Machine learning Based Clustering Using Spotify Audio Features
Surabhi Shinde, Sarang Kulkarni, Prashant Kulkarni, Abhishek Y. Bhatt · 2023
In this paper, we study the association between different clustering algorithms. Clustering is a crucial technique for data mining applications and research. Spotify audio features were analysed using machine learning techniques that separate the provided data into various clusters according to how close or how far apart they are from one another. Without any prior understanding of the data being considered, clustering analysis offers a helpful technique to group items. This study uses k-means, fuzzy c-means (FCM), and Possibilistic C- means (PCM) algorithms to classify datasets extracted from the Spotify API of random songs into different clusters. The results show that the K-means algorithms outperform fuzzy c-means (FCM) and probabilistic c-means (PCM), using silhouette scores as a performance comparison criterion. Values of silhouette coefficient that performed the best was for K-means (n=6) and is 0.465. Additionally, we observed that K-means performed better when value of n was greater; however, FCM showed a decline in performance as n increased, whereas PCM showed a steady performance across all n clusters. Utilizing clustering algorithms for categorizing songs based on audio data gives insightful analysis and useful applications, which has been increasingly popular in recent years.