Skip to content

Developing and exploring different statistic and ML techniques in order to understand the behaviour and predict the values of Spotify's Valence metric.

Notifications You must be signed in to change notification settings

pkaliosis/ML-Spotify-Valence-Prediction

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 

Repository files navigation

Spotify-Valence-Prediction

Spotify uses a metric called valence to measure the happiness of a track. The metric itself, however, was not developed by Spotify. It was originally developed by Echo Nest, a company that was bought by Spotify in 2014. We don't know exactly how valence is calculated. Some details are given by a blog post, which you can find here:

https://web.archive.org/web/20170422195736/https://blog.echonest.com/post/66097438564/plotting-musics-emotional-valence-1950-2013

The project's task is to untangle the mystery behind valence and propose how this is derived.

Spotify offers the following information that have been releveant to the task:

-Get Track's Audio Features and Get Tracks' Audio Features.

-Get Track's Audio Analysis.

There are two main goals on this project.

Goal #1: To explore which features influence the Valence: By using inferential statistic methods to study how features (track and possibly audio) influence valence. I am trying to find the best possible model for explaining the valence based on the features that you find significant.

Goal #2: Predict Valence: By using Machine Learning techniques to predict valence based on track features. There are both neural network and non-neural network techniques used on this project. This happened of course in order to explore ore techniques and expand my skillset, but at the same time to understand the power of the NNs.

This project was developed under AUEB's Applied Machine Learning course during the Winter Semester 2021-2022.

About

Developing and exploring different statistic and ML techniques in order to understand the behaviour and predict the values of Spotify's Valence metric.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published