According to our model, what is the profile or the typical defaulter? How does this compare to the actual results?

 
Screen Shot 2016-12-15 at 02.38.31.png
Screen Shot 2016-12-15 at 02.43.56.png
 

interest rate

The average interest rate for defaulting loans is 16.78% for our model, while in reality, the average interest rate for defaulters is 16.74%, which is almost exactly the same.

Number of defaulters

Our model predicts 5407 defaulters when in reality there are 1664 defaulters.

income

The average annual income for defaulters is $74,263 for our model, while in reality, the average annual income for defaulters is $75,234 - only a thousand dollar difference! This is notable since annual_inc, unlike interest rate, is not one of the top 10 important features in our model.

Employment

The most common length of employment is 10+ years. It makes up 28% of the true defaults and 29% of predicted defaults. Perhaps a higher degree of resolution would show that longer employment length decreases likelihood of default, but in this dataset, everything beyond 10 is grouped into one category.