Collaborative filtering recommendation algorithm based on enhanced similarity calculation method

This paper introduces a multi-factor collaborative filtering algorithm to address limitations in traditional methods, such as ignoring time effects and user attributes. By combining cosine similarity with a modified calculation method and incorporating a popularity correction factor, it reduces bias toward popular items. A time decay function tracks user interest changes, and user attribute similarity enhances cold-start and data sparsity issues. Experiments on public datasets validate its effectiveness, significantly improving recommendation accuracy and providing insights for optimizing personalized systems.