A Causality Learning of E-banking Operational Risk using Tree Augmented Naïve Bayes Classifier
Ako Rita Erhovwo1, Okpako Ejaita Abugor2

1Dr. Rita Erhovwo Ako, Mathematical Sciences, Edwin Clark University, Kiagbodo, Nigeria.
2Dr. Ejaita Abugor Okpako, Mathematical Sciences, Edwin Clark University, Kiagbodo, Nigeria.

Manuscript received on September 15, 2018. | Revised Manuscript received on September 19, 2018. | Manuscript published on November 30, 2018. | PP: 22-38 | Volume-8 Issue-4, November 2018. | Retrieval Number: D3168118418/2018©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: E-banking systems have been shown to increase and modify particularly Operational Risk (OR). It has increased the technical complexity of the banks operational and security issues. The mode of occurrence, magnitude, and consequences often takes on new dimensions. It has become increasingly important to effectively identify potential OR issues underlying the E-banking operations, their causal relationships, the effectiveness of controls implemented, the inherent risk exposure level, and the residual risk. This research work seeks to propose Tree Augmented Naïve Bayes (TAN) Classifier in the modeling of the causal relationships among operational risks factors. To validate the proposed use of TAN classifier, we comparatively analyzed the performance of the TAN classifier with three other soft computing tools; C4.5 Decision Tree, Naïve Bayes (NB) and Artificial Neural Networks (ANN). These soft computing tools were evaluated in terms of the CPU training time complexity, classification measured by prediction accuracy, ranking measured by AUROC, and the Mean and Relative absolute error rate. The dataset was pre-processed and transformed by conducting a factor analysis procedure using SPSS statistical measurement tool, to identify risks that may require urgent actions and to reduce the dimensionality of the dataset into a smaller subset of most significant measurable variables. WEKA was then used as the developmental tool for training and testing the soft computing classifiers. Through causality learning from the collected E-banking Customers’ data, we demonstrated that the proposed classifier cannot only discover causalities but also perform better in prediction than other algorithms, such as C4.5, NB, and ANN. The TAN network structure revealed the interdependencies among operational risk factors.
Keywords: Causal Relationships, Operational Risk, Soft Computing, Classifiers, E-banking.