A Novel Framework for credit card fraud detection

Abstract

Credit card fraud is the challenge of predicting fraudulent transactions based on specific rules. In this paper, Various classification algorithms are implemented on an imbalanced dataset concerning the performance analysis to detect fraud in the credit card. In this study, the dataset is sourced from Kaggle. There are 284,807 transactions, out of which 17% of transactions are fraudulent. Various classifiers that are logistic regression, naive Bayes, AdaBoost, and voting classifiers that are combinations of all mentioned above algorithms are refined. The AI model needs significant historical data to prepare the model. For this, a huge amount of information is given to the model as Training data, While dealing with a lot of information, the model's execution time is expanded, which influences execution. In this study, the voting classifier is applied for the expectation analysis which is a mix of different AI calculations.This voting based classifier increases the complexity of the prediction analysis and also increases execution time. In the future, a hybrid classification model will be designed to detect credit card scams.