A Human-AI Collaborative Approach for Credit Card Fraud Detection: Integration of LSTM Networks with Interactive Web Interface

Abstract

The increasing complexity of financial transactions has made fraud detection a critical challenge for financial institutions. Traditional rule-based systems are often unable to adapt to dynamic fraud patterns and require more advanced solutions. This paper proposes a Human-Artificial Intelligence Collaboration framework for credit card fraud detection that provides explainable predictions by integrating a pre-trained LSTM (Long Short Term Memory) model with the SHAP (SHapley Additive exPlanations) framework. The web application evaluates each transaction in real-time, classifies it as fraud or normal, and provides interpretable insights into the decision-making process using SHAP values. Users can provide feedback on the model's predictions, enabling continuous learning and improvement of the system. Experimental results show that the integration of user feedback significantly improves the performance of the model over time and leads to an increase in F1-score after iterative retraining. This work highlights the importance of explainability in AI-assisted fraud detection systems and the value of human feedback in improving machine learning models. The proposed framework not only improves detection accuracy, but also increases user trust through transparent decision-making processes.