Music Recommendation Mapping and Interface Based on Structural Network Entropy

Recommendation systems generally produce the results of their output to their users in the form of an ordinal list. In the interest of simplicity, these lists often obscure, abstract, or omit many relevant metrics pertaining to the measured "strength" of the recommendations or the relationships the recommended items share with each other. This information is often useful for coming to a better understanding of the nature of how the items are structured according to the recommendation data. This paper describes the ZMDS algorithm, a novel way of analyzing the fundamental network structure of recommendation results. Furthermore, it also describes a nonlinear repulsion plot method as a utility for mapping and interacting with the results generated by the ZMDS algorithm on music recommendation data. A novel "recommendation map" Web application implements both the ZMDS algorithm and the repulsion plot interface and are offered as an example of both components working together.