Gathering informations about feature engineering methods on Kaggle

[YuanYuan0608]

There are many useful ways to do feature engineering that is done by others on Kaggle. We should imitate them, while their methods can inspire us at the same time.

[xingjian-bai]

The following solutions contain some feature engineering:

• https://www.kaggle.com/code/mpwolke/wids-2023-xgbregressor
• https://www.kaggle.com/code/trushatalati/wids-2023-lgbm
• https://www.kaggle.com/code/ducanger/wids-2023-simple-basline-rmse-1-14
• https://www.kaggle.com/code/shrutisaxena/wids-2023-xgboost-catboost-sc-1-18

[xingjian-bai]

Some feature engineering ideas from me (feel free to criticize them!):

• Separating data by "climate region", then building a model for each region?
• (to be confirmed) the testing data's time all belong to Oct-Nov 2022. Therefore, our model should FOCUS on predicting temperature around Oct-Nov. This might inspire us to extract more information from the training data with the date ~Oct-Nov (e.g., add "the average temp in Oct-Nov" as a new feature, or simply train our model heavily on data within this period).

[YuanYuan0608]

It seems like we haven't try to drop the less correlated features. Should we try that?