This dataset contains 26 million fixed rate mortgages between Jan 1999 to Sept 2017 which can help us build more accurate credit performance models. It is publically available at http://www.freddiemac.com/research/datasets/sf_loanlevel_dataset.html
We had to develop a pipeline for this project which involved scraping the data from the website for the given time frame, this script would also carry out exploratory analysis and finally give us results.
We applied the best algorithm from the pipeline to alternative scenarios to verify their validity.
Highlights:
We employed AutoML technologies like H2O and TPOT to land with results that were validated by our manual investigations We used feature selection strategies like Forward, Backward and Recursive Feature selection