In this presentation, I explain the steps for data cleaning and modelling for an use case related to Banking domain.

The target is to develop a model to predict which customer is going to ask for a loan. Thanks to the unbalancement of the classes (people with a loan are just the 2% of the training set), this case represented a very interesting opportunity to practice with undersampling and oversampling.

Since, I have to anonymize it, I cannot publish the related iPython notebook and had to delete all the Business Intelligence answers: get the PDF