Novel Logistic Regression over Naive Bayes Improves Accuracy in Credit Card Fraud Detection

Abstract

Aim: To enhance the accuracy in credit card fraud detection using Naive Bayes and Novel Logistic Regression. Materials and Methods: This study contains Naive Bayes and Novel Logistic Regression. Each algorithm consists of a sample size of 70 and the study parameters include alpha value 0.05, beta value 0.2, and the power value 0.8. Their accuracies are compared with each other using different sample sizes also. Results: The Novel Logistic Regression is 93.59% more accurate than Naive Bayes of 85.88% in detecting fraudulent transactions. Significance value for accuracy and loss is 0.255 (p>0.05). Conclusion: The Novel Logistic Regression model is significantly better than Naive Bayes in detecting fraudulent transactions. It can be also considered as a better option for credit card fraud detection.