Magister Scientiae - MSc (Computer Science)

Permanent URI for this collection

https://hdl.handle.net/10566/14793

Browse

Now showing 1 - 7 of 7

Data science for health-care: Patient condition recognition
(University of the Western Cape, 2019) Mandava, Munyaradzi; Bagula, Antoine
The emergence of the Internet of Things (IoT) and Artificial Intelligence (AI) have elicited increased interest in many areas of our daily lives. These include health, agriculture, aviation, manufacturing, cities management and many others. In the health sector, portable vital sign monitoring devices are being developed using the IoT technology to collect patients’ vital signs in real-time. The vital sign data acquired by wearable devices is quantitative and machine learning techniques can be applied to find hidden patterns in the dataset and help the medical practitioner with decision making. There are about 30000 diseases known to man and no human being can possibly remember all of them, their relations to other diseases, their symptoms and whether the symptoms exhibited by the patients are early warnings of a fatal disease. In light of this, Medical Decision Support Systems (MDSS) can provide assistance in making these crucial assessments. In most decision support systems factors a ect each other; they can be contradictory, competitive, and complementary. All these factors contribute to the overall decision and have di erent degrees of influence [85]. However, while there is more need for automated processes to improve the health-care sector, most of MDSS and the associated devices are still under clinical trials. This thesis revisits cyber physical health systems (CPHS) with the objective of designing and implementing a data analytics platform that provides patient condition monitoring services in terms of patient prioritisation and disease identification [1]. Di erent machine learning algorithms are investigated by the platform as potential candidate for achieving patient prioritisation. These include multiple linear regression, multiple logistic regression, classification and regression decision trees, single hidden layer neural networks and deep neural networks. Graph theory concepts are used to design and implement disease identification. The data analytics platform analyses data from biomedical sensors and other descriptive data provided by the patients (this can be recent data or historical data) stored in a cloud which can be private local health Information organisation (LHIO) or belonging to a regional health information organisation (RHIO). Users of the data analytics platform consisting of medical practitioners and patients are assumed to interact with the platform through cities’ pharmacies , rural E-Health kiosks end user applications.
Design and Implementation of a Credible Blockchain-based E-health Records Platform
(University of the Western Cape, 2020) Xu, Lingyu; Bagula, Antoine
With the development of information and network technologies, Electronic Health Records (EHRs) management system has gained wide spread application in managing medical records. One of the major challenges of EHRs is the independent nature of medical institutions. This non-collaborative nature puts a significant barrier between patients, doctors, medical researchers and medical data. Moreover, unlike the unique and strong anti-tampering nature of traditional paper-based records, electronic health records stored in centralization database are vulnerable to risks from network attacks, forgery and tampering. In view of the data sharing difficulties and information security problems commonly found in existing EHRs, this dissertation designs and develops a credible Blockchain-based electronic health records (CB-EHRs) management system. To improve security, the proposed system combines digital signature (using MD5 and RSA) with Role-Based Access Control (RBAC). The advantages of these are strong anti-tampering, high stability, high security, low cost, and easy implementation. To test the efficacy of the system, implementation was done using Java web programming technology. Tests were carried out to determine the efficiency of the Delegated Byzantine Fault Tolerance (dBFT) consensus algorithm, functionality of the RBAC mechanism and the various system modules. Results obtained show that the system can manage and share EHRs safely and effectively. The expectation of the author is that the output of this research would foster the development and adaptation of EHRs management system.
Design and implementation of a credible blockchain-based e-health records platform
(University of Western Cape, 2020) Xu, Lingyu; Bagula, Antoine; Isafiade, Omowunmi
With the development of information and network technologies, Electronic Health Records (EHRs) management system has gained wide spread application in managing medical records. One of the major challenges of EHRs is the independent nature of medical institutions. This non-collaborative nature puts a significant barrier between patients, doctors, medical researchers and medical data. Moreover, unlike the unique and strong anti-tampering nature of traditional paper-based records, electronic health records stored in centralization database are vulnerable to risks from network attacks, forgery and tampering. In view of the data sharing difficulties and information security problems commonly found in existing EHRs, this dissertation designs and develops a credible Blockchain-based electronic health records (CB-EHRs) management system.
Internet-Of-Things for Cyber Healthcare (L0t4c): Information Dissemination, systems' Interoperability and security
(University of the Western Cape, 2017) Lubamba, Claude kakoko; Bagula, Antoine
Cyber Healthcare is becoming one of the fastest growing industries in the world due to an increasing elderly population and a more health conscious word population. On the other hand, IoT devices are emerging from niche areas to provide new services that we could not fathom without the technological advances made in IoT and healthcare elds [1]. Wireless Sensor Networking (WSN) is a promising approach to cyber healthcare as it can enable real-time monitoring of patients and early detection of emergency conditions and diseases [2, 3]. However, there are a number of issues that need to be addressed in order to bene t from the cyber healthcare promises.
Scalable Wireless Mesh Networks
(University of the Western Cape, 2016) Abdalla, Taha; Bagula, Antoine
Wireless Mesh Networks (WMNs) are wireless multi-hop networks built on wireless nodes that operate in an Independent Basic Set Identifier (IBSS) mode of the IEEE 208.11 wireless standard. IBSS is well known as an ad hoc mode which is found to build ad hoc wireless networks with the aid of routing protocols crafted to work in this mode. Ad hoc wireless mesh networks are always described as self-healing, self-configuring, easy to build, etc. However, these features do come at a cost because a WMN suffers performance degradation and scalability issues, which mainly come from the underlying IBSS mode that is used to form the physical network. Furthermore this is exacerbated by routing protocols in the upper layers which are intended to form a flat network architecture. Partitioning or clustering the flat network into smaller units has been proven to be a viable mechanism to counter the scalability problem in the communication network. The wired network for instance, presents a segmented, hierarchical architecture, where end user devices are organized in virtual local area networks (VLANs) using Ethernet switches and then Routers aggregate multiple VLANs. This thesis develops and evaluates a heterogeneous, clustering architecture to enhance WMN scalability and management. In the proposed architecture, the clustering is separated from the routing, where the clustering is done at the physical layer. At the routing level, each cluster is configured as a WMN using layer 2 routing for intra-cluster routing, and layer 3 routing for inter-domain routing between clusters. Prototypes for the proposed architecture have been built in a laboratory testbed. The proposed architecture reported better scalability and performance results compared to the traditional flat architecture.
Smart cities air pollution monitoring system - Developing a potential data collecting platform based on Raspberry Pi
(University of the Western Cape, 2019) Chen, Shu; Bagula, Antoine
Air pollution is becoming a challenging issue in our daily lives due to advanced industrialization. This thesis presents a solution to collection and dissemination of pollution data. Most of the devices that monitor air quality are costly and have limited features. The aim of this study is to revisit the issue of pollution in cities with the aim of providing a cheaper and scalable solution to the challenge of pollution data collection and dissemination. The solution proposed in this paper uses Raspberry Pi and Arduino micro-controller boards as the foundation, combined with specific sensors to facilitate the collection and transfer of pollution data reliably and effectively. While most traditional air pollution monitoring equipment and similar projects use memory cards as a medium for data storage, the system proposed in this research is built around a new network selection model that transfers data to the server by using either Bluetooth, Wi-Fi, GSM, or the LoRa protocol. The connectivity protocol is selected automatically and opportunistically by the network selection algorithm defined in the micro-controller board. The final data will be presented to the user through a mobile application and website interface effectively and intuitively after being processed in the server. This data transfer system can effectively reduce the cost and input of human resources. It is a viable solution. For other environmental research, this system can provide an air quality data support for analysis and reference. Modularity and cost-effectiveness are fully considered when designing the system. It is a viable solution. We can generalize the system by slightly changing the data transmission modules. In other case, it can be used as a platform for similar data transmission and offer help for other research directions.
Smart renewable energy : architectures, dimensioning and monitoring
(University of the Western Cape, 2017) Erasmus, Zenville; Bagula, Antoine
The Smart Renewable Energy project at the University of The Western Cape, under the guidance of the Intelligent Systems and Advanced Telecommunication (ISAT) group, aims at developing a dynamic system that enables users to (1) design smart architectures for next generation wind and solar systems to meet African power challenges (2) use these architectures to dimension the underlying solar and wind power systems and (3) simulate, implement and evaluate the performance of such power systems. The project's existing web and mobile monitoring system will undergo a much needed upgrade to cater for monitoring of the existing system's environmental and battery bank parameters. This will be implemented by allowing users to monitor input, storage and output trends over various time frames. These time frames would include hourly, daily, weekly and monthly readings. The visual evaluation of the system will be generated by mathematical, statistical and machine learning techniques. Trends will be discovered that will allow users to optimize the system's efficiency and their usage patterns. The accompanied dimensioning system will allow users to cater for their needs in a two way fashion. Users will be able to specify the number of devices that they want to run from a solar or wind based system and their power needs will be generated. They will also be able to determine what a given system is capable of producing and the number of devices that can be used simultaneously, as a result.

Browse

Browsing by Author "Bagula, Antoine"

Results Per Page

Sort Options