This is a project where I analyzed 100,000 bank transactions to detect fraud. I used logistic regression and gradient boosting to recognize fraudulent patterns and flag high risk transactions. For the logistic and gradient models, I created bins for fraud likleyhood ranges and did analysis based on the bins. I did so while keeping false positive rates low and prevented $53 million in my hypothetical dataset.
- R
- Data cleaning and transformation
- Logistic Regression
- Gradient Boosting
- Rule-based filtering
- Data visualization
- Fraud detection modeling