Abstract:

In order to solve the problem of the data sparsity and low recommendation accuracy for traditional collaborative filtering recommendation algorithm, a collaborative filtering recommendation algorithm for clustering optimization is proposed. According to the difference of the user’s scores, the original rating matrix is preprocessed. And the user category preference matrix, which combined by the user item rating matrix and item type matrix is constructed for reflecting users’ preferences better and alleviating the sparsity of data. Then the fuzzy clustering algorithm optimized by flower pollination is used on the matrix to cluster users and enhancethe clustering effect of users. And the similarity of item preference information and the similarity of item rating matrix are weighted and summed to obtain multiple nearest neighbors. The time factor is integrated to predict the item score of target users and improve the influence of user interest change on the recommendation effect. The experimental results on the MovieLens 100k dataset show that the proposed algorithm mitigates the data sparsity and improves the recommendation accuracy.

Key words: collaborative filtering, recommendation algorithm, fuzzy clustering, flower pollination algorithm

摘要：

针对传统的协同过滤推荐算法存在评分数据稀疏和推荐准确率偏低的问题，提出了一种优化聚类的协同过滤推荐算法。根据用户的评分差异对原始评分矩阵进行预处理，再将得到的用户项目评分矩阵以及项目类型矩阵构造用户类别偏好矩阵，更好反映用户的兴趣偏好，缓解数据的稀疏性。在该矩阵上利用花朵授粉优化的模糊聚类算法对用户聚类，增强用户的聚类效果，并将项目偏好信息的相似度与项目评分矩阵的相似度进行加权求和，得到多个最近邻居。融合时间因素对目标用户进行项目评分预测，改善用户兴趣变化对推荐效果的影响。通过在MovieLens 100k数据集上实验结果表明，提出的算法缓解了数据的稀疏性问题，提高了推荐的准确性。

关键词: 协同过滤, 推荐算法, 模糊聚类, 花朵授粉算法