Causal model for recommondation
- Publication Type:
- Thesis
- Issue Date:
- 2025
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Recommender systems (RSs) are central to digital ecosystems, offering personalized item suggestions by analyzing users’ historical interactions. However, a major challenge in RSs is the prevalence of spurious correlations—misleading patterns in interaction data caused by confounding effects or external factors such as exposure mechanisms—which can obscure true user preferences and degrade recommendation quality. Fortunately, causal inference provides a principled statistical framework for modeling causal relationships between variables while accounting for confounding effects, helping identify true causal factors influencing recommendation outcomes. This thesis introduces a comprehensive causal-based framework to address spurious correlations in RSs through three core contributions: (1) the development of foundational causal models to reduce confounding biases for unbiased recommendations, (2) advanced causal techniques that improve model robustness in complex recommendation scenarios, and (3) methods leveraging causal insights to enhance recommendation explainability via counterfactual reasoning. Extensive experiments on real-world datasets demonstrate that the proposed causal models consistently outperform state-of-the-art baselines in both accuracy and interpretability. These findings validate the effectiveness of causal inference in distinguishing genuine user preferences from spurious correlations, advancing the robustness and transparency of modern recommender systems.
Please use this identifier to cite or link to this item: