Credit Card Fraud Data Analysis and Prediction Using Machine Learning Algorithms
K. V., S. Thillai Ganesh, D. Muralidharan, G.R. Brindha, Muthu Thiruvengadam · Security and Privacy · 2025
ABSTRACT Credit card fraud detection has become increasingly important due to the surge in digital transactions occurring every minute. Banks and credit card companies require a robust system to alert users about potential misuse of cards at PoS terminals and on online platforms. Credit card fraud is defined as an unauthorized transaction made using a credit or debit card, resulting in financial loss. The identification of fraud is based on demographics and usage patterns, and exploratory data analysis such as identifying duplicates and outliers, feature encoding and scaling, dataset balancing, and plotting techniques was conducted to gain insights prior to modeling. In addition to traditional machine learning methods such as Logistic Regression, kNN, and SVM, this study investigates a novel approach that replaces the conventional kNN distance metric with probability values derived from logistic regression. The objective of the study is twofold: (1) to verify how well the proposed probability‐based kNN method works for imbalanced datasets by comparing it with oversampling techniques such as ADASYN and SMOTE and (2) to propose a more efficient alternative to computationally intensive methods like XGBoost by introducing the probability‐based kNN, which aims to enhance classification performance without significantly increasing computational costs. Cross‐validation was used to estimate model performance and minimize overfitting, ensuring that the proposed method and other models were evaluated comprehensively.