Spotify Skip Prediction

APS360 Final Project

The challenge: predict if users will skip or listen to the music they're served on the Spotify music streaming platform. A solution to the Spotify Sequential Skip Prediction Challenge 2018 on AIcrowd. A supervised sequential classification task.

Recommender systems play a crucial role in the digital age where media is abundant, and serving the right media to the right user at the right time means everything in terms of revenue generation. It is in the best interest of digital media providers the likes of Spotify, Youtube, Facebook, Google, and others to maximize audience retention. A model which is able to accurately predict user intent given significant nuance is necessary; machine learning is consequently the right tool for the task.

System Architecture

We employ an autoencoder model for feature extraction and a vanilla RNN model for inference on the autoencoder embeddings. Our results are evaluated against a gradient boosting model for classification.

Contribution

Setup

This section will take you through the procedure to configure your development environment. At a glance:

1. Install project's python version
2. Install Git
3. Install poetry
4. Clone repository
5. Run poetry install
6. Configure IDE virtual environment
7. Install pre-commit hooks

Begin by installing the project's python version. See the badges at the top of the README for the version number.

If not already installed, install git.

The repo employs poetry as its dependency and environment manager. Poetry can be installed through the Windows Powershell via:

(Invoke-WebRequest -Uri https://raw.githubusercontent.com/python-poetry/poetry/master/install-poetry.py -UseBasicParsing).Content | python -

Clone the repo using Github Desktop or the commandline via:

git clone https://github.com/DM1122/spotify-skip-prediction.git

From within the cloned repo, run poetry's install command to install all the dependencies in one go:

poetry install

Configure your IDE to use the virtual environment poetry has created at C:\Users\<USERNAME>\AppData\Local\pypoetry\Cache\virtualenvs. In the case of VSCode , enter the command pallet by going to View>Command Palette and search for Python:Select Interpreter. Select the appropriate poetry virtual environment for the repo.

Install the pre-commit script and hooks using:

pre-commit install --install-hooks

You're now ready to start contributing!

Adding Packages

To add a new package to the poetry virtual environment, install it via:

poetry add <package>

This is poetry's version of pip install <package>.

Testing

This repo uses pytest for unit testing. To run all unit tests, call:

pytest -v

You can find an interactive report of test results in ./logs/pytest-report.html. Individual tests can also be specified as follows:

pytest tests/test_<filename>.py::<function name>

Groups of tests can be run using markers. Assign a marker decorator to the group of functions you want to test like this:

@pytest.mark.foo
def my_test_function():
    # some test

To use the custom marker foo, it must be added to the list of custom pytest markers in pyproject.toml>[tool.pytest.ini_options]>markers. The tests marked with foo can then be run by calling:

pytest -v -m foo

Or to avoid all tests with a particular marker, call:

pytest -v -m "not foo"

Commits

Pre-Commit

This repo is configured to use pre-commit hooks. The pre-commit pipeline is as follows:

1. Isort: Sorts imports, so you don't have to.
2. Black: The uncompromising code autoformatter.
3. Pylint: It's not just a linter that annoys you!

Pre-commit will run the hooks on commit, but when a hook fails, they can be run manually to debug using:

isort . ; black . ; pylint_runner