This AI-Powered Fraud Detection System is a full-stack solution that bridges the gap between complex machine learning and human-readable analysis. The application utilizes the Isolation Forest algorithm to identify transaction anomalies by measuring how easily data points can be "isolated" based on their amount and timing.
To make these flags actionable, the system integrates Gemini 2.5 Flash to generate professional, natural language explanations for each suspicious case. A key technical highlight is the implementation of Batch Processing, which bundles multiple transaction rows into a single API request to ensure high performance and stability within the Gemini Free Tier's rate limits.
Built with Streamlit for an interactive user interface and Firebase Firestore for persistent data storage, the project demonstrates a resilient, production-oriented approach to AI engineering.
Built With
- gemini
- pandas
- plotly
- python
- python-dotenv
- scikit-learn
- streamlit
Log in or sign up for Devpost to join the conversation.