JavaScript is disabled for your browser. Some features of this site may not work without it.
| dc.contributor.author | Binyam, Bekele | |
| dc.date.accessioned | 2022-11-18T07:55:16Z | |
| dc.date.available | 2022-11-18T07:55:16Z | |
| dc.date.issued | 2022-07 | |
| dc.identifier.uri | http://ir.bdu.edu.et/handle/123456789/14477 | |
| dc.description.abstract | ATM (Automated Teller Machine) service is among the important strategies used by banks for quality service. But this service is not as expected by the bank and customers. This results in customer’s dissatisfaction of the ATM service and bank also loses revenue. This study was conducted to develop model detecting ATM transaction failure at CBE by using machine learning algorithm. Based on various reasons these transactions not successful as intended by the customers and also by the bank resulting dissatisfied customers and additional activity of frequent reconciliation. Mainly the problem comes from the status of the ATM. But sometimes, failure of the transaction may be due to problems of customers. Previous studies were not conducted to detect transaction failure by using machine learning techniques. Therefore, this study applied a ML algorithm to detect the transaction failure. The study used transaction status for selected month in a year (2022). Data for the study was collected 27,783 instances from ATM reconciliation data handled by Bole branches in Addis Ababa. The study applied supervised machine learning from classifier algorithms such Decision Tree Classifiers from decision tree learners, Random Forest Classifier from ensemble learners, K-neighbors classifier from neighbors’ learner, Gaussian NB from Naïve Bayes learner, Support Vector Classifier (SVC) from Support Vector Machine (SVM) learner. The machine learning experimentation was conducted by using python 3.7 in Jupyter Notebook (anaconda3). The ATM transaction failure predicting model was developed by using SVC algorithm. Naive Bayes is least performing classifier and SVC is best performing classifier with accuracy of 85.39% and 93.92% correct classification respectively. This study reveals that ATM transaction status can be predicted by using machine learning approach. In addition, this study suggests that application of machine learning model developed by the study enables efficient handling of ATM transaction in CBE. This study showed that ATM transaction failure is linked with transaction properties. Some of the main factors associated with ATM transaction were amount of cash requested by a customer, transaction committed time, type of transaction, workday type; whether the transaction date is committed at work day or not. Key Words: ATM CBE, Supervised ML, Classifier Algorithms, SVC, CNN, | en_US |
| dc.language.iso | en_US | en_US |
| dc.subject | Faculty of Electrical and Computer Engineering | en_US |
| dc.title | Automated Teller Machine (ATM) Transaction Failure detection using Machine learning: The case of commercial Bank of Ethiopia (CBE) | en_US |
| dc.type | Thesis | en_US |
