BPI-Challenge-Rabo

Mining of Freo's loan application process 📊 💱

Scripts/notebooks used in a research project for Freo (Rabobank/DLL subdivision) in which event logs related to their loan application process are analysed. The project involved, among other things, the following explorations/examinations/sub-tasks:

  1. Devising a clear overview of their loan application processes.
  2. Building predictive models using TensorFlow and RandomForest. The goal is to predict whether an applicant will accept a loan offer or not.
  3. Assessing whether the 'frequency of incompleteness' (i.e. requesting documents multiple times) has an effect on the final outcome. In other words: evaluating whether a more streamlined/less redundant document verification process would be beneficial to their loan application conversion rates.
  4. Discover/identify group of customers (i.e. applicants) and/or patterns in 'application subsets' such as immediately rejected applications and applications with declined/accepted/cancelled offers.
  5. Discover/identify group of customers and/or patterns in high throughput time applications.