Paper Contents
Abstract
In the contemporary digital economy, credit card fraud has increased in tandem with the sharp rise in online financial transactions, presenting a serious risk to both financial institutions and individuals. The extremely unbalanced nature of transaction data, where legitimate cases greatly outnumber fraudulent ones, makes it difficult to detect fraudulent transactions effectively and accurately. This study offers a machine learning-based framework for detecting credit card fraud that incorporates several supervised learning algorithms, such as XGBoost, Decision Tree, Random Forest, and Logistic Regression.The study's dataset, which included more than 280,000 anonymized transactions, came from Kaggle's Credit Card Fraud Detection Dataset 14. In-depth data pre-processing methods were used, such as standard scaling for normalization, log transformation to address transaction amount skewness, and missing value analysis. The SMOTE (Synthetic Minority Oversampling Technique) algorithm was used to address class imbalance, allowing the creation of synthetic minority (fraudulent) samples and enhancing the sensitivity of the model.A number of performance metrics, including accuracy, precision, recall, F1-score, and ROC-AUC, were used to assess the models. The results of the experiment showed that the XG Boost classifier performed the best and most balanced, with high precision and recall. This made it especially useful for identifying fraudulent activity while reducing false alarms. This study emphasizes how crucial ensemble-based models and data pre-processing are to creating dependable and scalable credit card fraud detection systems for practical financial applications.
Copyright
Copyright © 2025 Rajnish Sharma, Sagar Chaudhary, Nitin Kumar, Megha. This is an open access article distributed under the Creative Commons Attribution License.
