The Challenge
A leading retail bank needed to reduce false positives in its fraud detection system, which were leading to high operational costs and poor customer experience.
The Solution
Neotheta developed and deployed a real-time transaction scoring engine using a Gradient Boosting Machine (GBM) model. The model was trained on a massive dataset of historical transactions and was able to identify subtle patterns indicative of fraud with high precision.
The Results
- 40% reduction in false positive fraud alerts
- 95% precision in detecting fraudulent transactions
- Improved customer satisfaction by minimizing legitimate transaction declines
Technologies Used
AWS SageMaker, Python, XGBoost, Kinesis, Lambda
