- Title
- Statistical modelling for detection of fraudulent activity on banking cards
- Creator
- Nasila, Mark Wopicho
- Subject
- Mathematical statistics Mathematical models
- Subject
- Statistics Bank fraud
- Date Issued
- 2014
- Date
- 2014
- Type
- Thesis
- Type
- Doctoral
- Type
- DPhil
- Identifier
- http://hdl.handle.net/10948/45887
- Identifier
- vital:39314
- Description
- The current global recession has highlighted the fragile banking and related systems exposure to risks and acts of fraud. As a result of the ever changing information technology environment, where the internet has become an important retail sector channel, new fraud challenges are being encountered. The rapid growth in credit and cheque card transactions as a payment mechanism has led to an increase in card fraud. Approximately 70% of consumers utilising credit and cheque cards, as payment mechanisms, are significantly concerned about fraud (McAlearney, 2008). Additionally, credit card fraud has broader negative implications, such as funding organised crime, international narcotics trafficking and even the financing of terrorist activities. The first section of this study develops classification models that will improve on existing methods used to detect fraud and, as a result thereof, reduce the number of fraudulent transactions. Using confidential data obtained from a South African Bank, logistic regression and scoring techniques have been combined to develop a classification model that improves on the existing fraudulent identification methods. Using the methods developed in this study, a higher percentage of fraudulent transactions are classified correctly when compared to discriminant analysis, a method often used to identify fraudulent transactions. These models enable the banking business to identify demographic, socio-economic and banking-specific determinants which contribute significantly towards fraudulent transactions. The early detection methods will allow banks to put in place measures that will reduce the occurrence of fraudulent transactions on customer’s cards. The second section involves understanding how card holders and merchants contribute towards the occurrence of fraudulent incidents. This was achieved through two surveys which were carried out in the Johannesburg metropolitan area. These surveys aimed at understanding the perceptions of card holders and merchants with regard to aspects pertaining to card fraud contributed towards the occurrence of card fraud. Multinomial logistic regression (MLR) is used to classify card holders and merchants according to their likelihood of experiencing card fraud incidents. These results are based on their perceptions of certain aspects related to card fraud as obtained from the survey instruments.
- Format
- x, 170 leaves
- Format
- Publisher
- Nelson Mandela Metropolitan University
- Publisher
- Faculty of Science
- Language
- English
- Rights
- Nelson Mandela Metropolitan University
- Hits: 1122
- Visitors: 1181
- Downloads: 211
Collections
NMMU School of Computer Science, Mathematics, Physics and Statistics
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | SOURCE1 | PHD_SUBMIT.pdf | 2 MB | Adobe Acrobat PDF | View Details Download |