A Survey of Learning-Based Methods for Cold-Start, Social Recommendation, and Data Sparsity in E-commerce Recommendation Systems

GCZRec: Generative Collaborative Zero-Shot Framework for Cold Start News Recommendation

“…Similarly, in cold start item recommendations, the aim is to predict whether an item should be recommended to a particular user without any available historical interactions for that item. In cold start user case, items are to be recommended to a particular user for which there are no existing historical information [13]. Following this intuition, the features of news items and users can be used to deduce the behavioral context of cold start items and users in recommendation scheme just like a class label can be predicted for an unseen data sample using the generalization from known samples in zero-shot classification.…”