What correlates with what — and which genres rewire the rules
The Sonic Genome — Feature Correlation.Top: pairwise Pearson correlations among nine audio features over n = 88,732 usable tracks (2022 Spotify snapshot). Production features dominate: energy–loudness r = +0.76, energy–acousticness r = −0.73, acousticness–loudness r = −0.58; the strongest mood-pair is danceability–valence at r = +0.48. Bottom: the same matrix recomputed inside five super-genre families, shown as the difference Δr from the global matrix (white = the family behaves like the corpus; saturated = it rewires the pair). The structure is not universal: within electronic music, danceability–loudness collapses from +0.25 to −0.07 (Δr = −0.32); within classical, acousticness–instrumentalness strengthens to r = +0.50 (Δr = +0.41) and energy–instrumentalness deepens to −0.58; within hip-hop & r&b, the global instrumentalness–loudness anticorrelation (−0.43) vanishes (Δr = +0.41). Family membership is each track's modal super-genre across its tags. Diverging color is centered on 0 in both panels; n per family is printed in each title. Hover any cell for exact values.
Pearson r on the 9 continuous features (z-scaling does not affect r); quality-flagged tracks excluded (1,009). Family sizes: electronic 16,964 · rock 10,605 · classical 3,608 · hip-hop & r&b 4,661 · latin 8,941. Source: Kaggle maharshipandya/-spotify-tracks-dataset; Spotify deprecated the /audio-features endpoint Nov 2024.