Identifying Unauthorized Transactions On Credit Cards By Using Machine Learning Methodologies

Abstract

It is essential for organizations that issue credit cards to be able to recognize fraudulent credit card transactions. This will prevent consumers from being charged for products that they did not buy with their credit card. The purpose of this project is to demonstrate the modelling of a data set via use of machine learning for the detection of credit card fraud. The problem of detecting fraudulent use of credit cards requires modelling previously completed credit card transactions using the information from those that were determined to be fraudulent. After that, this model is put to use to determine whether or not a new transaction constitutes fraudulent activity. Our goal is to appropriately handle misclassified categories by reducing the number of false Negative cases. During this stage of the process, our primary focuses have been on the analysis and preprocessing of data sets, as well as the application of multiple anomaly detection algorithms these algorithms include the local outlier factor and the isolation forest algorithm. We have used IEEE_CIS Fraud dataset, provided by the kaggle .we applied feature extraction technique to reduce the dimensionality of large dataset by extracting only those principle components with highest variance. Given the class imbalance ratio, we measured the accuracy using the Area Under the Precision-Recall Curve (AUPRC) which gives better results than any other previously used models.