A Hybrid Approach for Credit Card Fraud Detection Using Naive Bayes and Voting Classifier

Abstract

The prediction analysis is the approach which is applied to predict future possibilities from the current data. One need not pay cash, the cardholder just gives his card to the shopkeeper, he will swipe the card and the payment will be done in just fractions of a second. Sometimes the credit card can be stolen, and then the attacker can make false transactions. In this type of conditions, the credit card company faces a huge loss. In the existing research work, the voting-based classification approach is applied for credit card fraud detection. The voting based classification is a combination of multiple classifiers like SVM, decision tree etc. The classifier will have maximum accuracy will display its predicted result. To improve the accuracy of prediction analysis the voting based classification method will be replaced with naïve bayes classification approach. The naïve bayes classifier is the probability based classifier for credit card fraud detection. In the probability based classification method, the probabilities of the target classes are calculated and the probability of the test data is calculated. The test set which is near to the probability class is identified as the target set. The naïve bayes classification approach will improve the accuracy of credit card fraud detection. The proposed methodology will be implemented in python and results will be analyzed in terms of accuracy, precision, recall and F-measure. The naïve Bayes classifier optimizes the results in terms of accuracy, precision, recall and f-measure is optimized up to 10 to 15% for the credit card fraud detection.