基于CE3:k-means的协同过滤推荐模型研究

Research on collaborative filtering recommendation model based on CE3: k-means

随着网络数据量的迅速增长,传统数据处理方式的推荐算法已经不能满足互联网发展的需求,为了追求推荐精确性与人性化,协同过滤算法以其更高的推荐满意率逐渐取代其他推荐算法。然而,协同过滤算法推荐的准确程度取决于用户或者物品相似度的计算,成员偏好的多元性使得用户相似度并不能很好的体现用户之间的关联程度。针对这一问题,将CE3:k-menas引入协同过滤推荐,借鉴其基本思想,以成员偏好为特征,根据成员与簇类中心的距离进行偏好划分,由于边界成员与簇类中心成员在一定程度上有着相似的偏好同时也存在较大的差异。因此,针对成员距离类簇中心的远近采取不同的偏好融合策略。实验结果表明,所提出的算法相比LM-CF,UCCF和UBCF算法在准确率、召回率和平均绝对误差上效果提升明显。

With the rapid growth of network data, the traditional data processing method of recommendation algorithms can no longer meet the needs of Internet development. However, the accuracy of collaborative filtering algorithm recommendation depends on the calculation of user or item similarity. The diversity of member preferences makes user similarity not a good reflection of the degree of association between users. In response to this problem, this paper introduces CE3: k-menas clustering into collaborative filtering recommendation, draws on the basic idea of CE3: k-menas clustering, features member preferences, and divides preferences based on the distance between members and the cluster center. The boundary members and the cluster center members have similar preferences to a certain extent but also have larger differences. Therefore, different preference fusion strategies are adopted for the distance between the members and the cluster center. The experimental results show that the algorithm proposed in this paper has significantly improved accuracy, recall and average absolute error compared to LM-CF, UCCF and UBCF algorithms.