This repository is for Data Ming project 'Airbnb' from CI6227, NTU, SG. Created and maintained by Arthur, Peggy, Rashmitha and Meenu. Sept.6th, 2016.

This repository contains:

1. Reference codes.
2. Project Analysis.
3. Our own codes.

In the final implementation, we use Sandro's method (http://blog.kaggle.com/2016/03/07/airbnb-new-user-bookings-winners-interview-3rd-place-sandro-vega-pons/) to generate features and train hierarchical classifier.

How to use it:

1. Use 'WorkSpace/Sandro's Feature' to generate feature,
2. use 'WorkSpace/ModelTraining_DifferentClassifier.py' to train various model in the first layer.
3. Use 'WorkSpace/ModelTraining_SecondLayer.py' and 'WorkSpace/ModelTraining_SecondLayer2.py' to train the second layer.
4. Use 'WorkSpace/Voting1.py' to perform the final prediction.

Due to time, I haven't finished all the code refinement. If you follow the previous steps to use the code, you may encounter some problems(Like get trouble about the data used in Voting1.py). Feel free to contact me at csyhhu@gmail.com if you get interest or trouble using this codes.