基于邻居模型的协同过滤算法的应用研究

Application of collaborative filtering algorithm based on neighbor model

协同过滤算法通过对目标用户的行为数据进行分析和建模,预测其可能感兴趣的项目并进行推荐,从而缓解信息过载问题。对不同特征的Movie Lens数据集分别进行了基于用户和基于项目邻居模型的协同过滤推荐算法的性能对比实验。实验结果表明,基于用户的邻居模型适用于用户数远远小于项目数的推荐系统,相反情况下基于项目的邻居模型则具有更高的效率。对于实际系统中的推荐算法选择具有一定的参考价值。

The collaborative filtering algorithm mitigates the problem of information overload by analyzing and modeling the behavior data of the target users,predicting the items that may be of interest and recommending them. The performance comparison experiments of user-based and item-based neighbor model collaborative filtering recommendation algorithm are carried out for different characteristics of the Movie Lens datasets. The experimental results show that the user-based neighbor model is suitable for the recommendation system with the number of users far less than the number of items. In the opposite case,the item-based neighbor model is more efficient. It has certain reference value for the selection of the recommended algorithm in the actual system.