融合相对评分的个性化兴趣点推荐算法

A RECOMMENDED ALGORITHM BASED ON PERSONALIZED POINT OF INTEREST AND RELATIVE SCORES

现有的基于位置的社交网络中兴趣点推荐算法大都没有充分挖掘个性化信息,没有综合考虑影响用户评分行为的各种因素,致使不能很好地理解用户的兴趣,不能满足用户对个性化兴趣点推荐的需求。针对这些问题,提出一种融合相对评分的个性化兴趣点推荐算法。通过对个性化问题的理解和用户评分行为的研究,以效用模型和比较模型为依据,提出用户相对评分概念;通过相对评分来学习用户的个性化偏好,并将相对评分信息转化为元路径的权值,使用推荐算法生成有序的推荐列表,形成一个效果理想的个性化兴趣点推荐系统。通过选取用户专家的评分数据进行实验,保证数据的充实性,缓解用户冷启动造成的数据稀疏性问题。通过召回率、准确率函数与其他算法进行对比,实验结果证明,该算法在个性化兴趣点推荐方面具有明显的优势。

Most existing location-based recommendation algorithm for interest points in social networks do not fully exploit personalized information and do not take into account various factors that affect the users＇ rating behaviors, resulting in poor understanding of the users＇ interest and failure to satisfy- users＇ demands for personalized interest points recommendation. Aiming at these problems, we proposed the personalized point of interest recommendation algorithm based on relative scores. On the basis of utility model and the comparison model, the concept of user relative scores was presented through the study of understanding of personalized problems and user rating behavior. The user＇s personalized preference was learned by relative scores, and the relative score information was converted into the weight of the meta- path. We utilized the recommendation algorithm to generate an ordered list of recommendation and form an ideal personalized interest point recommendation system. Experiments were pefformed by selecting the user experts＇ rating data to ensure the data enrichment and ease the data sparseness caused by the cold start of the user. Comparing the recall rate and the accuracy rate of the proposed algorithm with those of other algorithms, the experimental results show that the algorithm has obvious advantages in personalized point of interest recommendation.