Credit Card Fraud Detection Using Hybrid Machine Learning Algorithms

Abstract

Evolution and civilization are two faces of the same coin. Human civilization has evolved from the stone age to the iron age and currently in 2024 we are living in the digital age. Now we have easy access to the internet and electronic devices. Improved cost of living, abundant availability of internet and gadgets has facilitated the use of cashless transactions. Nowadays we can pay using net banking, mobile banking, mobile apps, tap and swipe of cards. It increases the possibility of fraud in financial transactions. Fraud is a phenomenon when we do certain things without the consent of authorized stakeholders. Now a days physical presence of a credit card is not necessary, we can transact if we have details of the credit card. So, this increases the possibility of credit card fraud. To stop credit card fraud, we should have a strong fraud prevention and detection mechanism. Credit card fraud is a loss for both the issuer as well as the user. In this article, we are going to discuss a fraud detection mechanism using Machine learning algorithms. Fraud detection is a binary classification problem where we have to classify the transactions as fraud or non-fraud. In this article, we are using a hybrid ensemble voting classifier for fraud detection. Logistic Regression, Decision Tree and K Nearest Neighbor algorithm are being used to set up the hybrid model. Ensemble voting classifiers consist of several homogeneous weak learners and the prediction capability of the Ensemble hybrid model is better than individual models. Sometimes the hybrid machine learning model out performs legacy systems. When the precision of the model is high it predicts the fraud correctly.so in our work we are building a hybrid model whose precision is 0.97, where the individual models have the precision Logistic regression (LR) 0.94, Decision Tree (DT) 0.85 and K nearest Neighbor (KNN) 0.93.