<![CDATA[The rapid growth of music streaming platforms has created very large catalogs, making popularity patterns difficult to understand using a single indicator. Total streams reflect cumulative success, but they do not always represent current listening momentum. This situation motivates the need for song segmentation based on more informative popularity patterns to support decision-making for streaming platforms, artists, and labels. This study applied a data mining approach using K-Means clustering to group Spotify most-streamed songs based on streaming popularity indicators. The main contribution was a segmentation framework that combined total streams, daily streams, and a daily-to-total streams ratio to better capture current momentum. The method included data cleaning, missing value imputation, logarithmic transformation to reduce skewness, feature engineering of a ratio variable, feature standardization, K-Means training, cluster number selection using the elbow method and Silhouette Score, and evaluation using Inertia, Silhouette Score, the Calinski–Harabasz Index, and the Davies–Bouldin Index. The final model with k = 4 achieved an Inertia of 2673.011 and a Silhouette Score of 0.364835 and produced four interpretable segments. Cluster 0 represented super-trending songs with the highest daily-to-total ratio, cluster 1 represented legacy popular songs with low daily activity, cluster 2 represented mega hits with extremely high total streams and still strong daily activity, and cluster 3 represented consistently performing songs with stable daily streams. These segments provided practical insights for promotion prioritization, playlist curation, and trend interpretation.]]>
