融合用户影响力和灰关联分析的Slope One推荐算法

Slope One Algorithm Integrating User Influence and Grey Association Analysis

Slope One算法在推荐系统中得到了广泛的应用,但该算法并未考虑用户间的差异性,并且存在数据稀疏、冷启动及灰羊用户等一系列问题.因此,采用均衡接近度灰关联方法计算用户间相似性,有效捕捉用户之间的潜在关联;并引入PageRank算法,利用图结构来建模用户间的信任关系,将计算出的PR值作为体现用户影响力水平的参数;最后将均衡接近度和用户影响力作为权重因子加权到评分预测过程中.实验结果显示,该算法可以有效提高推荐精度,缓解数据稀疏、冷启动和灰羊用户等问题对算法的影响,在数据稀疏、邻域较少的情况下,仍然可以达到较高的推荐准确率和较快的收敛速度.

Slope One algorithm has been widely used in recommendation systems,however,the algorithm does not consider the variability among users and suffers from a series of problems such as data sparsity,cold start,and gray sheep users.Therefore,the gray correlational analysis method by the balanced adjacent degree is adopted to measure the similarity between users and effectively explore the potential connections between users;in addition,the PageRank algorithm is introduced to model the trust relationship among users using the graph structure,and the calculated PR value is used as a parameter to reflect the level of user influence;finally,the balanced adjacent degree and the user influence are weighted as weight factors in the rating prediction process.The experimental results demonstrate that the algorithm can effectively improve the prediction accuracy of the algorithm,alleviate the impact of problems such as data sparsity,cold start and gray sheep users on the algorithm,and still achieve high recommendation accuracy and fast convergence speed in the case of sparse data and fewer neighbors.