基于多维特征聚类和用户评分的景点推荐算法  

Scenic spots recommendation algorithm based on multi-dimensional feature clustering and user score

作　　者：程鹏 柳林[1,2] 刘晓 许传新[3] 郭慧 CHENG Peng;LIU Lin;LIU Xiao;XU Chuan-xin;GUO Hui(Geomatics College,Shandong University of Science and Technology,Qingdao 266590,China;Key Laboratory of Surveying and Mapping Technology on Island and Reef,NationalAdministration of Surveying,Mapping and Geoinformation,Qingdao 266590,China;Geological Surveying and Mapping Institute ofShandong Province,Jinan 250002,China)

机构地区：[1]山东科技大学测绘科学与工程学院,山东青岛266590 [2]国家测绘局海岛(礁)测绘技术国家测绘地理信息局重点实验室,山东青岛266590 [3]山东省地质测绘院,山东济南250002

出　　处：《计算机工程与设计》2019年第5期1322-1327,共6页Computer Engineering and Design

基　　金：山东省自然科学基金项目(ZR2012FM015);海岛(礁)测绘技术国家测绘地理信息局重点实验室基金项目(2014B08);卫星测绘技术与应用国家测绘地理信息局重点实验室经费基金项目(KLAMTA-201407)

摘　　要：针对传统的协调过滤推荐算法利用单一评分矩阵带来的数据稀疏性问题,提出一种基于多维特征聚类和用户评分的景点推荐算法。划分用户类别,使用基于属性权重的加权K-means聚类算法将表示用户特征的多维指标数值进行聚类;确定目标用户类别,引入用户的推荐可信度和质量可信度并形成评分可信度,将评分可信度和评分相似度结合平衡因子来计算用户之间的相似度,优化传统的相似度推荐算法。实验结果表明,该算法降低了数据的稀疏性,提高了推荐精度,具有更好的稳定性。To solve the problem of data sparsity of traditional coordinated filtering recommendation algorithm by using single score matrix,scenic spots recommendation algorithm based on the multi-dimensional feature clustering and user score was proposed.The multidimensional index values of user characteristics were clustered using the k-means algorithm based on optimization of attribute weight.Target user ’s category was determined.The score reliability was designed by recommendation credibility and quality credibility and it was used to combine with rating similarity to compute the similarities between users.The simila- rity algorithm was optimized.Experimental results show that this algorithm not only reduces the data sparsity,but also improves the recommended precision with better stability.

关 键 词：多维特征 用户聚类 评分可信度 评分相似度 景点推荐 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]