0
$\begingroup$

The YearPredictionMSD dataset, is a machine learning practice dataset extracted from the Million Song Dataset and found in the UCI Machine Learning Repository. I couldn't find the field list that describes the features. Does anyone know where I can find them?

Also, I'm confused why there are only 54 features in the original data from LabRosa but 90 in the subset of the data on UCI.

$\endgroup$
1
  • 1
    $\begingroup$ The attributes are: 90 attributes, 12 = timbre average, 78 = timbre covariance The first value is the year (target), ranging from 1922 to 2011. Features extracted from the 'timbre' features from The Echo Nest API. We take the average and covariance over all 'segments', each segment being described by a 12-dimensional timbre vector. There's contact info for T. Bertin-Mahieux on the dataset page for more information. I actually count 91 attributes, the year, the 12 timbre average, then the 78 timbre covariances. $\endgroup$ Dec 14 '16 at 14:13
1
$\begingroup$

You don't get a full data dictionary with this dataset. Instead the explanation (from your link) is:

90 attributes, 12 = timbre average, 78 = timbre covariance The first value is the year (target), ranging from 1922 to 2011. Features extracted from the 'timbre' features from The Echo Nest API. We take the average and covariance over all 'segments', each segment being described by a 12-dimensional timbre vector.

This is basically features derived direct from the audio. This is typical for problems in signal analysis, and the UCI dataset has been cut down so that it is purely about matching audio summaries to year of production.

The data dictionary from the original MSD dataset shows much more metadata. There are 54 feature types, but actually thousands of features - each type can be repeated many times. The UCI subset is not only subset by row, but by columns too, and only uses some of the MSD data.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.