- Loan_Amount_Requested was converted to numeric by removing commas
- Catboost requires minimum pre-processing while lightgbm requires some more pre-processing
- All NaN's were converted to "NaN"
- Created 2 new features.
- For lightgbm all string features were label encoded
- Catboost was somewhat tuned and the first set of prediction probabilities were generated
- LightGBM was somewhat tuned and the second set of prediction probabilities were generated
- Finally, a weighted average of the probabilities from both classifiers are used to generate the final predictions
- Python for programming
- sklearn and numpy libraries for methodology
- lightgbm and catboost library for the final model
- matplotlib and seaborn was used for plotting and analyzing the data
Rank: 2nd on public LB and 4th on private LB
Link to LeaderBoard!