KKBox's Music Recommendation System

Project Structure

project/
│
├── src/                    # Source files
│   ├── models/             # each one's folder
│
├── data/                   # CSV files
│
├── notebooks/              # Jupyter notebooks
|
├── tensorboard/            # TensorBoard
|
├── checkpoints/            # Checkpoints
│
├── .gitignore              # Specifies intentionally untracked files to ignore
│
├── README.md               # Project overview and setup instructions
│
└── requirements.txt        # The dependencies file

Data

• Put the csv files under the "data" folder.
• In the training data, we use the first 80% as the training set, and last 20% as the validation set.

Setup Env

pip install -r requirements.txt

Train

Uncomment or add train_pipeline in train.py, then

cd ./src
python train.py

Models will be saved in the "checkpoints" folder.

Test

cd ./src
python bagging.py

An "output.csv" will be generated under the "data" folder.