This project explores unsupervised learning on a large-scale music dataset from Spotify, containing over 1.2 million tracks. By leveraging audio features such as danceability, energy, and tempo, the goal is to group songs into meaningful clusters and uncover hidden patterns in musical characteristics.

Spotify Tracks Dataset (1.2M+ songs with audio features)
Original Source: Kaggle (Rodolfo Figueroa)
Note: I have manually deleted few columns from original dataset which were not useful & increased the dataset size significantly according to me.

To cluster Spotify songs based on their audio features using K-Means, identify the optimal number of clusters using the Elbow Method, and assign intuitive names to each cluster by analyzing representative songs.

• Trained K-Means clustering for a range of cluster values
• Applied the Elbow Method to determine the ideal number of clusters
• Selected representative tracks from each cluster
• Assigned custom names to clusters based on retrieved audio traits

• Successfully grouped tracks into distinct musical categories
• Provided clear visualizations of the Elbow curve and cluster distributions
• Analyzed the number of songs in each custom-named cluster to interpret the overall structure of the Spotify dataset

Spotify/
│
├── Spotify-Clusters.ipynb    🔹 Jupyter notebook containing entire ML Workflow
├── Spotify-Clusters.py       🔹 Python File
├── newplot.png               🔹 Result Image
├── LICENSE                   🔹 MIT License
└── README.md                 🔹 This file !!

Anuj Kulkarni - aka - steam-bell-92