Statistical and Mathematical Learning: an application to fraud detection and prevention
- Authors: Hamlomo, Sisipho
- Date: 2022-04-06
- Subjects: Credit card fraud , Bootstrap (Statistics) , Support vector machines , Neural networks (Computer science) , Decision trees , Machine learning , Cross-validation , Imbalanced data
- Language: English
- Type: Master's thesis , text
- Identifier: http://hdl.handle.net/10962/233795 , vital:50128
- Description: Credit card fraud is an ever-growing problem. There has been a rapid increase in the rate of fraudulent activities in recent years resulting in a considerable loss to several organizations, companies, and government agencies. Many researchers have focused on detecting fraudulent behaviours early using advanced machine learning techniques. However, credit card fraud detection is not a straightforward task since fraudulent behaviours usually differ for each attempt and the dataset is highly imbalanced, that is, the frequency of non-fraudulent cases outnumbers the frequency of fraudulent cases. In the case of the European credit card dataset, we have a ratio of approximately one fraudulent case to five hundred and seventy-eight non-fraudulent cases. Different methods were implemented to overcome this problem, namely random undersampling, one-sided sampling, SMOTE combined with Tomek links and parameter tuning. Predictive classifiers, namely logistic regression, decision trees, k-nearest neighbour, support vector machine and multilayer perceptrons, are applied to predict if a transaction is fraudulent or non-fraudulent. The model's performance is evaluated based on recall, precision, F1-score, the area under receiver operating characteristics curve, geometric mean and Matthew correlation coefficient. The results showed that the logistic regression classifier performed better than other classifiers except when the dataset was oversampled. , Thesis (MSc) -- Faculty of Science, Statistics, 2022
- Full Text:
The diversity and dry season habitat associations of fish communities in the Kabompo River Basin, Upper Zambezi, Zambia
- Authors: Rennie, Craig Lawrence
- Date: 2022-04-06
- Subjects: Biodiversity Zambia Kabompo District , Ecology Zambia Kabompo District , Fish communities Zambia Kabompo District , Fishes Ecology Zambia Kabompo District , Freshwater habitats Zambia Kabompo District , Fishes Effect of human beings on Zambia Kabompo District , Fishes Climatic factors Zambia Kabompo District , Mesohabitat
- Language: English
- Type: Master's thesis , text
- Identifier: http://hdl.handle.net/10962/232261 , vital:49976
- Description: The Zambezian headwaters contain diverse fish communities which support thriving fisheries. This region and its fishes are under pressure from multiple anthropogenic threats, including overexploitation and large-scale developments with potential knock-on effects for the riparian communities. Previous studies have focused on known fisheries areas and the mainstem Zambezi River, neglecting large tributaries such as the Kabompo River. Consequently, little literature is available on the diversity and ecology of the fishes that inhabit these large tributaries, hindering the effective management and protection of biodiversity. This study aimed to fill the current knowledge gaps in the diversity and habitat associations of fish communities in the Upper Zambezi, using the Kabompo River as a case study. The first objective was to provide an updated checklist of the fishes of the Kabompo River basin using a compilation of historical data and field surveys. This study detailed the occurrence and distribution of 83 fish species within the Kabompo River basin. All these species have been recorded in the Upper Zambezi, with some of their ranges extending into the Middle (29 species) and Lower Zambezi (23 species) while others have more restricted distributions. The most diverse families were the Cyprinidae (26 species) and the Cichlidae (15 species). A number of potential undescribed species, whose taxonomic distinctiveness need further investigation were also collected. Taxonomic conflicts are also highlighted for some of the taxa that were previously considered to have broad geographic ranges or disjunct distributions. Consistent with findings from other studies within the region, the current taxonomy underestimates the diversity of fishes in the Kabompo River and Upper Zambezi. The second objective was to assess the habitat use of small-bodied fish communities during the low-flow period in 2019. Reconnaissance trips identified dominant mesohabitats along the middle Kabompo River around Jivundu. A total of 139 mesohabitats were sampled across the five dominant mesohabitats identified; Phragmites mauritianus, wood, rock, Vallisneria aethiopica and bare substrate. Catch per unit effort, species richness, Shannon diversity and Pielou’s evenness differed significantly between these mesohabitats. Twenty-six species showed significant associations (p < 0.05) with a specific mesohabitat type or environmental variable (current velocity or depth). Eight species were associated with the woody habitat, with three of these, E. radiatus, E. unitaeniatus, and P. ngamensis being almost exclusively associated with this habitat. Enteromius kerstenii, E. lineomaculatus and S. depressirostris were almost exclusively associated with P. mauritianus and represent potential indicator species for this habitat. A number of species were also with both P. mauritianus and woody habitats. Nine species showed statistically significant associations with the rocky mesohabitat, with Amphilius uranoscopus and Petrocephalus longicapitis, being almost exclusive to rocky sections of the river. Therefore, species such as A. uranoscopus are potential indicators for monitoring the integrity of rocky habitats under threat from sedimentation. The strong associations indicate that this comprehensive baseline may be valuable indicators/ proxies for identifying anthropogenic induced change in the Kabompo basin. This would provide a basis to determine fish responses to regional environmental changes associated with human activities. , Thesis (MSc) -- Faculty of Science, Ichthyology & Fisheries Science, 2022
- Full Text:
The invasion autecology of Iris pseudacorus L. (Iridaceae) in South Africa
- Authors: Sandenbergh, Emma
- Date: 2022-04-06
- Subjects: Iris pseudacorus South Africa , Invasive plants South Africa , Aquatic weeds South Africa , Plant genetics South Africa , Freshwater ecology South Africa , Iris pseudacorus Geographical distribution , Phytogeography
- Language: English
- Type: Master's thesis , text
- Identifier: http://hdl.handle.net/10962/232272 , vital:49977
- Description: Iris pseudacorus L. (Iridaceae) is an emergent aquatic macrophyte originating from Europe, north Africa, and western Asia, and is becoming an increasingly problematic invader in South Africa. By forming dense rhizomatic mats in the absence of natural enemies, I. pseudacorus outcompetes co-occurring indigenous biota, causing serious environmental and socioeconomic challenges. Iris pseudacorus is a declared invader in South Africa, Argentina, New Zealand, the United States of America, and Canada, but little information is known regarding the species’ invasive potential, particularly in the southern hemisphere, hindering the effectiveness of control efforts. This study addresses this knowledge gap in a South African context, providing valuable insight into the invasion autecology of I. pseudacorus in South Africa. For effective management and control of I. pseudacorus in South Africa and the global south, its distribution and invasive potential must be determined, and its population genetics understood. Hence, this study aimed to map the current confirmed distribution of I. pseudacorus populations in South Africa, investigating the relative abundance of I. pseudacorus individuals in each population, and comparing their sexual reproductive outputs. Moreover, this study assessed the competitive interactions between I. pseudacorus and co-occurring native species T. capensis, and examined the genetic diversity present between and within South African I. pseudacorus populations. Through field surveys, I. pseudacorus infestations were confirmed in eight of the country’s nine provinces, with the highest number of infestations recorded in the urban hubs, and greatest population abundances reported in the warmer, wetter regions of South Africa. These surveys indicated that South African I. pseudacorus populations have enhanced their sexual reproductive output relative to native range populations, and a germination rate of ~ 83 % was determined in the laboratory. The results of a common garden competition experiment indicated that T. capensis may be a superior competitor over I. pseudacorus, but this was not supported by field observations, and may be a result of the short duration of the experiment. Using inter-simple sequence repeats (ISSRs), high genetic diversity was observed within and between populations of I. pseudacorus, indicating the employment of sexual reproductive strategies, and providing evidence for gene-flow between and within populations. Moreover, a weak negative correlation was observed between geographic distance and genetic similarity, ii indicating a largely anthropogenic spread of I. pseudacorus, and suggesting the occurrence of fewer founding events than reported in the United States. This study provides useful insight into the invasion autecology of I. pseudacorus in South Africa, contributing to the ongoing research surrounding I. pseudacorus invasions worldwide, particularly in the southern hemisphere. These results contribute to the development of appropriate adaptive and integrated management strategies to control I. pseudacorus invasions in South Africa, and should be implemented before South African I. pseudacorus infestations reach the severity observed elsewhere. , Thesis (MSc) -- Faculty of Science, Botany, 2022
- Full Text: