Enhancing Credit Card Fraud Detection: Analyzing Time and Amount Distributions with Computational Intelligence Algorithms

Credit card fraud presents a significant risk to both financial institutions and cardholders, necessitating the implementation of robust detection measures. This study focused on evaluating the distributions of time and transaction amounts to enhance the effectiveness of fraud detection. By analyzing distribution patterns, activity anomalies, and trends, it becomes possible to identify fraudulent transactions. Specifically, the amount distribution provides valuable insights into the monetary dimension of fraud, aiding in the identification of suspicious activities. Logistic regression (LR), support vector machines (SVM), and k-nearest neighbors (kNN) were used for the model analysis and evaluation. The results demonstrate that SVM outperforms both LR and kNN, exhibiting high precision and recall rates, as well as an improved F1-measure. In particular, the precision achieved for the no fraud and fraud categories is 0.92 and 0.96, respectively. The corresponding recall rates are 0.96 for the no fraud category and 0.9 for the fraud category. The F1-scores achieved are 0.91 for the no fraud category and 0.93 for the fraud category. Moreover, the average accuracy attained is 0.95. These findings also emphasize the significance of analyzing time and amount distributions and employing suitable model for the effective detection.