Context The Million Song Dataset (MSD) is a freely-available collection of audio features and metadata for a million contemporary popular music tracks. This is a subset of the MSD and contains audio features of songs with the year of the song. The purpose being to predict the release year of a song from audio features. Content The owners recommend that you split the data like this to avoid the 'producer effect' by making sure no song from a given artist ends up in both the train and test set. - train: first 463,715 examples - test: last 51,630 examples Field descriptions: - The first value is the year (target), ranging from 1922 to 2011. - Then there are 90 attributes - TimbreAverage[1-12] - TimbreCovariance[1-78] These features were extracted from the 'timbre' features from The Echo Nest API. The authors took the average and covariance over all 'segments' and each segment was described by a 12-dimensional timbre vector. Acknowledgements Original dataset: Thierry Bertin-Mahieux, Daniel P.W. Ellis, Brian Whitman, and Paul Lamere. The Million Song Dataset. In Proceedings of the 12th International Society for Music Information Retrieval Conference (ISMIR 2011), 2 Subset downloaded from: https://archive.ics.uci.edu/ml/datasets/yearpredictionmsd Inspiration Use this dataset to predict the years that each song was released based on it's audio features