The prosper dataset provided initially contained 113,937 loans with 81 variables. After performing cleaning operations, i was left with 82,708 unique loan listings and 15 variables that characterize each given loan.
The dataset used can be downloaded from here
The feature documentation of the dataset can be found at here
At the onset of my analysis, i encountered the presence of duplicate values within my dataset which i dropped (This was made possible through the listing key).
Also, i decided to pick the prosper rating as one of my major features to inspect; This led to my dropping of every loan listing without a value for prosper rating.
I discovered a high positive correlation between the borrower APR and borrower rate, as well as between prosper rating and prosper score.
In my investigation on how some variables (employment status, occupation, income range and available bank card credit) impacted the borrower APR; I found (surprisingly) a negative correlation existing between each of these variables and the borrower APR. This suggests that borrowers who earn less are most likely going to have a higher borrower APR and rate.
I also found that the prosper rating classified loans with high borrower APR as high risk loans. To check that prosper rating classifier was accurate enough, i explored the relationship between prosper Loan status and borrower APR. I discovered that lower borrower APR loans which prosper rating associated as low risk loans are the major constituents of the 'completed' and 'final payment in progress' loan status classes; while the charged off, defaulted and delinquent loan classes have relatively high borrower APR.
here is the link to my analysis