- competition-link
- jupyter notebook
- Slide
Solution: Hierarchical XGBoosting model
To avoid imbalanced data we followed the SVM one-against-all method provide to using binary-classification to predict the answer in hierachial way, and in the third model using the five-class XGBoost to predict the answer - Best score of our model!
- The private score only 0.3 percentage points from the first place!
h30306/Kaggle-Airbnb-New-User-Booking
Kaggle competition. Our proposed using Hierarchical XGBoosting model to predict the target country of customer preference
Jupyter Notebook