lake Victoria Basin (LVB) and systematically evaluating Machine Learning and Deep Learning approaches for rainfall forecasting in a data-scarce African environment. Regression experiments benchmarked Random Forest, Support Vector Regression, Neural Network Regression, Least Absolute Shrinkage and Selection Operator, Gradient Boosting, and Extreme Gradient Boosting (XGBoost) Regression. Among these, XGBoost consistently achieved the lowest error, with Mean Absolute Error values as low as 0.006, 0.018, and 0.005 mm h−1 for Uganda, Kenya, and Tanzania, respectively. To complement these continuous forecasts, rainfall classification was implemented using Multi-Layer Perceptrons, Convolutional Neural Networks, and Long Short-Term Memory networks (LSTM). LSTM outperformed alternative architectures, achieving weighted F1-scores above 90% at multiple stations. To overcome data scarcity, a transfer-learning strategy was developed by fine-tuning pre-trained LSTM models from data-rich stations and applying them to the data-limited station of Kisumu, yielding performance improvements of up to 3%. An ensemble of these transfer-learned models using an Exponential Weighting Algorithm further enhanced robustness, delivering gains of up to 5% in F1-score. Overall, the dissertation demonstrates that an ensemble-based transfer-learning framework, grounded in a regionally curated dataset, can substantially improve rainfall forecasting in East Africa. The integration of regression, DL classification, and transfer learning provides methodological advancement and operational potential, contributing to more reliable weather services in data-scarce regions such as the LVB.
  • Item
    Fraud detection in mobile money transactions using differentially private machine learning techniques
    (Makerere University, 2025) Kagaba, Dennis
    Mobile money has become a vital financial tool in regions with limited banking infrastructure, fostering financial inclusion for underserved populations. However, its rapid growth has led to rising fraud threats such as refund scams, account takeovers, and identity misuse. Machine learning (ML) models show promise in detecting fraud through transaction patterns, but they rely heavily on sensitive personal data, raising concerns about privacy, security, and regulatory compliance. These risks may reduce user trust and hinder adoption. This thesis explores the use of privacy-preserving ML, specifically differential privacy to enable effective fraud detection while safeguarding user data. This study adapts privacy preserving ML models by integrating differential privacy into supervised algorithms, including Gaussian Naive Bayes, Logistic Regression, Decision Trees, and Random Forest. A labeled synthetic Mobile money transaction dataset is used to train these models, leveraging the IBM Diffpriv library to enforce differential privacy constraints. Model performance was further assessed across varying privacy levels using the differential privacy parameter e to evaluate the privacy utility trade off. The results showed that Logistic Regression experienced substantial degradation under strict privacy € = 1, with accuracy dropping to 0.496 and recall to 0.047, indicating high sensitivity to privacy noise. Meanwhile, Gaussian Naive Bayes, Decision Trees, and Random Forest remained robust, maintaining high accuracy and recall even at lower epsilon values. Across the full range of privacy budgets, Random Forest and Gaussian Naive Bayes demonstrated the most stable performance, highlighting their suitability for privacy preserving fraud detection. These results indicate that differentially private machine learning models can effectively detect frandulent transactions while preserving user privacy. The findings contribute to the development of secure, ethical and privacy sensitive fraud detection systems for mobile financial services.