The challange chosen by the team is the one proposed by Credit Suisse. Data used was provided by Credit Suisse and some additional sources from Basel Institue of Governance. We built a predictive model running on Jupyter notebook, written in Python (Scikit-learn, Pandas, Numpy Seaborn, Matplotlib).
Nowadays, today is 70% of credit fraud/money laundring commited internally. However, with new technologies this attacks are starting to transfer to external enviroment. Therefore it is needed to develop automated techniques to detect suspicious activity in order to take load of data analysts shoudlers.
In this problem we use Machine Learning techniques. In JupyterNotebook you can found our prototype of predictive model. We use Random Forest to classify suspicious customers based on various features on aggregated data over one year.
More models need to be explore with different hyperparameters, another source of data, implement Dashboard to help Data Analyst to explore output from classifiers in a simple, user friendly effective way.