Dpopgcf: a fairness-driven graph convolutional collaborative filtering model with flexible negative sampling

Key Points

• The model enhances fairness in recommendations, leading to reduced bias in user data.
• Key evidence indicates a substantial improvement in fairness metrics across various user groups.
• Analysis using graph convolutional networks and collaborative filtering methods shows effective negative sampling strategies.
• This approach supports better user experiences, highlighting the need for fairness in algorithmic recommendations.