Sample codes and dashboards for 4th place solution for Kaggle's Home Credit Default Risk competition. These are some dashboardas and codes we used through this competition, our approach has been mainly based on the following approach:
- everyone creates his own features and oofs to the end of the competition,
- pooling our features and creating more oofs,
- training different 1-layer stacking models,
- tuning models and blending.
As most of ML enthusiasts know feature engineering is the most important part of every ML competition and most informal one, everyone does it his/her way so instead of talking about feature engineering, here we present some feature engineering dashboards based on many good public kernels and our own ideas that are used to create diverse models.
We created a lot of oofs (150+) by changing parameters and features used in different models and dashboards. some of those models are presented here as examples...