Implementasi Klaster K-Means untuk Analisis Tren Musik Spotify dengan Menggunakan Metode LPP dan UMAP

Deswinda, Nafisa Sakha Ramadhani (2025) Implementasi Klaster K-Means untuk Analisis Tren Musik Spotify dengan Menggunakan Metode LPP dan UMAP. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002211068-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB)

Abstract

Tren musik mengalami evolusi seiring waktu, dipengaruhi oleh preferensi pendengar, inovasi industri musik, dan perkembangan teknologi streaming seperti Spotify. Analisis tren musik menjadi penting untuk memahami perubahan pola preferensi dalam jangka panjang. Penelitian ini bertujuan untuk menganalisis tren musik dari tahun 1990 hingga 2023 melalui penerapan algoritma klasterisasi K-Means, serta membandingkan performa dua metode dimensionality reduction, yaitu Locality Preserving Projection (LPP) dan Uniform Manifold Approximation and Projection (UMAP). Dataset yang digunakan mencakup sembilan fitur audio, yaitu acousticness, danceability, energy, instrumentalness, liveness, loudness, speechiness, tempo, dan valence. Hasil klasterisasi diverifikasi menggunakan ground truth berupa genre dari Billboard Year-End Hot 100 Singles. Hasil evaluasi menunjukkan bahwa setiap kombinasi metode memiliki keunggulan yang berbeda. Metode UMAP 2D terbukti paling konsisten dalam merepresentasikan genre musik dari waktu ke waktu berdasarkan ground truth, sekaligus paling andal secara topologis, dengan nilai trustworthiness tertinggi sebesar 0,9663. Sementara itu, LPP 3D unggul dalam merefleksikan transisi era musik, ditunjukkan oleh nilai Adjusted Rand Index (ARI) tertinggi sebesar 0,6519. UMAP 3D menunjukkan keseimbangan terbaik antara evaluasi internal dan eksternal, ditandai dengan nilai Normalized Mutual Information (NMI) yang tinggi sebesar 0,7343. Adapun LPP 2D memiliki struktur klaster paling stabil dari segi pemisahan dan kekompakan, tercermin dari nilai silhouette score sebesar 0,6502, Calinski-Harabasz Index sebesar 86,7517, dan Davies-Bouldin Index terendah sebesar 0,4510. Dari segi tren, klaster yang terbentuk mengindikasikan kemunculan subgenre digital yang memiliki karakteristik sonik khas namun tidak terwakili secara eksplisit dalam data Billboard. Penelitian ini memberikan wawasan terhadap evolusi karakteristik audio musik, dan berpotensi menjadi referensi bagi musisi, produser, serta platform streaming dalam memahami preferensi dan arah tren musik di masa depan.
==================================================================================================================================
Music trends evolve over time, influenced by listener preferences, music industry innovations and the development of streaming technologies such as Spotify. Analyzing music trends is important to understand changes in preference patterns over the long term. This study aims to analyze music trends from 1990 to 2023 through the application of the K-Means clustering algorithm, and compare the performance of two dimensionality reduction methods, namely Locality Preserving Projection (LPP) and Uniform Manifold Approximation and Projection (UMAP). The dataset used includes nine audio features, namely acousticness, danceability, energy, instrumentalness, liveness, loudness, speechiness, tempo, and valence. The clustering results are verified using ground truth in the form of genres from Billboard Year-End Hot 100 Singles. The evaluation results show that each combination of methods has different advantages. The UMAP 2D method has been proven to be the most consistent in representing music genres over time based on ground truth, while also being the most reliable topologically, with the highest trustworthiness value of 0.9663. Meanwhile, LPP 3D excels in reflecting musical era transitions, as evidenced by the highest Adjusted Rand Index (ARI) value of 0.6519. UMAP 3D demonstrates the best balance between internal and external evaluation, marked by a high Normalized Mutual Information (NMI) value of 0.7343. LPP 2D has the most stable cluster structure in terms of separation and compactness, reflected in a silhouette score of 0.6502, a Calinski-Harabasz Index of 86.7517, and the lowest Davies-Bouldin Index of 0.4510. In terms of trends, the clusters formed indicate the emergence of a digital subgenre with distinctive sonic characteristics that are not explicitly represented in Billboard data. This study provides insights into the evolution of music audio characteristics and has the potential to serve as a reference for musicians, producers, and streaming platforms in understanding future music preferences and trends.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	K-Means Clustering, Locality Preserving Projections (LPP), Spotify, Tren Musik, Uniform Manifold Approximation and Projection (UMAP), K-Means Clustering, Locality Preserving Projections (LPP), Music Trends, Spotify, Uniform Manifold Approximation and Projection (UMAP)
Subjects:	M Music and Books on Music > M Music Q Science > QA Mathematics Q Science > QA Mathematics > QA278.55 Cluster analysis Q Science > QA Mathematics > QA76.6 Computer programming. Q Science > QA Mathematics > QA76.9.D33 Data compression (Computer science) Q Science > QA Mathematics > QA9.58 Algorithms
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Nafisa Sakha Ramadhani Deswinda
Date Deposited:	01 Aug 2025 02:34
Last Modified:	01 Aug 2025 02:34
URI:	http://repository.its.ac.id/id/eprint/125426

Actions (login required)

View Item