Advanced Machine Learning Models for Detecting and Classifying Financial Fraud in Big Data-Driven

Abstract

The banking sector faces a major challenge in identifying credit card fraud, especially as online transactions increase. This study employs the Kaggle Credit Card Fraud Detection dataset to present a machine learning (ML)-based method to credit card fraud detection. The collection contains de-identified transaction information from European cardholders. With 284,807 transactions, only 492 included fraud, suggesting a significant disparity in class.Therefore, data balancing techniques were used to improve model training. Among the data pretreatment procedures were label encoding for categorical conversion and standardization to normalize feature scales. Using Euclidean distance, for the purpose of identifying the majority-belonging k-nearest neighbor class, A classifier called K-Nearest Neighbors (KNN) was created. The model was assessed using ROC-AUC, F1-score, accuracy, precision, recall,and K-fold cross-validation, among other important performance measures. The KNN model outperformed benchmark models like MLP and Naïve Bayes,obtaining 98.56% accuracy and an AUC of 96.07, according to experimental data, demonstrating great classification efficacy. The promise of KNN in creating reliable and accurate fraud detection systems for cybersecurity applications is confirmed by these findings.