Abstract
The banking sector remain
s the heartbeat of every economy since it is the source of finance for all
economic activiti es . Despite being the critical sector of any economy, banks are under attack from
cyber criminals. Consequently, banks and other financial institutions are under pressure to find
fraud defences which can effectively grant security to their assets and finances from these errant
criminals. The current study so ught to develop a framework for using machine learning to combat
fraud in the South African banking sector. The Fraud Triangle Theory and Technological
Determinism Theory informed the stu dy’s theoretical framework . The study adopted the mixed
research method approach to complement the weaknesses of using a sing le research method .
T hus a mixture of quantitative and qualitative tools was applied in the current research. In
analysing quantitative data from the survey , p artial l east s quare s tructural e quation m odel ling
(PLS SEM ) was used and thematic analysis in Maxqda 24 software was also used for qualitative
data analysis The study established that there is high E banking fraud prevalence in South Africa,
and it is causing both financial and non financial losses to banks and their customers. P ressure,
opportunity and advances in technolog y were found to be the key drivers of electronic banking
fraud prevalence in the South African banking sector. However, other electronic banking fraud
prevalence driving factors found in South Africa incl ude, inadequate systems, poor internal controls
and senior management oversight, lack of budget for advanced technological defences, bribery
and corruption within the judiciary system, lack of investigation skills, lack of fraud data, and lack of
client and employee education and awa reness.
To effectively
combat E banking fraud, the study found that machine learning algorithms are very
effective in preventing and detecting fraud because of their distinct capabilities to interrogate and
analyse complex data or scenario s . A lso , the study show s that mach ine learning methods can be
integrated with already existing risk management frameworks to effectively reduce E banking fraud
cases . In addition, through the explanatory factor analysis utilising principal axis factor ing in IMB
SPSS , the study established that factors such as continuous monitoring of fraud events (CMFE),
s trong internal controls and systems (SICS), s trict laws and regulations (SLR) and e mployee and
customer education and awareness (ECEA) can be implemented by the South African banking
sector to eradicate E banking fraud. Therefore, the proposed framework of the current study clearly
outlines how machine learning algorithms can b e synchronised with the four pil lars of fraud risk
management namely deterrence, prevention, detection and reporting and other factors to combat
E banking fraud t o enhance the accountabili ty of banks’ assets and finances .
The
implications drawn from the findings and analyses of the current study are very significant to
the academia, accounting practi ti oners, bank management and executives for bo th local and
international institutions and policymakers .
iv
KEYWORDS
KEYWORDS
Machine
Machine LLearning, earning, South African banking sector, South African banking sector, EElectronic lectronic BBanking, anking, EElectronic lectronic BBanking anking FFraudraud, , Fraud Fraud DDeterrence, Fraud Prevention, Fraud eterrence, Fraud Prevention, Fraud DDetectionetection, , Fraud Fraud RReporting, eporting, and and Fraud Fraud RRisk isk MManagementanagement..