الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 113 |  |  |
| | | from | 113 | | |
Abstract
Credit card transaction fraud results in massive losses for consumers and banks in recent years. Subsequently, both cardholders and banks need a strong fraud detection system (FDS) to reduce cardholder losses. Credit Card Fraud Detection (CCFD) serves as an essential way of preventing fraud. However, many barriers exist to establishing an optimum fraud detection system for banks. First, due to data security and privacy concerns, various banks and other financial institutions are not permitted to exchange their transaction datasets. These issues make traditional systems find it difficult to learn and detect fraud depictions. Therefore, this thesis proposes federated learning for CCFD over different frameworks (TensorFlow federated, PyTorch).
Furthermore, optimized federated models in which models are updated many times before connecting with the server. This dramatically reduces the communication required to train a model on a federated learning platform. Second, there is a significant imbalance in credit card transactions across all banks, with a small percentage of fraudulent transactions outweighing most valid ones. The dataset must be balanced to demonstrate the urgent need for a comprehensive investigation of class imbalance management techniques to develop a powerful model to identify fraudulent transactions.
In order to address the issue of class imbalance, this study also seeks to give a comparative analysis of several individual and hybrid resampling techniques. In several experimental studies, the effectiveness of various resampling techniques in combination with classification approaches is compared. This research finds that the hybrid resampling methods perform well for machine learning classification models compared to deep learning classification models.
The proposed optimization strategy is implemented over four phases: Firstly, credit card data is highly skewed, leading to inefficient fraudulent transaction prediction. To achieve a better result, imbalanced or skewed data is preprocessed with the re-sampling technique for better results. Secondly, a federated learning model has been built over multiple frameworks to preserve the data security and privacy challenges. Thirdly, the proposed optimization phase focuses on decreasing communication costs when proceeding with federal training to accelerate convergence speed by optimizing the initial global model before the federated learning phase. Finally, Measuring the general performance of optimized the federated learning model and comparing these results with the previous works.
This thesis uses the seven-metaheuristic optimization algorithms, namely Particle Swarm Optimization (PSO), Cat Swarm Optimization (CSO), Ant Colony Optimization (ACO), Harris Hawks Optimization (HHO), Heap Optimization (HBO), Fire Hawk Optimizer (FHO), Flying Fox Optimization Algorithm (FFO). The performance of these algorithms is recorded with their comparative analysis. The work is implemented in Python, and the performance of the algorithms is measured based on accuracy, precision, recall, F-measure, loss, and computation time.
The experimental results show that the HBO algorithm with the Federated Learning (FL) Model can achieve high detection performance and the minimum loss ratio across different datasets. The proposed approach is compared with the six of the previous works for more reliability.