A carregar...
Publicação
Predicting user activity in music streaming platforms through early behavior signals: a machine-learning approach
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.2 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
This thesis examines whether early behavioural signals can predict future user activeness on a music-streaming platform. Using impression-level data from NetEase Cloud Music, user-level features are constructed from the first four actions after registration to capture interaction
intensity, social richness, depth and temporal persistence, and exposure to popular content. Several machine learning models are evaluated, including Logistic Regression, Random Forest, a multilayer perceptron, and XGBoost, with robustness checks across alternative activeness
definitions. XGBoost achieves the highest ROC-AUC and PR-AUC. SHAP analyses reveal that richness, depth, and persistence in early interactions are stronger predictors of future activeness than sheer interaction volume
Descrição
Palavras-chave
Business analytics Machine learning User activity prediction Digital platform