We help our US-based client build patented technology that guards your company and workflow and minimizes duration wasted on mechanically revision operations and transactions. It’s easy to set up and it’s already used by more than 80 thousand merchants.
Our team created a non-linear, monitored machine learning model. The pattern identified fraud in real-time. We try to build a model for highly imbalanced data that are regular for this market. We used a mix of Classification methods like Random Forest, Bootstrap, Catboost, and LightGBM and implemented about 80 features.
Tasks we covered
- Cost of operation reduction by 24%
- Customer Satisfaction increase by 36%
- Analyzed around 140K transactions with 7% of fraudulent data items
Tech stack we used