This is a flask micro website for real time LendingClub loan requests risk prediction with a xgboost model trained on LendingClub 2015 data.
The risk predicted is ranging from 0 to 1. The larger the risk is, the high probability the loan will be charged off.
This model only gives prediction for 36 months terms loans.
For study purpose only.
Below is a ROC figure. The ROC is obtained from the training with "1" label for "Charged Off" and "0" for "Fully Paid". The test AUC of ROC reaches 0.71
The figure below shows the relative importance of the most important 20 features. The importance score is the xgboost model training fscore divided by total fscore for all features.
python 3.5 and above
xgboost
plotly
pandas
numpy
flask
Add your own api key to /review/views.py line 13 "api_key="
Nevigate to the folder and type python runsever.py in terminal