A Collaborative Filtering Recommendation Algorithm Based On Conditional Entropy Trust Model

The traditional collaborative recommendation algorithm doesn't take the non-linear dependence between users into consider, which is not accurate enough in prediction. The collaborative filtering algorithm based on entropy can measure the nonlinear characteristics of users, but it can't reasonably describe the relationship between users and is subject to sparsity. To address this problem, conditional entropy trust model is proposed, which uses the conditional entropy to describe the non-linear dependence between users, and Laplace estimation is introduced to alleviate sparsity. A collaborative filtering algorithm based on the conditional entropy trust model (CECF) is designed. The experiments show that this algorithm doesn't increase the time complexity and significantly improve the degree of accuracy.