This project analyzes and predicts customer churn of a music streaming service using Spark on a large dataset.
I took the starter code for this repository from a Udacity assignment project and modified it to the present form, which deviates significantly from the original form; see starter.
The project focuses on an imaginary music streaming service, similar to Spotify, where users can listen to streamed music. In that service:
- We have: (1) free-tier users and (2) premium users who pay a subscription.
- Every time an user is involved in an event, it is logged with a timestamp; example events:
songplay,logout,like,ad_heard,downgrade, etc.
The goal is to predict customer churn, either (1) as a downgrade from the premium to free plan or (2) in form of a user leaving the service. With churn predictions, the company can target those users with incentives, such as discounts, etc.
🚧 On-going work.
Contents:
- A
- B
- ...
🚧 TBD.
🚧 TBD.
The directory of the project consists of the following files:
.
├── Instructions.md
...
If you already have a Python environment with the usual ML libraries and you'd like to add PySpark:
# Install PySpark manually
python -m pip install pyspark
python -m pip install findsparkAlternatively, if you want to create a new Python environment (recommended), you can do it with conda:
# Create an environment
conda create -n sparkify python=3.9 pip
conda activate sparkify
# Install pip-tools
python -m pip install -U pip-tools
# Generate pinned requirements.txt
# PySpark is listed there
pip-compile requirements.in
# Install pinned requirements, as always
python -m pip install -r requirements.txt
# If required, add new dependencies to requirements.in and sync
# i.e., update environment
pip-compile requirements.in
pip-sync requirements.txt
python -m pip install -r requirements.txt
# To track any changes and versions you have
conda env export > conda.yaml
pip list --format=freeze > requirements.txt
# To delete the conda environment, if required
conda remove --name sparkify --all🚧 TBD.
🚧 TBD.
🚧 TBD.
🚧 TBD.
🚧 TBD.
Mikel Sagardia, 2023.
No guarantees.
If you find this repository useful, you're free to use it, but please link back to the original source.