Next Hit - Data Science Project
Overview
Timeline
February 2018 - April 2018
Role
Data Scientist, Lead
Team
6 people
Tools
Python
Jupyter Notebook
Introduction
Data Analysis of the different attributes that make a song popular using data from Spotify song charts from 2004 to 2018. Songs are ranked by total number of streams, which also implies that popularity of a song can be seen through total number of streams it has. Anaylsis and data visualizations are shown in NextHit.ipynb. More details on project could be found in ProjectProposal.ipynb (Final project results may differ from initial ideas in proposal, but the proposal carries the inital ideas)Data has been scraped from website and saved to a csv file to avoid scraping every time the notebook is run. But if needed can re-scrape data using the cells under the Data Cleaning section.
Major music streaming services, such as Spotify and Pandora, find music for consumers that cater to their individual tastes. For example, Spotify can create radios for specific artists, albums, or tracks to help users find new music based on their listening history. How does Spotify figure this out? Such companies do this by using data to predict the type of music that listeners would enjoy. With this data, Spotify can predict the type of music that best suits the user. Gathering data based on user preferences is also useful for determining which features of a song make a track popular. Using already known data, one could pinpoint what exactly makes pop music, for example, so popular because each track comes with information on the key, tempo, energy, etc. The fact that this information is available is lucrative because artists and musicians could find out what kind of sounds could elicit a favorable response. This would help people in the music industry discover current music trends.
Background
Our research question involves the use of music’s audio features, so we will need to understand them first. They are listed below:
Pitch class notation: A method to classify the 12 pitches from integers 0 to 11.
Table taken from: https://developer.spotify.com/web-api/get-audio-features/
Research Question
Is there going to be a correlation between certain audio attributes and position on the top charts? If there is a correlation, which features of a song make it more popular? Such features include danceability, energy, key, loudness, mode, speechiness, acousticness, etc. as shown in the above chart.
Hypothesis
We predict that songs that make the top charts will have certain features in common over others. More specifically, we expect tracks with higher valence, energy, and danceability to be more likely to hold a higher position on the charts.
