Identification of Credit Card Fraud Based on Machine Learning and Deep Learning Approaches

Nowadays, digital transactions have become prevalent across the globe which as well raises alarming concerns in the security sector. Unauthorized transactions become frequent in the crowd of digital transactions. Identifying fraudulent credit card transactions is crucial to minimize monetary losses and protect consumers' trust in the financial sector. Therefore, this study focuses on developing an efficient and accurate model to identify fraudulent transactions using machine learning techniques and ensemble techniques (combination of Artificial Neural Network (ANN with base classifiers). By analyzing a dataset containing labeled transactions, this approach applies data preprocessing (standardization technique), and algorithmic optimization to differentiate between legitimate and fraudulent activities. We have done a comparative analysis among all the classifiers for fraud detection. To validate the study result, a cross-validation technique has been applied. Among all the classifiers, Random Forest classifiers show the highest performance level by achieving 95% precision, 90% recall 92% f1 score, and 99% accuracy. In the comparison of base classifiers and their ensemble techniques, we found ANN with Gaussian Naïve Bayes (GNB) performs better than usual GNB.