Fraud detecection based on 3-stage filtering:
- Benford's Law credit and debit transaction analysis
- Psychological profiling based on availble researches
- Trained Machine Learning Logistic Regression model. Model was trained on our dataset but eventualy should be trained on clients dataset.
Application created in sake of increasing demand on AML analysis without proportional increase of AML specialists. Application can be adjusted depending on business needs and constraints of the client. Main goal of The Fraud detector is to prioritize potential fraudsters for further investigation by qualified employees. Technologies in use: Python 100% (Pandas, Sci-kit learn, Numpy).