Temporal Dynamics of User Preferences

User preferences are often assumed to be static, but in reality, interests can shift over time due to changing contexts, seasons, or personal growth. This thesis explores how user preferences evolve over time in recommender systems. Using timestamped interaction data long-term datasets, we aim to analyze trends in e.g., genre preferences, item novelty, and user engagement. The goal is to evaluate recommendation strategies that account for temporal decay (e.g., giving more weight to recent interactions) and more comprehensive approaches for preference-shift-aware recommendations and to evaluate their impact on recommendation quality.