Abstract

Harmonic progressions are central to the popular music listening experience, yet corpus-based, data-driven classification of song-level harmonic progression patterns remains underexplored. This study analyzes 178 songs from the Billboard Hot 100 year-end charts for 2023–2025 using chord charts collected from Ultimate Guitar. The chord charts were converted into key-neutral chord-root sequences through Roman numeral analysis, and six indices were developed to capture harmonic diversity, repetition, concentration, and entropy-based information content. Using these standardized indices, K-means clustering with silhouette analysis identified a four-cluster solution with interpretable harmonic types and descriptively distinct mean profiles. A Genre × Type chi-square analysis revealed no significant association between broad genre categories and harmonic type. Chord transitions within each type were then modeled as directed networks, and high-frequency transition patterns were interpreted within a music-theoretical framework. The resulting networks exhibited type-specific structures consistent with distinct harmonic grammars. By integrating quantitative clustering with a network-based interpretation of complete songs, this study presents an empirical methodology that translates data-driven harmonic patterns into musically meaningful insights into contemporary popular music harmony.