Theses and Dissertations

Permanent URI for this collectionhttps://dspace.nm-aist.ac.tz/handle/123456789/3186

Browse

Now showing 1 - 20 of 57

Mechanism for detection and mitigation of address resolution protocol spoofing attacks in large-scale software-defined networks
(NM-AIST, 2025-03) Patrice,Laurent
Address Resolution Protocol (ARP) spoofing has been a long-standing problem, with no clear remedy until now. The attacks can be launched easily, utilizing an enormous number of publicly available tools on the web. However, they are extremely tough to counterattack due to ARP's stateless nature for not authenticating ARP replies for subsequent requests. Previous studies have demonstrated significant efforts to counterattack these assaults in Software Defined Networks (SDN). However, they mainly focused on detecting the assaults, with little effort being made to address performance bottlenecks, scalability, and Single Point of Failure (SPOF) issues in large-scale LANs. This study focuses on developing ARP spoofing attacks detection and mitigation mechanism in large-scale SDN that is resistant to SPOF, performance bottlenecks, and scalability constraints. It enables controllers to intercept and analyze incoming ARP packets, learn address mappings, and store them in the application’s memory for ongoing ARP cache comparisons while maintaining a global ARP cache in the controller. Simulation experiments were carried out in a closed network environment to precisely monitor network traffic and result patterns. Mininet and Open Network Operating System were used to implement the data plane and OpenFlow-based control plane, respectively. The results show that the proposed solution is resistant to ARP spoofing attacks, with an average detection and mitigation time of 4.3 and 26.19 milliseconds, respectively. Further significant improvement has been observed in alleviating SPOF, performance bottlenecks, and scalability constraints. Further improvement can be done to enhance the proposed solution to counterattack multiple types of assaults through machine learning models.
Machine learning model for early detection of sexually transmitted infections
(NM-AIST, 2025-07) Shija,Juma
Sexually Transmitted Infections are diseases transmitted mostly through unprotected sex with an infected partner. Every day, about one million people throughout the world get sexually transmitted infections. The most vulnerable groups in Tanzania are commercial sex workers, truck drivers who travel long distances and grocery and hotel workers. Common sexually transmitted infections in Tanzania are Gonorrhoea, Syphilis, Chlamydia and Trichomoniasis. The STIs have several effects if they are not cured on time or use the wrong medications. The STIs can induce infertility or sterility, make the body prone to more serious diseases like HIV, and even cause death. The stigma and humiliation associated with sexually transmitted infections create significant hurdles to seeking effective diagnosis and treatment. This study aimed to develop a machine learning model integrated into a web application to facilitate seamless communication between patients and health centres, specifically addressing communication challenges between sexual health clinics and STI patients. Both qualitative and quantitative research methods were employed in the study. Qualitative data were gathered through interviews with health practitioners and ICT officers from the respective hospitals, while quantitative data were collected using survey questionnaires from four hospitals, supported by the Government of Tanzania Health Operation Management Information System (GoT-HoMIS). Dataset with features which included several STI symptoms and the label features which are laboratory diagnosis results. The model was trained on a local dataset using five machine learning algorithms: AdaBoost, Support Vector Machine (SVM), Random Forest, Decision Tree, and Stochastic Gradient Descent (SGD). In this study, results revealed that the highest accuracy score was 97.45% and the F1 score of 97.70% from the AdaBoost classifier. Thus, the model from the AdaBoost algorithm was serialised for integration with the web app. The validation of the web app system was done with a higher number of people recommending the system to be used in the Health Information Management System. The developed machine learning model can benefit policymakers and health practitioners by using telemedicine to enable remote diagnosis and patient monitoring. Apart from telemedicine, the model can remove stigmatisation barriers among STI patients. And lastly, a machine learning-powered system can increase patient adherence to medication and treatment strategies by anticipating future noncompliance and offering timely reminders or interventions.
Development Of A Web-Based System To Enhance Monitoring Laboratory Order And Result Dissemination A Case Study Of Softmed Company Limited
(NM-AIST, 2025-08) Abashe, Japheth
Laboratory testing plays a critical role in clinical decision-making and patient care. However, in many healthcare settings, especially in developing countries, laboratory order management remains manual and paper-based, leading to inefficiencies, errors, and delays in result dissemination. These challenges compromise the quality of care and timely treatment decisions. Despite existing studies highlighting errors in laboratory workflows, especially in the pre-analytical phase, few interventions target private, multi-hospital laboratory services in Tanzania. This study aimed to improve operational efficiency and result dissemination at SoftMed, a private pathology laboratory in Arusha, Tanzania, through the development of a web-based laboratory order and result dissemination system. The system was designed to automate test acquisition, streamline inter-facility coordination, and reduce operational bottlenecks. A qualitative approach was used, involving interviews and system validation by six clinical and laboratory staff directly involved in lab order processes. Development followed the Agile Extreme Programming (XP) methodology, using Angular for the web interface, Laravel for the backend, and MySQL for data storage. Validation assessed improvements in turnaround time, communication, and error reduction. Results indicated notable improvements in workflow efficiency, reduced turnaround time, and enhanced communication between healthcare providers and lab staff. Although the validation sample was limited, feedback confirmed the system's operational effectiveness in the specific private lab context. This study contributes practical evidence supporting lab workflow automation in Tanzania’s private healthcare sector and recommends futureintegration with electronic medical records (EMRs) for comprehensive health information management.
An Interoperable Data-Informed Procurement And Distribution Information System For Enhanced Stock Control
(NM-AIST, 2025-01) Babu, Jacinta Akinyi
This report presents a comprehensive study on developing and implementing an Interoperable Data-Informed Procurement and Distribution Information System designed to enhance stock control across multiple branches of Transchem Pharmaceuticals. The pharmaceutical industry relies heavily on efficient supply chain management, and the research addresses the challenges of poor visibility and coordination between procurement and distribution functions, which result in operational inefficiencies and increased costs. The study employs qualitative research methods, including focus group discussions, semi-structured observations, survey questionnaires and a benchmark test to gather data and analyse Transchem Pharmaceuticals' operations. The Agile Approach was used for system development, specifically Extreme Programming. The developed system exhibited a modestly superior overall multi-core performance compared to the baseline system. The baseline system performed better in file compression by a slight margin, surpassing the developed system's navigation efficiency. The developed system showed a marginally better performance in HTML5 handling and the baseline system demonstrated a slight advantage in PDF rendering. Key findings include the system's unforeseen benefit of providing valuable insights into consumer preferences and demand patterns, supporting more informed decision-making and strategic planning. The project's implications in supply chain management and information systems underscore the importance of leveraging real-time data for decision-making in contemporary business practices. This research contributes to the advancement of enhanced stock control practices across company branches and demonstrates the effectiveness of data- driven insights in supply chain management. Future research directions include exploring machine learning-driven predictive demand forecasting, emerging technologies integration such as blockchain technology for transparency, and adoption of robotic process automation to accelerate routine processes.
Mobile-based application to predict children under five nutritional status using machine learning techniques
(NM-AIST, 2025-07) Selemani,Bakary
Malnutrition tends to be one of the most important reasons for child mortality in Tanzania and other developing countries, in most cases during the first five years of life. This research was conducted to develop a machine-learning model for predicting fetal nutritional status. Several machine learning techniques, such as AdaBoost, Logistic Regression, Support Vector Machine, Random Forest, Naïve Bayes, Decision Tree, K-nearest neighbor and Stochastic Gradient Descent. Because of their ability to predict more categorical data, so were used to categorize the children in the test dataset as "malnourished" or "nourished". These algorithms' prediction abilities' accuracy, sensitivity, and specificity were compared using performance measures such as accuracy, sensitivity, and specificity. Results show that Random Forest machine learning techniques outperformed other techniques with an accuracy of 98%. The study findings of this research are indicating a need for more attention on the nutritional status of expected mothers and children under five to be well administered with the government and the society at large by putting relevance to the suggestion that cooperation between government organizations, academia, and industry is necessary to provide sufficient infrastructure support for the future society.
Development of a mobile-based climate information and advisory system for crop management: a case of musanze district, rwanda
(NM-AIST, 2025-08) Angelique,Mukaneza
Mobile applications and internet access enable climate change adaptations. Climatic information services package and deliver climatic data to customers, including temperature, rainfall, wind, and soil moisture. In Rwanda, farmers are accessing climate information through radios, television, trained agents through the Rwanda Meteorological Agency, and by weather applications. However, farmers claim that because of weak dissemination channels and not interacting, they are facing the main challenges in making decisions at the right time for achieving sustainable food production and security, which provides lower incomes and famine to society. We developed a bilingual (Kinyarwanda and English) mobile application using the Flutter framework with a Firebase backend to address these challenges. We integrated the Open Weather API for real-time and forecast data. An SMS gateway was incorporated to ensure notifications reach farmers even in low-connectivity areas. The system was evaluated in Musanze District, where it delivered current and forecast temperatures between 15.5 ℃ and 21.9 ℃, overcast or rainy skies, 55 % humidity, 801.5 hPa pressure, 1.5 m/s wind speed, sunrise at 07:04 AM, and sunset at 07:06 PM directly to users’ devices. Problem reporting and real-time conversation with agronomic officers provided individualized advising help. The Dfarmer app enabled farmers to make timely decisions and increase crop output by providing interactive climatic information and advisory services in their local area.
Cybersecurity framework for e-waste disposal in tanzanian public institutions
(NM-AIST, 2025-08) Mustapha,athuman
The proliferation of electronic devices in Tanzanian public institutions has increased electronic waste (e-waste), posing cybersecurity risks due to improper disposal. Despite increased digitization in Tanzanian public institutions, there remain critical gaps in the secure disposal of electronic waste. Current practices often lack integration between IT asset lifecycle management and appropriate data sanitization protocols tailored to institutional contexts, resulting in heightened risks of sensitive data exposure during device disposal. This research aimed to investigate e-waste disposal practices, identify vulnerabilities, and develop a cybersecurity framework for e-waste disposal. Conducted in Tanzania, the study employed mostly qualitative methods with semi-structured interviews and focus groups to gather data from IT personnel, e-waste officers, and policymakers from 11 public institutions, and quantitative surveys for validation. Thematic analysis, guided by Routine Activities Theory, revealed inconsistent regulations, reliance on ad hoc disposal (e.g., auctions), and inadequate data sanitization, leaving sensitive data vulnerable. The Cybersecurity Framework for E-Waste Disposal (CFED) was developed, integrating IT Asset Management with risk-based sanitization protocols (e.g., cryptographic erasure for high-sensitivity devices). Validated by 15 experts (86.7% feasibility), the CFED aligns with NIST SP 800-88 and GDPR. Recommendations include policy reforms, CFED adoption, and capacity-building. This study bridges cybersecurity and e-waste management gaps, offering a scalable solution for resource-limited settings, with future work exploring broader implementation.
A Web-Based Private Permissioned Blockchain for Ionizing Radiation Management: A Case Study of The Tanzania Atomic Energy Commission
(NM-AIST, 2024-09) Khwatenge, Elvira Immaculate
This dissertation is about a study on insurance companies that have experienced ruin but have a possibility of recovery from ruin. The study has proposed a perturbed mathematical model, analysed and used it for modelling the portfolio of insurance companies with possibilities of recovery after ruin. Return on investment and refinancing have been used as approaches for overcoming ruin. The model was analysed for various cases of possibilities of recovery after ruin in the closed interval [0, 1]. The basic perturbed classical risk process was later compounded by refinancing and return on investment. The Hamilton-Jacobi-Bellman and Integro-Differential Equation of Volterra type were obtained. The Volterra Integro-Differential Equation for the survival function of an insurance company was converted to a third order ordinary differential equation and later converted into a system of first order ordinary differential equations which was solved numerically using the fourth order Runge-Kutta method. The results indicate that the return on investment plays a vital role in reducing ultimate ruin and that as the possibility of recovery for insurance companies increases, the return on investment reduces ruin much faster. Also, the survival function increases with the increasing intensity of the counting process but decreases with an increase in the instantaneous rate of stock return and return volatility. Because an insurance company faces more risks, these results also suggest that insurance companies should increase their counting process since doing so will help the insurance companies in servicing more customers.
Mathematical Models for Chikungunya Virus: Effects of Heterogeneity and Periodic Environmental Variations
(NM-AIST, 2024-08) Lusekelo, Eva
In recent years, chikungunya, a mosquito-borne viral disease has spread globally and invaded new habitats and, as such, it is now regarded as one of the global threats to humanity because of its highly debilitating nature and unprecedented magnitude of its spread. Precisely, there is inadequate mechanistic understanding of how environmental factors, socioeconomic factors and disease intervention strategies, combined, affects epidemic magnitude and duration. In this study, we developed four mechanistic models for Aedes aegypti mosquitoes and chikungunya virus transmission that incorporates relevant ecological and biological factors, socio-economic factors and disease intervention strategies. The first model quantifies the effects of biological control and temperature on the growth of Aedes aegypti mosquito population in the environment. Utilising empirically derived temperature functions in literature, we observed that temperature ranges from 29^0 C to 35^0 C supports maximum egg hatching as well as development of larva and pupa. We also computed the vector reproduction number and examined the influence of entomological parameters on its magnitude. In addition, results also revealed that the attack rate of aquatic predators has higher impact to reduce mosquito population compared to the size of the predator population in the environment. In the second model, we developed a mathematical model to determine optimum timing of rolling out intervention strategies during a chikungunya virus outbreak. The proposed model incorporates three intervention strategies, physical barriers, larvicide and insecticide. Making use of optimal control theory, parameter sensitivity analysis, and numerical simulations, we performed a cost-effective analysis of the aforementioned intervention strategies. Findings from the proposed model offer a framework for designing cost-effective strategies for chikungunya with multiple intervention methods. Temperature and heterogeneous biting exposure are known to be integral factor capable of altering the spread of chikungunya during an outbreak. To quantify the role of these factors, we developed two mechanistic models, an autonomous and a non-autonomous. In a non-autonomous model temperature varies with time while in the autonomous it is regarded to be constant. In all scenarios, analysis of the model showed that both temperature and heterogeneous biting exposure have a substantial influence on shaping the transmission of the disease during an outbreak. Besides, temperature and heterogeneous biting exposure, the non-autonomous model incorporated mass media campaigns. Upon evaluating the implications of mass media campaigns during chikungunya outbreak, we observed that if 20% of infections are detected and reported, with mass media campaigns at 90% efficient, then the new infections produced may decrease by 68.8% over a four-year period. Overall, our results showed that temperature, predation, heterogeneous biting exposure and mass media campaigns play an essential component in determining both the short and long chikungunya virus dynamics.
Enhancement for the access and utilization of library resources using machine learning techniques
(NM-AIST, 2024-06) Kato, Agrey
The growing demands for online information have motivated researchers to explore the most effectively use of digital library (DL) resource tools. The main challenges of online DL are information search and retrieval attributes related to label relevance and feature correlation segments. Previous research mainly relied on unbalanced multi-label data and therefore could not develop a reliable tool to access online information. To improve availability and usefulness of online DL, this work uses machine learning techniques to enhance the access and utilization of library resources. The research data were collected at The Nelson Mandela African Institution of Science and Technology (NM-AIST), Mzumbe University (MU), and the University of Dar es Salaam (UDSM) through questionnaire and purposeful sampling technique were then analysed with python and MAXQDA tools respectively. The survey found that 1,217 (73%) of respondents were aware of electronic information resources (EIRs) but faced accessibility limitations due to social and technical issues. Then, the proposed ensemble model (PEM) for machine learning (ML) methods was used to develop a resource discovery tool (RDT). The effectiveness of the PEM was then evaluated by comparing the accuracy of the PEM, logistic regression (LR), support vector machine (SVM), and knearest neighbor (kNN) algorithms. The experimental results reveal that PEM offers the highest precision of 95%, as compared to LR's 84%, SVM's 65%, and kNN's 57%. The Web Content Accessibility Guidelines (WCAG) 2.1 standards had been successfully used to test the four digital library tools, the developed RDT, NM-AIST, MU, and UDSM to see how well the developed system performs. The developed RDT had the highest established compliance score for online content accessibility, which is 90% with only one violation, compared to NM-AIST's 80% with 16 violations, MU's 55% with 12 violations, and UDSM's inability to be evaluated because of the excessive number of infractions. Therefore, the results of this study show the need to regularly check the accessibility of an online resources as well as optimization of the digital libraries.
A framework for timely payments of property taxes in local government authorities: a case of Tanzania
(NM-AIST, 2024-08) Gration, Justuce
Many governments in developing counties, including Tanzania depend on different taxes charged from various sources such as income taxes from an individual, income taxes from cooperate, taxes on services and goods, property taxes and taxes from social insurance as main sources of revenues. Numerous studies conducted worldwide have shown that tax avoidance and late payments by taxpayers still occur often. Some taxpayers' actions include hiding their income from tax authorities and believing that many other taxpayers also break the applicable laws because they are confident there is little chance of being caught. Even though African countries' laws appear to offer sufficient methods for tax enforcement, some of them are politically unpalatable and are therefore not put into operation. This study focused on designing a framework for enforcing prompt property taxes payment in Tanzania. A strategy based on case studies was used to collect data from two regions, Dar es Salaam and Dodoma, to explore the current difficulties Tanzania is having with collecting property taxes. The questionnaire and interviews involved 10 tax collectors, 150 property owners, and 10 property taxes information systems administrators. The findings revealed that Tanzania still has some challenges in collecting property taxes, including ineffective involvement of Local Government Authorities in provision of taxpayers’ education, higher costs related to tax collection, a high number of tax defaulters, slow payment methods, and impractical enforcement measures. The design science research methodology was used in designing the framework, and extreme programming was adopted in developing the application prototype to put the designed framework into action. The prototype application's results demonstrated that the best way to enforce timely payment of property tax is cutting off electricity to tax defaulters. Through the developed application prototype, the framework was validated and more than 80% of the participants agreed that the designed mechanism is useful, more than 83% of participants said that the mechanism is appropriate and 86.7% of participants said the application is easy to use. Since the study is based on enhancing existing information systems functionality, it is contributing to understanding of how to improve and optimize the information systems to better meet user needs. The study consequently advises property taxes authorities to consider the designed enforcement mechanism that would aid in overcoming the highlighted challenges and enable them to capture more income from these highly viable revenue sources.
Modelling the transmission dynamics and control of aflatoxins crops and its associated health risks in livestock and humans
(NM-AIST, 2024-07) Mgandu, Filimon
Aflatoxin contamination poses a significant challenge to food safety and security, as it affects both the health of consumers and the entire supply chain. Doses of aflatoxins beyond accept able levels are dangerous and may lead to poisoning, also called aflatoxicosis, a life-threatening illness. Liver damage or liver cancer, especially for people who may have conditions such as hepatitis B infection, is also caused by aflatoxin consumption. This study aimed to investigate the transmission dynamics and control of aflatoxin contamination in crops and its associated health risks in livestock, and humans. A deterministic mathematical model to study transmis sion dynamics was formulated and analyzed. Partial Rank Correlation Coefficients (PRCCs) for global sensitivity analysis were calculated using Latin Hypercube Sampling (LHS) to de termine how sensitive and significant the parameters are for each variable. Three controls, namely good farming practices, biological control, and public education and awareness cam paigns, were analyzed. The optimal control theory and cost-effective analysis were performed to identify the most effective strategy for aflatoxin contamination mitigation in crops, live stock, and humans. Four machine learning algorithms: Gaussian Process Classification (GPC), Support Vector Machine (SVM), Random Forest Classifier (RFC), and K Nearest Neighbors (KNN) have been used to predict aflatoxin contamination in maize and groundnuts. The anal ysis of the mathematical model formulated shows that aflatoxin contamination-free equilib rium (ACFE) and aflatoxin contamination-persistence equilibrium (ACPE) exist. The ACFE is globally asymptotically stable if the basic aflatoxin contamination number R0 < 1 whereas the ACPE is globally asymptotically stable if R0 > 1. Numerical simulations showed that a decrease in crop contamination and shedding rates and an increase in the death rate of aflatoxin fungi in the environment by 50% reduced the basic contamination number by above 92%. Re sults from the optimal control analysis suggest that implementation of all controls performs better than other strategies in controlling aflatoxin contamination in crops, livestock, and hu mans. Therefore, to control aflatoxin contamination, initiatives should focus on good farming practices, biological control, and public education and awareness campaigns. In predicting aflatoxin contamination, GPC outperformed other models with an accuracy of 96% and 95% in groundnut and maize samples, respectively. Moreover, the study revealed that humidity and rainfall have a greater influence on predicting aflatoxin contamination compared to tempera ture.
Modelling the impacts of anthropogenic activities on forest biomass and dependent wildlife population
(NM-AIST, 2024-05) Fanuel, Ibrahim
The depletion of forest biomass and declining of forest-dependent wildlife populations are ur gent ecological and societal issues resulting from human activities such as deforestation and land-use changes. This study aims to comprehend the influence of anthropogenic activities on forest biomass and the population of wildlife dependent on forests, and to formulate appropri ate management measures. Specific objectives include forecasting forest land loss in Tanzania, analysing mathematical model describing the impact of human activities on forests and wildlife, examining the influence of fuzzy parameters on model dynamics, and evaluating the effects of economic measures and technological efforts on conservation. The study considered Tanzania where local communities heavily rely on forest resources for their livelihoods. This region also supports a rich biodiversity of wildlife species, where the forest provides essential habitats and resources for their survival. Furthermore, the study also acknowledged the existence of vari ous human activities carried out in the area which may have a significant impact on the forest ecosystem and the wildlife populations that depend on it. Four models are presented: a time series model and three dynamical system models. The key findings include: (i) The univariate time series model accurately predicts an increase in forest land loss in Tanzania with a 96.2% accuracy rate (MAPE = 0.0377), highlighting the urgency for sustainable forest management practices and conservation policies. (ii) Depletion of forest biomass due to human activities has severe implications for wildlife survival and ecological balance. Achieving this balance re quires ensuring the growth rate of forest biomass and wildlife populations exceeds their rates of utilisation and depletion, respectively. (iii) Incorporating fuzzy parameters improves model re liability by accounting for uncertainties in climate, geography, and human activities, enhancing decision-making processes. (iv) Economic measures and technological efforts have the po tential to conserve forest biomass and wildlife populations. However, careful implementation and comprehensive understanding of forest ecosystem dynamics are crucial to prevent desta bilisation. The study emphasises the need for interdisciplinary collaboration and stakeholders engagement to ensure sustainable forest use and conservation, considering the complexities and uncertainties of natural systems. Balancing forest conservation with socio-economic needs is key for the well-being of local communities and future generations.
A monitoring system for transboundary foot and mouth disease considering livestock keepers demographic characteristics
(NM-AIST, 2023-03) Kijazi, Ahmed
Foot and Mouth disease (FMD) is a transboundary disease caused by a virus that affects domestic and wild cloven-hooved animals such as sheep, goats, pigs, and buffalos. FMD is transmitted from one animal to another through direct or indirect contact. Apart from other animal diseases, FMD has been given great attention due to its unique behaviour, such as being potentially dangerous, rapidly spreading disease, and it has no cure. Therefore, immediate information flow among livestock stakeholders could help to mitigate FMD. Realizing the importance of animal disease surveillance, many agencies developed systems for monitoring animal health (fast disease reporting and response). The challenge is that they were developed using advanced technologies like web-based and android, requiring skills, internet connectivity, computers, and smartphones to access them. However, most livestock keepers lack these facilities, especially in developing countries. In that case, they deny access to livestock keepers positioned at the grass-root of animals’ disease reporting chain since illnesses always begin with their animals. Therefore, their lack of participation in reporting or receiving animal disease information through the electronic-based animal disease surveillance system causes a delay in identifying and reporting disease cases and provides insufficient information for controlling contiguous diseases like FMD, which require more precautionary measures through timely information sharing. This study aims to bridge the gap between livestock keepers and top-level stakeholders by developing an animal diseases surveillance system named “Monitoring System for Transboundary Foot and Mouth Disease Considering Livestock Keepers Demographic Characteristics (AMoS4T- FMD)”. The system provides a standard platform for sharing FMD-related information between top- level stakeholders and livestock keepers in time using various mobile technologies based on their demographic characteristics. Gairo district in the Morogoro region was selected as a study area. Therefore, the surveillance system was developed and tested in Gairo district settings. However, it has flexible settings to work elsewhere. In Gairo, livestock keepers’ mobile phone usage and demographic data were collected to determine the appropriate mobile technologies to communicate animal disease surveillance information among themselves and top-level stakeholders through AMoS4T-FMD. After that, an algorithm (FMD communication algorithm) which enables livestock keepers to communicate with AMoS4T-FMD using Unstructured Supplementary Service Data (USSD), Short Message Service (SMS) and Robot calls (Robocalls) based on their demographic data was developed. Also, a Model for predicting and alerting FMD outbreaks in the Gairo district using an Agent-Based Simulation modelling technique was developed. Lastly, the FMD communication algorithm and the Agent- ii Based Simulation model were combined into the software using the waterfall model for system development. Finally, the system was tested using verification and validation techniques.
Development of an implementation framework and social media analytics tool for Tanzania's tourism small and medium-sized enterprises
(NM-AIST, 2023-08) Madila, Shadrack
Tourism is among the sectors that contribute greatly to the economic development of many countries. The industry contributes the growth of countries' economies and employment in 2019 it contributed 10.3% of the global gross domestic product and 330 million jobs. Majority of organisations in the tourism sector operate as small and medium enterprises (SMEs). Tourism SMEs prominently use ICT services including social media in their daily business activities. Performing social media analytics has the potential to bring maximum advantage to social media business users as it can provide insights of social media data for added business competitiveness. Tourism SMEs are conducting SMA in their business activities even though there is no framework to govern the process. There is a need for these tourism SMEs to have an implementation framework that governs the implementation and management of the SMA process. This study aims to develop a social media analytics implementation framework and a social media analytics tool for tourism SMEs in Tanzania. The study used questionnaires to survey tourism SMEs in the Arusha and Kilimanjaro regions to determine their social media analytics practices. The study found that majority of tourism SMEs 73% have not adopted the use of social media analytics tools and technologies. The 27% of SMEs that have adopted the use of social media analytics tools perform simple analytics using built-in tools from social media platforms, and don’t have guidelines or procedures to aid the implementation of social media analytics. The study is unique in that, it proposes a social media analytics implementation framework to assist tourism managers in implementing social media analytics in their social media platforms and the social media analytics tool. The framework was evaluated by tourism SMEs and the majority 86.66% agreed that it is appropriate for their usage. The study also introduces the social media analytics tool which will provide insight into the social media data of tourism SMEs. This research contributes knowledge and information about social media analytics to tourism SMEs managers and owners and provides the implementation framework and the social media analytics tool to tourism managers.
Fabrication of liposome-chitosan-zno nanohybrid integrated with Carissa Spinarum extract for antibacterial application
(NM-AIST, 2023-08) Rubaka, Clarence
Innovative biomaterials provide a stimulating and adaptable platform for the implementation of new and more effective methods to prevent bacterial infection. Built on biomimetic- inorganic hybrid material, Dual Nanohybrid Delivery System (DN-DS) has advantageous properties for biomedical applications, such as the delivery of herbal formulations for the treatment of bacterial infections. Using microwave assisted extraction (MAE), the polyphenols of Carissa spinarum were extracted. The Dual Nanohybrid Delivery System (LipCsP-ZnONPs)-CT was formed by combining LipCsP-Chitosan and ZnO-Chitosan, which were both generated using different methods of co-precipitation and ion gelation, respectively. A Zetasizer was used to characterize the nanosystems' size, zeta potential, and polydispersity index (PDI). A UV-visible spectrophotometer was utilized for the optical study, and a scanning electron microscope was employed to investigate at the surface morphology. The interaction of coated chitosan with liposomes and ZnONPs was evaluated using Fourier Transformation Infrared (FTIR) spectroscopy. Different kinetic models were fitted to the results of the encapsulation and release profiles of polyphenols in the liposome nanosystems to determine the mechanism of release. Antibacterial activity of the nanoformulations was assessed by an agar diffusion assayand the micro plate blue assay (MABA). The Zeta potential of LipCsP changed from -45.3 ± 0.78 to +34.43 ±1.36 due to chitosan coatings. Polyphenol-encapsulation efficiency was higher in LipCsP-Chitosan (81 ± 2.5%) than in LipCsP (66.11 ± 1.11%). Conversely, the size of LipCsP (176.17 ± 1.05 nm) increasedto 365.2 ± 0.70 nm. FTIR analysis revealed the interaction of the liposome with chitosan due to the disappearance of N-H primary amine. Interaction between chitosan and zinc oxide was revealed by the formation of new absorption peaks at 670 cm-1 and 465 cm-1 as observed in the FTIR analysis. (LipCsP-ZnONPs)-CT presented high bioaccessibility of polyphenols in the simulated gastric phase (82.14 ± 0.80%) than in simulated intestinal phase (71.60 ± 0.86%), a stable system for sustained release of polyphenols, and prominent antibacterial activity. (LipCsP-ZnONPs)-CT exhibited a relative inhibition zone diameter (RIZD) of 89.60 ± 1.32, significant high viability reduction (P˂0.05) against Klebsiella pneumoniae as compared to LipCsP-Chitosan and ZnO-chitosan. The nanohybrid systems (LipCsP-Chitosan and ZnO- chitosan) exhibited synergistic effect against Klebsiella pneumoniae. This study successfully demonstrated the utility of the nanohybrid as a potential antibacterial agent against K.pneumoniae, therefore, the fabricated dual nano delivery system is an efficacy material for treatment of pneumococcal infections.
Integrated machine learning based quality measurement model for maternal, neonatal and child health services in Tanzania
(NM-AIST, 2022-08) Nyanjara, Sarah
The high maternal and neonatal mortality rate has remained a challenge for most developing countries. Scholars link the high death occurrences to the poor quality of health services provided to pregnant women and children. It is further revealed that most deaths could be prevented if women and children could access high-quality maternal, neonatal and child health services. Quality measurement, a process of using data to evaluate healthcare plans and performance, is essential in improving the quality of health services and reducing mortality rates. However, most developing countries and Tanzania lack effective approaches to measure and report the quality of Maternal, Neonatal and Child Health services provided. The Lack of an effective quality measurement approach limits the quality measurement processes and may jeopardize the quality measurement results. Additionally, failure to establish the quality of health services hampers healthcare plans and governance of healthcare supplies and other resources. The available quality measurement approaches require trained data collectors, dedicated datasets and the physical presence of quality measurement personnel at each health facility; therefore, labour intensive and resource inefficient. This study proposed and developed an integrated machine learning-based quality measurement model for maternal, neonatal and child health services in Tanzania. The study employed a machine learning technique, a K-means clustering algorithm, and a dataset selected from the national health information system and data warehouse: “District Health Information System (DHIS 2)”. The developed model clustered the Maternal, Neonatal and Child Health (MNCH) dataset into two groups (clusters), and cluster analysis was performed to discover the knowledge about the quality of health services in each cluster formed. The study also performed model validation to establish the usefulness of the developed integrated machine learning-based model for quality measurement in MNCH. This study brings to the body knowledge an integrated machine learning-based quality measurement model for maternal, neonatal and child health services and a list of important indicators for quality measurement, the essential inputs for an effective quality measurement process. The current quality measurement model requires only data to measure the quality of health services readily available in DHIS 2, making the quality measurement model resource-efficient and ideal for quality measurement in resource-constrained countries such as Tanzania.
Developing a high-performance soil fertility status prediction voting ensemble using brute exhaustive optimization in automated multiprecision weights of hybrid classifiers
(NM-AIST, 2023-08) Josephat, Augustine
With the advent of machine learning (ML) techniques, various algorithms have been applied in previous studies to develop models for predicting soil fertility status. However, these models are observed to use varying fertility target classes, and variations have been reported in these models' predictive performances. As a result, practical applications of these models for obtaining the most accurate predictions may become hindered. While the weighted voting ensemble (WVE) ML technique can be used to improve soil fertility status prediction by aggregating individual models prediction, guaranteeing finding of an optimal WVE assignment weights is challenging. Whereas a brute exhaustive search procedure can be applied for the mentioned task, there is a lack of exploration on the exploitation of automated classifiers' precise weights combinations as search spaces for successful optimization. This research aims to develop a high-performance soil fertility status prediction voting ensemble using brute exhaustive optimization in automated 1EXP(-)Z+ multi-precision weights of hybrid classifiers. Soil chemical properties and ML modeling algorithms for modeling soil fertility status were identified. Base hybrid ML classification models for predicting soil fertility status were evaluated using Tanzania as a case study. Finally, the base ML hybrids WVE models were optimized using brute exhaustive search procedure’s novel developed search spaces generation algorithm for guaranteed optimal solution finding. The research was designed using design science research methodology, with the application of unsupervised machine learning K-mean algorithm with a knee detection method to find the optimal number of soil fertility status target classes, and supervised learning algorithms were applied to model classifiers for those optimal classes. Three soil fertility target classes were identified by clustering technique. The model achieved on test data a predictive accuracy of 98.93%, with respective AUC of 82%, 83%, and 87% for low, medium, and high soil fertility targets classes. Whereas these performances are observed higher compared to models in previous studies, 92% correct classifications were obtained on validation against external unseen laboratory-based tested soil results. Therefore, soil testing laboratories and farmers should consider using the model to smartly manage soil fertility which may lead to improved crop growth and productivity. The government could set agricultural-related policies that require the use of the model by farmers with the provision of agricultural inputs subsidies. Future work could be to develop an integrated real-time web and mobile application for providing farmers with soil fertility status information.
A monitoring system for transboundary foot and mouth disease considering livestock keepers demographic characteristics
(NM-AIST, 2023-03) Kijazi, Ahmed
Foot and Mouth disease (FMD) is a transboundary disease caused by a virus that affects domestic and wild cloven-hooved animals such as sheep, goats, pigs, and buffalos. FMD is transmitted from one animal to another through direct or indirect contact. Apart from other animal diseases, FMD has been given great attention due to its unique behaviour, such as being potentially dangerous, rapidly spreading disease, and it has no cure. Therefore, immediate information flow among livestock stakeholders could help to mitigate FMD. Realizing the importance of animal disease surveillance, many agencies developed systems for monitoring animal health (fast disease reporting and response). The challenge is that they were developed using advanced technologies like web-based and android, requiring skills, internet connectivity, computers, and smartphones to access them. However, most livestock keepers lack these facilities, especially in developing countries. In that case, they deny access to livestock keepers positioned at the grass-root of animals’ disease reporting chain since illnesses always begin with their animals. Therefore, their lack of participation in reporting or receiving animal disease information through the electronic-based animal disease surveillance system causes a delay in identifying and reporting disease cases and provides insufficient information for controlling contiguous diseases like FMD, which require more precautionary measures through timely information sharing. This study aims to bridge the gap between livestock keepers and top-level stakeholders by developing an animal diseases surveillance system named “Monitoring System for Transboundary Foot and Mouth Disease Considering Livestock Keepers Demographic Characteristics (AMoS4T-FMD)”. The system provides a standard platform for sharing FMD-related information between top- level stakeholders and livestock keepers in time using various mobile technologies based on their demographic characteristics. Gairo district in the Morogoro region was selected as a study area. Therefore, the surveillance system was developed and tested in Gairo district settings. However, it has flexible settings to work elsewhere. In Gairo, livestock keepers’ mobile phone usage and demographic data were collected to determine the appropriate mobile technologies to communicate animal disease surveillance information among themselves and top-level stakeholders through AMoS4T-FMD. After that, an algorithm (FMD communication algorithm) which enables livestock keepers to communicate with AMoS4T-FMD using Unstructured Supplementary Service Data (USSD), Short Message Service (SMS) and Robot calls (Robocalls) based on their demographic data was developed. Also, a Model for predicting and alerting FMD outbreaks in the Gairo district using an Agent-Based Simulation modelling technique was developed. Lastly, the FMD communication algorithm and the Agent- ii Based Simulation model were combined into the software using the waterfall model for system development. Finally, the system was tested using verification and validation techniques
Insurance Ccompanies portfolio optimisation with possibilities of recovery after ruin
(NM-AIST, 2023-01) Komunte, Masoud
This dissertation, is about a study on insurance companies that have experienced ruin but have a possibility of recovery from ruin. The study has proposed a perturbed mathematical model, analysed and used it for modelling the portfolio of insurance companies with possibilities of re covery after ruin. Return on investment and refinancing have been used as approaches for over coming ruin. The model was analysed for various cases of possibilities of recovery after ruin in the closed interval [0, 1]. The basic perturbed classical risk process was later compounded by refinancing and return on investment. The Hamilton-Jacobi-Bellman and Integro-Differential Equation of Volterra type were obtained. The Volterra Integro-Differential Equation for sur vival function of an insurance company was converted to a third order ordinary differential equation and later converted into a system of first order ordinary differential equations which was solved numerically using the fourth order Runge-Kutta method. The results indicate that the return on investment plays a vital role in reducing ultimate ruin and that as the possibility of recovery for insurance companies increases, the return on investment reduces ruin much faster. Also, the survival function increases with the increasing intensity of the counting pro cess but decreases with an increase in the instantaneous rate of stock return and return volatility. Because an insurance company faces more risks, these results also suggest that insurance com panies should increase their counting process since doing so will help the insurance companies in servicing more customers.

Browse

Recent Submissions