Petrokosta, Thekla (2026) Modelling Spotify Popularity Using Beta Generalized Linear Mixed Models. Bachelor's Thesis, Mathematics.
|
Text
BachelorsProjects5513286.pdf Download (2MB) | Preview |
Abstract
This thesis develops statistical models that predict the Spotify popularity score of songs appearing on Billboard’s Hot 100 chart between 2015 and 2020, using audio features assigned by Spotify as predictors. Accounting for the genre-clustered structure of the data and the bounded, skewed nature of popularity scores, the Beta Generalized Linear Mixed Model (GLMM) is proposed as the baseline framework, with the random component following a Beta distribution. The designed models are assessed in terms of their predictive fit, accuracy, and their ability to reproduce the relative ordering of songs on the Billboard Hot 100. The models are also used to explore the effect of audio features on Spotify popularity. Finally, a Monte Carlo Simulation Study is conducted to investigate whether the Gaussian distribution could serve as a reliable random component for the models, instead of the Beta distribution. Considering the known limitations of the Gaussian GLMM, when applied to bounded and skewed data, this study aims to assess the accuracy of inferences and determine a ”boundary of skewness” beyond which the Gaussian model fails to perform reliably.
| Item Type: | Thesis (Bachelor's Thesis) |
|---|---|
| Supervisor name: | Grzegorczyk, M.A. and Krijnen, W.P. |
| Degree programme: | Mathematics |
| Thesis type: | Bachelor's Thesis |
| Language: | English |
| Date Deposited: | 28 Jul 2026 06:52 |
| Last Modified: | 28 Jul 2026 06:52 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/37469 |
Actions (login required)
 |
View Item |