<p>Session-based recommendation (SBR) is a challenging task that involves predicting user interactions within short and anonymous browsing sessions. One common user behavior is repeat consumption, where a user repeatedly clicks the same item over time, providing insight into the user’s consumption intents. User-item interactions in real-world sessions are influenced by various factors such as passing time and matching interests. Current SBR methods consider repeat consumption to discern user’s propensity to consume old (visited) and new (unvisited) items based on historical interactions, but they fail to explore the diverse underlying factors that affect interactions and suffer from data sparsity issues. To address these challenges, this paper proposes a multi-factor decoupling repeat-aware network for session-based recommendation, named MFDReNet. This network incorporates a multi-factor decoupling module and a contrastive learning module into the repeat-explore intent mechanism. The former learns multiple factors underlying user behaviors to capture intricate user preferences, while the latter constructs different self-supervised signals to augment session sequences for better training. The repeat-explore intent mechanism enables the model to consider the recommendation of old and new items separately, refining user behavior patterns and enhancing recommendation performances. Extensive experiments conducted on three benchmark datasets, Diginetica, Nowplaying, and Gowalla, demonstrate the efficacy and superiority of our model.</p>