Music recommendations for sleep: using Spotify features and subjective ratings to sample sleep-inducing music

This deposit contains sample data and R code for recommending music for sleep based on objective (musical features) and subjective (listener ratings) data. We present a paradigm for sampling music based on features from Spotify and investigate the reliability of subjective ratings for predicting what music is sleep inducing. It consists of three parts:

1) Analysing and clustering using PCA of features of music associated with sleep, relaxation and energising in order to create a predictive model of music for sleep based on a set of objective features. This is a replication and extension of the method and code used in Kirk & Timmers (2024), with data from Kirk et al. (2024) (see Related Materials).

2) Analysing subjective properties of music for sleep, relaxation and energising in order to create a predictive model of music for sleep based on subjective properties. This is a replication of the method and code used in Kirk et al. (2024).

3) Defining a final model that combines in an optimal way subjective and objective features to predict music for sleep. This is a replication of the method and code used in Kirk et al. (2024). This final model is proposed as a model that can be employed by future users.

The ratings data included was collected in a prior study that received ethical approval from the University of Sheffield, reference No. 036383.

For more information on the data and research background see Related Materials.