Music Industry Scalar Analysis Using Unsupervised Fourier Feature Selection

In the continuing investigation of understanding the connection between music information retrieval (MIR) and the economics of music this paper focuses on mining economically viable song attributes. Identical features and attributes found in songs in one genre or place in time can have completely inverse translations as to the economic viability of a song. Partial cognates are pairs of features in two songs that have the same affect in portions of scalar theory, but not completely. Learning when partial cognates differ can be useful for MIR and while feature selection is effective in dimensionality reduction the meta-dimensionality of data in music that generates money it is a non-trivial challenge to determine viable feature selection methods in the field of data mining economically viable music. We present a novel methodology for music industry scalar analysis using unsupervised fourier feature selection.