Evaluating GNNs for Success Prediction in Artist Collaboration Networks

This research analyzes artist collaboration networks in the music industry, building on prior studies of Italian and Danish networks by introducing a new dataset from the Polish music scene. It facilitates direct comparison across these distinct European music landscapes and combines them into a merged network. Furthermore, the study proposes a framework to evaluate the effectiveness of Graph Neural Networks (GNNs) in predicting artist popularity, utilizing metadata and network position as key factors.

Why it matters

Understanding the dynamics of collaboration networks and leveraging advanced analytical techniques like GNNs for prediction can offer significant strategic advantages. This enables more informed decision-making regarding talent identification, partnership formation, and resource allocation within collaborative ecosystems, potentially enhancing efficiency and fostering growth.

Key insights

• The study expands on existing analyses of artist collaboration networks by integrating a new Polish music scene dataset.
• It enables direct comparative analysis across Italian, Danish, and Polish music landscapes.
• The research merges these distinct European networks into a unified network structure.
• A framework is introduced to assess Graph Neural Networks (GNNs) for predicting artist popularity.
• Artist popularity predictions are based on metadata and an artist's position within the collaboration network.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.02920