This project predicts hotel booking cancellations using the Hotel Bookings dataset. The goal is to show that feature engineering and preprocessing improve model performance.
raw.githubusercontent.com/swapnilsaurav/Dataset/refs/heads/master/hotel_bookings.csv
Hotel Bookings dataset - 119,380 rows, 33 columns Target: is_canceled (binary classification)
| Version | ROC-AUC | F1 |
|---|---|---|
| Baseline | 0.9051 | 0.7257 |
| After Preprocessing | 0.9558 | 0.8447 |
| After Feature Engineering | 0.9809 | 0.9066 |
| After Selection (Top 20) | 0.9739 | 0.8855 |
- Open FeatureEngineering_Capstone.ipynb in Google Colab
- Runtime > Restart and Run All
- No local setup needed, dataset loads from URL automatically
See requirements.txt
- FeatureEngineering_Capstone.ipynb - main notebook
- src/helpers.py - reusable feature construction functions
- report/ - contains Report PDF
- requirements.txt - dependencies