Improving Model Generalization in Songs’ Popularity Prediction Based on Datasets with Diverse Distributions

Hongzhan Yao

2024

Abstract

Online streaming music platforms offer publics a more convenient getaway to enjoy music. It also benefits the music creators that getting fortunes from their songs. However, the factors for a songs’ success are not apparent for most of the listeners. With the help of Artificial Intelligence, making prediction of songs popularity gets easier. This study chooses an updated dataset including various song track information on Spotify. Three relevant models including Random Forest, Decision Tree, K-Nearest Neighbor are train and tested with different data splitting strategy and training strategy with the dataset about Spotify. Model’s performances are analyzed and visualized after the test. The research attempt to improve the original models with different improving strategies. The research finds that using more and relatively diverse data can help to improve the performance of the data. Using data that corresponded to the target prediction task could strongly improve the models on some specific tasks. Using random data to adjust original models may have negative impact to the models’ performance.

Paper Citation

in Harvard Style

Yao H. (2024). Improving Model Generalization in Songs’ Popularity Prediction Based on Datasets with Diverse Distributions. In Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management - Volume 1: MLSCM; ISBN 978-989-758-738-2, SciTePress, pages 342-348. DOI: 10.5220/0013330900004558

in Bibtex Style

@conference{mlscm24,
author={Hongzhan Yao},
title={Improving Model Generalization in Songs’ Popularity Prediction Based on Datasets with Diverse Distributions},
booktitle={Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management - Volume 1: MLSCM},
year={2024},
pages={342-348},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013330900004558},
isbn={978-989-758-738-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management - Volume 1: MLSCM
TI - Improving Model Generalization in Songs’ Popularity Prediction Based on Datasets with Diverse Distributions
SN - 978-989-758-738-2
AU - Yao H.
PY - 2024
SP - 342
EP - 348
DO - 10.5220/0013330900004558
PB - SciTePress