Discover with me patterns in audio features of spotify songs and follow me on my way searching for a good model, to predict a songs popularity by using audio features created by Spotify for analysis.
My current results of data analysis, using : https://eddabra.github.io/popular-spotify-tracks/
About the data set: The dataset is collated from Spotify's API using two separate python scripts to extract popular and non-popular songs and their associated audio and descriptive features.
How much songs belongs to a certain genre?
In which year and which months have the songs been released?
More than a half of the songs ave been released in the last 6 years. Most songs have been been released in october. The beginning of autumn seems to be a good time for spreading music.
How are danceability and energy are connected?
Intuitively, I would say there is a connection between the energy of a song and its danceability. Could we find evidence for that looking at the following figures?
Most of the songs in the data set have a medium to high energy and danceability! Also, quite silimar distribution, could be a hint, that there is connection.
Which role plays the tempo?
Do the features have a relationsship to each other?
The distribution of popularity
Is very happy, sad or angry music more popular?
When predicition of a linear model fail and we can't see a linear line
Making the model more flexible identifying appropriate splines
