The core data of the Million Song dataset is the Taste Profile subset released by The Echo Nest, which consists of three parts: "user ID, song ID, and play count."

• The data is divided into two parts:
• The training set contains more than 1m users, with complete historical releases (available on the MSD website);
• The validation and test sets contain 110k users with half of their historical releases (available on Kaggle).

This dataset was released by kaggle in 2012.