Paper Details

Along with the great increase of internet and e-commerce, the use of credit card is an unavoidable one. Due to theincrease of credit card usage, the frauds associated with this have also increased. Fraudsters are continuously trying to findnew ways and tricks to misuse the card and transparency of online payment. To detect such frauds, comparing the usagepattern and current transaction of a user over the past transactions, then classify it as either fraud or a legitimate transaction.Thus, to overcome these fraud activities we need a powerful fraud detection technique. To detect outliers, different machinelearning algorithms such as logistic regression, Random forest, Naive Bayes, Support Vector Machine, KNN, NeuralNetwork Algorithm are used. However, credit card dataset is imbalanced and the classification model can’t apply directly onthe imbalanced dataset because prediction may incline toward the majority cases so the resulted prediction can be wrong. So,dataset need to be converted into balanced dataset which is done by sampling methods. In this study, classificationalgorithms were applied on balanced and imbalanced dataset after that calculate the accuracy for each algorithm tomeasure the performance of algorithms and then compare the result of different machine learning algorithms to determinewhich algorithm give best result for identifying fraud transactions

Simran Sharma, Dr. N. C. Barwar, “A Comparative Study of Machine Learning Techniques for Fraud Detection in Imbalanced Credit Card Datasets”, International Journal of Advance Engineering and Research Development (IJAERD), Vol. 8, Issue 3, pp. 26-34, March 2021.