AI +ML

Developing e-Health systems to improve healthcare service delivery in South Africa

Patients’ medical records are critical to services, it helps to prevent errors in prescriptions and medications. Despite the understanding of these consequential risks to patients, medical errors remain prevalence in South Africa and many parts of the world. Instances of incorrect medications, prescription misrepresentations, and mismanagement within healthcare facilities have been observed in South Africa. Additionally, patients' limited access to their medical records and the utilization of incomplete medical records contribute to these challenges. A significant concern arises when patients seek treatment in different provinces, as the lack of shared medical records impedes seamless continuity of care. In critical situations where patients are unable to respond, doctors are forced to make crucial decisions without access to sufficient medical history information. However, by implementing a unique identifier system for patients, linked with biometrics and AI technologies, and integrating real-time medical information into a centralized system, this issue could be effectively addressed. Such a solution would enable healthcare providers to access comprehensive patient records promptly, facilitating informed decision-making and improving patient safety and care outcomes.

Research output

Phase 1 :Design of the application (completed)

Presented a conference paper titled “Towards a design of E-Health systems to improving healthcare service delivery” in the International Conference on Information Resources Management held in Ontario on October 18-20, 2022.

Conference proceedings

Tom, S. L., Iyamu, T. (2022). Towards a design of E-Health systems to improving healthcare service delivery. International Conference on Information Resources Management. CONF-IRM 2022 Proceedings. 25. ISBN:78-0-473-65864-9. https://aisel.aisnet.org/confirm2022/25

Predicting NBA outcomes using machine learning: Can emerging leagues benefit? 

The National Basketball Association provides strong potential for predictive analytics due to its global fan base, structured season and rich historical data. This research evaluates the ability of machine learning models such as logistic regression, support vector machines and random forests to predict the 2023-2024 NBA Championship winner. The research used 20 years of team performance statistics to select predictive features including field goals percentage, two-point percentage, three-point percentage, assists, rebounds and blocks through correlation-based feature selection. The models achieved evaluation through accuracy, precision, recall and F1 score measurements. Logistic regression produced the most balanced and interpretable results despite the class imbalance in the dataset because it correctly predicted Boston Celtics as the championship winner at 94.99% probability. Random forest produced the highest accuracy rate of 93%, yet it did not correctly predict the winner because it performed poorly on minority class detection. Support vector machine yielded intermediate performance. The success of logistic regression demonstrates its effectiveness for sports datasets with imbalanced outcomes and proves that conventional team statistics remain valuable for predictions. The research findings expand sports analytics knowledge while showing how machine learning systems can assist professional basketball teams with strategic planning. The developed framework can be adapted to support performance management and talent development in emerging basketball leagues. Future research could use advanced metrics together with neural networks and other alternate modelling techniques to boost predictive power and solve binary classification challenges.

Research outputs

Tom, S. & Chen, Y.-Y. D.(2025). Predicting NBA outcomes using machine learning: Can emerging leagues benefit? Proceedings of the IEEE International Conference on Emerging Trends in Networks and Computer Communications (ETNCC 2025). (In press).

 Stereoscopic 3D Video Generation

The summer of 2010 witnessed the introduction of 3DTVs by several prominent TV manufacturers, employing the use of shutter-glasses technology. The advent of 3D video applications has significantly impacted our daily lives, particularly in the realm of home entertainment. While the production of 3D movies continues to rise, the available 3D video content is still not extensive enough to meet the growing demands of the future 3D video market. As a result, there is a pressing need for innovative techniques that can automatically convert 2D video content into captivating stereoscopic 3D video displays. In this research paper, we propose an automatic monoscopic video to stereoscopic 3D video conversion scheme that utilizes block-based depth from motion estimation and color segmentation for enhancing the depth map. The incorporation of color-based region segmentation ensures accurate region boundary information, which is then fused with the block-based depth map to eliminate visual artifacts and assign appropriate depth values within each segmented region. The experimental results in this study demonstrate that this scheme yields high-quality output, successfully delivering immersive and engaging stereoscopic 3D video experiences.

Research outputs

Presented poster on “3D SCENE MODELING” in International Women’s conference conducted by ACM on Sep 16-17, 2010.

Robo-Cop

During my pursuit of a BTech degree in Computer Science and Engineering (2009), I undertook a remarkable project that involved developing a  robot, designed to operate in sensitive and hazardous environments including military and spy services . Leveraging my knowledge in programming and engineering, I designed and implemented a comprehensive program that enabled the robot to carry out covert operations and surveillance activities. This involved integrating advanced algorithms, sensor technologies, and secure communication systems into the robot's architecture. Throughout the project, I utilized my expertise in computer science to ensure optimal performance, reliability, and adaptability of the program in various challenging environments. This endeavor not only deepened my understanding of robotics and software development but also emphasized the significant role that technology plays in enhancing security and intelligence operations.

Technology Used: Embedded C, Java