CBLN: A User Behavior Prediction Model Considering Content Correlation

The core of user behavior prediction is Click-through Rate (CTR) prediction, which requires deep analysis of user actions such as clicks and browse. It is crucial to model user behavior sequences, content attributes, and user attributes to understand user preferences and improve CTR prediction accuracy. While many CTR prediction models are based on historical behavior sequences, most treat historical behavior representations merely as preferences, neglecting the representation of correlations between users and content. To address this, we propose a new model called the Correlated Behavior Learning Network (CBLN), for the task of user behavior prediction. This model includes a User Behavior Tracking Network (UBTN) and a Content Relevance Network (CRN), which dynamically modeling the evolution of user preferences related to target content and introduce content relevance to more comprehensively capture user behavior, thus more accurate predictions of user behavior can be made. We conducted extensive experiments on two public datasets to validate the effectiveness of our model, and the results showed that our proposed model achieved better performance than many complex CTR prediction models.