Abstract: Collaborative filtering is one of the most effective way in the recommended system. The forecast accuracy of recommendation algorithm depends on two key points：the extraction of the nearest neighbors and the calculation of project/user similarity. The paper extracts rated most neighbors with the current project as a nearest neighbor candidate set, and proposes a weighted cosine similarity method to calculate the project/user similarity, then optimizing the weight by the Particle Swarm Optimization（PSO） algorithm. The experimental results show that using these methods can efficiently improve the accuracy of the score predicted, and provide better recommendation results than traditional collaborative filtering algorithms.

Key words: recommended system, Particle Swarm Optimization（PSO）, collaborative filtering

摘要： 协同过滤是推荐系统中最有效的方法之一，推荐算法评分预测的精确性受到最近邻居的提取以及项目或用户相似度计算的两个关键点的影响。根据用户行为相似性原理，采用最大交集法提取与当前项目共同评分最多的邻居作为最佳邻居候选集，同时提出了加权余弦相似性方法对相似度进行计算，并采用粒子群优化算法（PSO）对权重进行优化求解。实验结果表明，采用上述方法相对于传统方法来说，能较好地改善评分预测的精确度，有效地提高推荐系统的推荐质量。

关键词: 推荐系统, 子群算法, 协同过滤