第15卷第6期 智能系统学报 Vol.15 No.6 2020年11月 CAAI Transactions on Intelligent Systems Nov.2020 D0L:10.11992tis.201710024 融合用户特征优化聚类的协同过滤算法 梁丽君，李业刚，张娜娜，张晓，王栋 (山东理工大学计算机科学与技术学院，山东淄博255049) 摘要：针对推荐系统领域中应用最广泛的协同过滤推荐算法仍伴随着数据稀疏性、冷启动和扩展性问题，基 于用户冷启动和扩展性问题，提出了基于改进聚类的PCEDS(pearson correlation coefficient and euclidean distance similarity)协同过滤推荐算法。首先针对用户属性特征，采用优化的K-means聚类算法对其聚类，然后结合基于 信任度的用户属性特征相似度模型和用户偏好相似度模型，形成一种新颖的PCEDS相似度模型，对聚类结果 建立预测模型。实验结果表明：提出的PCEDS算法比传统的协同过滤推荐算法在均方根误差(RMSE)上降低 5%左右，并且推荐准确率(precision)和召回率(recal)均有明显提高，缓解了冷启动问题，同时聚类技术可以节 省系统内存计算空间，从而提高了推荐效率。 关键词：推荐系统：协同过滤；冷启动：扩展性：优化聚类；信任度；用户属性特征：用户偏好 中图分类号：TP311文献标志码：A文章编号：1673-4785(2020)06-1091-06 中文引用格式：梁丽君，李业刚，张娜娜，等.融合用户特征优化聚类的协同过滤算法J引.智能系统学报，2020,15(6)： 1091-1096. 英文引用格式：LIANG Lijun,,LI Yegang,.ZHANG Na'na,et al.Collaborative filtering algorithm combining user features and pref- erences in optimized clustering J].CAAI transactions on intelligent systems,2020,15(6):1091-1096. Collaborative filtering algorithm combining user features and preferences in optimized clustering LIANG Lijun,LI Yegang,ZHANG Na'na,ZHANG Xiao,WANG Dong (College of Computer Science and Technology,Shandong University of Technology,Zibo 255049,China) Abstract:The collaborative filtering recommendation algorithm in the field of recommendation systems is still accom- panied by the data sparsity,cold start,and scalability problems.To solve the cold start and scalability problems,we pro- pose a PCEDS(pearson correlation coefficient and euclidean distance)collaborative filtering recommendation algorithm based on optimized clustering.First,the optimized K-means clustering algorithm is used to cluster the attributes of users. Then,based on the trust-based similarity model of user attribute features and the similarity model of user preference,a novel PCEDS similarity model is established to create a prediction model for the clustering results.The experimental results indicate that,compared with the traditional collaborative filtering recommendation algorithm,the proposed PCEDS collaborative filtering recommendation algorithm reduces the root mean square error by approximately 5%,sig- nificantly improves the recommendation precision and recall,and solves the cold start problem.Simultaneously,the clustering technology can save the memory space of the recommendation system,thereby improving its efficiency. Keywords:recommendation system;collaborative filtering;cold start,scalability;optimization clustering;trust degree; user attribute;user preference 随着互联网和移动技术的飞速发展，现在越 导致了信息超载问题。当用户搜索其感兴趣的信 来越多的人拥有智能手机、平板电脑和其他的智 息时，会花费大量的时间和精力去过滤掉无用的 能终端，这使得生产信息的速度呈爆炸式增长， 信息，然而结果往往得不到用户的满意，于是个 性化推荐技术应时而生。个性化推荐技术是指利 收稿日期：2017-10-29 基金项目：国家自然科学基金项目(61671064). 用用户某种兴趣点和购买特点，向用户推荐感兴 通信作者：李业刚.E-mail:liyegang@sdut.edu.cn 趣的内容，是缓和信息超载问题的有效途径。在DOI: 10.11992/tis.201710024 融合用户特征优化聚类的协同过滤算法 梁丽君，李业刚，张娜娜，张晓，王栋 （山东理工大学 计算机科学与技术学院，山东 淄博 255049） 摘 要：针对推荐系统领域中应用最广泛的协同过滤推荐算法仍伴随着数据稀疏性、冷启动和扩展性问题，基 于用户冷启动和扩展性问题，提出了基于改进聚类的 PCEDS(pearson correlation coefficient and euclidean distance similarity) 协同过滤推荐算法。首先针对用户属性特征，采用优化的 K-means 聚类算法对其聚类，然后结合基于 信任度的用户属性特征相似度模型和用户偏好相似度模型，形成一种新颖的 PCEDS 相似度模型，对聚类结果 建立预测模型。实验结果表明：提出的 PCEDS 算法比传统的协同过滤推荐算法在均方根误差 (RMSE) 上降低 5% 左右，并且推荐准确率 (precision) 和召回率 (recall) 均有明显提高，缓解了冷启动问题，同时聚类技术可以节 省系统内存计算空间，从而提高了推荐效率。 关键词：推荐系统；协同过滤；冷启动；扩展性；优化聚类；信任度；用户属性特征；用户偏好 中图分类号：TP311 文献标志码：A 文章编号：1673−4785(2020)06−1091−06 中文引用格式：梁丽君, 李业刚, 张娜娜, 等. 融合用户特征优化聚类的协同过滤算法 [J]. 智能系统学报, 2020, 15(6): 1091–1096. 英文引用格式：LIANG Lijun, LI Yegang, ZHANG Na’na, et al. Collaborative filtering algorithm combining user features and pref￾erences in optimized clustering[J]. CAAI transactions on intelligent systems, 2020, 15(6): 1091–1096. Collaborative filtering algorithm combining user features and preferences in optimized clustering LIANG Lijun，LI Yegang，ZHANG Na’na，ZHANG Xiao，WANG Dong (College of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China) Abstract: The collaborative filtering recommendation algorithm in the field of recommendation systems is still accom￾panied by the data sparsity, cold start, and scalability problems. To solve the cold start and scalability problems, we pro￾pose a PCEDS(pearson correlation coefficient and euclidean distance) collaborative filtering recommendation algorithm based on optimized clustering. First, the optimized K-means clustering algorithm is used to cluster the attributes of users. Then, based on the trust-based similarity model of user attribute features and the similarity model of user preference, a novel PCEDS similarity model is established to create a prediction model for the clustering results. The experimental results indicate that, compared with the traditional collaborative filtering recommendation algorithm, the proposed PCEDS collaborative filtering recommendation algorithm reduces the root mean square error by approximately 5%, sig￾nificantly improves the recommendation precision and recall, and solves the cold start problem. Simultaneously, the clustering technology can save the memory space of the recommendation system, thereby improving its efficiency. Keywords: recommendation system; collaborative filtering; cold start; scalability; optimization clustering; trust degree; user attribute; user preference 随着互联网和移动技术的飞速发展，现在越 来越多的人拥有智能手机、平板电脑和其他的智 能终端，这使得生产信息的速度呈爆炸式增长， 导致了信息超载问题。当用户搜索其感兴趣的信 息时，会花费大量的时间和精力去过滤掉无用的 信息，然而结果往往得不到用户的满意，于是个 性化推荐技术应时而生。个性化推荐技术是指利 用用户某种兴趣点和购买特点，向用户推荐感兴 趣的内容，是缓和信息超载问题的有效途径。在 收稿日期：2017−10−29. 基金项目：国家自然科学基金项目 (61671064). 通信作者：李业刚. E-mail：liyegang@sdut.edu.cn. 第 15 卷第 6 期 智 能 系 统 学 报 Vol.15 No.6 2020 年 11 月 CAAI Transactions on Intelligent Systems Nov. 2020