How did popularity of the genres change over time?

Simplify your music!

Spotify's multitude of genres Clustered into four distinct groups

A dendrogram represents the hierarchical structure of objects. In our case, we have represented the songs in such a dendrogram to show the similarities of them. The analysis provides at the same time a basis and a control for further clustering of the songs into genres. The basis for the dendrogram is again the audio features. Based on the dendrogram, we see that four clusters emerge, which can be separated from each other. This number was also used as the basis for the clustering algorithms.

We want our own genres!

Spotify offers an enormously high number of different genres. For the songs used in our analysis alone, there were over 1600 different genres. Spotify differentiates very meticulously into genres like "pittsburgh rap", "pop rap", "rap", "southern hip hop", for example. Therefore, we decided to create our own classes based on the features. The clustering was done over the normalized feature vectors. For the visualization a PCA was performed afterwards, to reduce the feature vectors to three dimensions for plotting purposes. The four resulting cluster were named after the genre, which occurred most frequently in the respective cluster. Using these four classes, we can compare the change over time much better than with 1600+ different genres.