Enhancing Credit Card Fraud Detection with K-Nearest Neighbours (KNN): A Machine Learning Approach

The rise in digital transactions has made credit card fraud an even more serious worldwide issue. The study examines the use of a highly biased dataset of transactions from European cardholders to identify credit card fraud using the K-Nearest Neighbours (KNN) method. The dataset was improved by utilising Principal Component Analysis (PCA) to improve feature relevance. It consisted of 284,807 transactions with only 492 fraudulent cases. After training and evaluation, the KNN model showed a good accuracy of 94.04%. However, a considerable number of false positives and undetected frauds are indicated by the model's low accuracy (0.0136) and moderate recall (0.5074). The study highlights the importance of effective data preprocessing, feature selection, and parameter optimization in improving model performance. This study improves fraud detection rates in real time using KNN, providing information for future research on advanced machine learning methods and ensemble techniques.