This paper proposes to design a majority vote ensemble classifier for accurate detection of credit card frauds. There is a lack of research studies on analyzing real-world credit card data owing to the issues of confidentiality, some features were hidden. Credit Card Fraud Dataset from kaggle.com was used for this study. The dataset is made up of 284,807 the number of legitimate transactions was found to be 284,315 while the number of fraudulent transactions was found to be 492, this shows that dataset is highly imbalanced, skewed data set like this may lead to an unreliable prediction performance, which is a major classification problem. Hence this study review literatures on methods of addressing the problem imbalance in dataset, and profile a methods of using the Voter Ensemble approach to address the classification problem. The real-world credit card dataset is analyzed to check for missing values and results show that there were no missing or duplicate values in the dataset. In this study, machine learning algorithms are used to detect credit card fraud. The base learners (Logistic regression, Bagging and Naïve Bayes) and the voter ensemble methods are combined. In this experiment, data are splits into 70% training and 30 % testing, the base (learners) classifiers were trained from the training sets, and predictions evaluations were made on data using 10-fold cross-validation. Afterward, the predictions of the learned base classifiers are combined into an ensemble using a stacked generalization strategy. The experimental results positively indicate that the Voter Ensemble method achieves good performances based on Precision, Recall, F1-score and Accuracy rates in detecting fraud cases in credit cards with 95.77%, 99.99%, 97.83%, 99.99% respectively. The result is evaluated and compared with other existing works. Keywords: Voter Ensemble, Logistic Regression (LR), Naïve Bayse, Bagging, imbalance ratio (IR), CCFD.

No abstract