Clustering: Creating a Song Recommender
This project focuses on using clustering techniques to recommend songs based on their audio features. The dataset includes 20,311 tracks released between the years 2000 and 2020. The data are evaluated and verified to be suitable for a machine learning model. A pipeline is created to scale the data and reduce dimensionality to two dimensions. KMeans is then used to cluster the tracks. Euclidean distance is used to find the "closest" songs to a selected song within a cluster. The album images of these recommended songs are retrieved using the Spotipy library to provide a graphical representation of the 10 recommended songs similar to the chosen one.
The project focuses on the following:
- Understand how music recommendation systems work.
- Learn and use unsupervised machine learning methods for music classification.
- Identify the criteria and features of songs used for recommendations.
- Create a music recommender and connect it with the Spotify API.
- Gain in-depth knowledge of the Spotify API using the Spotipy library.