This is a project carried out for the course CE4041, Machine Learning, in Nanyang Technological University (Singapore) for Semester 2 2017/18.

The project consists in predicting if a user makes or does not make (churn) a new service subscription transaction in a streaming music online platform (KKBOX) within 30 days after the membership expiration date for that user.

To do this project we used Jupyter Notebooks under Python 3.6.

1. Data exploration's notebooks are stored at .\Data Exploration directory. There are 3 files for each of the datasets: transactions, user logs and members. notebooks are stored at .\Data Exploration directory.
2. Feature Engineering's notebooks are stored at .\Feature Engineering directory.
3. Notebooks for the training and testing of our model are stored at .\Prediction Model directory.

Installing Python and its libraries is generally easy. The installation process is shown below for Windows and Mac OS.

First of all check that you don't already have Python installed by entering python in a command line window. If it is already installed, it will show Python version.

1. Download and install Python 3.6 with the .exe file from here.
2. Download and install Anaconda (package that contains Jupyter Notebooks) from here.
3. Launch Anaconda from Windows Start Menu.
4. Open Jupyter Notebooks and go to the folder where you have your Jupter Notebook's files.

1. Download and install Python 3.6 with the by typing .pkg fie from here.
2. Download and install Jupyter Notebooks by typing into the terminal: python3 -m pip install --upgrade pip python3 -m pip install jupyter
3. Launch Terminal and type jupyter notebook.
4. Once Jupyter Noteboo is open, browse to the folder where you have your Jupter Notebook's files.

The following libraries are needed: numpy, pandas, time, sklearn, xgboost, matplotlib, mpld3, seaborn and datetime.

Note that some of them are installed together with either Anaconda or Jupyter Notebook.

We achived TOP 8.5% with a Kaggle private result (log loss) of 0.11859.

• Pablo Javier de Paz Carbajo
• Rick Schneider
• Maxime Kayser
• Ragnhild Skirdal Froehaug
• Martha Honganvik Andersen

• Python oficial website.
• Python 3.6 documentation.
• Anaconda documentation.
• Jupyter Notebooks documentation.