改进信任度的商品推荐算法研究

Research on Commodity Recommendation Algorithm Based on Improved Trust Metrics

对于有个性化推荐需求的电子商务系统,传统协同过滤推荐算法对商品的用户项目矩阵构建比较单一,难以解决数据稀疏以及推荐结果精度较低等问题。为此,提出一种改进的基于信任度的协同过滤算法,根据用户历史行为,对用户项目评分矩阵进行细分量化,综合考虑用户间关系,引入信任因子维持用户信任关系中的非对称性,通过共同评分项计算用户评分信任度。最终融合信任度与信任因子,计算获得最佳邻居集并产生最终推荐列表。在淘宝官方UserBehavior数据集下进行实验,结果表明,该算法降低了推荐稀疏性,提高了推荐精度。

For the e-commerce system with personalized recommendation requirements,traditional collaborative filtering algorithm has a simple construction of the user item matrix of the commodity,which is difficult to solve the problem of data sparseness and low accuracy of recommendation results.For this reason,an improved trust recommendation algorithm is proposed with trust metrics.According to the historical behavior of the user,the user project scoring matrix is subdivided and quantified,and then the relationship between users is comprehensively considered.The trust factor is introduced to maintain the asymmetry in the user trust relationship,the user's trust is calculated by the common score item,and finally the trust metrics and the trust factor are integrated to calculate the best neighbor set and generate the final recommendation list.Experiments were carried out under the official UserBehavior dataset of Taobao.The results show that the proposed algorithm reduces the recommended sparsity and improves the accuracy of the recommendation.