协同过滤推荐算法的性能对比与分析  

Performance Comparison and Analysis of Collaborative Filtering Recommendation Algorithms

作　　者：王子茹 宋尚文 阎红灿 WANG Zi-ru;SONG Shang-wen;YAN Hong-can(College of science,North China University of Science and Technology,Tangshan Hebei 063210,China;Hebei Key Laboratory of Data Science and Applications,Tangshan Hebei 063210,China)

机构地区：[1]华北大学理学院,河北唐山063210 [2]数据科学与应用重点实验室,河北唐山063210

出　　处：《计算机仿真》2022年第9期435-440,共6页Computer Simulation

基　　金：基于云课堂的Python语言混合教学模式研究(ZXXJ2019030);中国信息协会“十三五”规划课题(“实用物联网医学”课程体系建设,教育部产学研协同育人项目)(201902137008)。

摘　　要：对基于用户属性的协同过滤和基于模型的聚类型协同过滤推荐算法综述研究并进行实验重现,发现计算用户间相似性时加入用户信任值、降低用户邻居个数可增强推荐系统的可扩展性,同时基于用户属性的协同过滤和结合密度聚类的协同过滤推荐更适合高维数据集;引入C均值的协同过滤推荐不受近邻个数的影响,推荐效果稳定;K-means聚类的协同过滤在选取聚类中心时可能会陷入局部最优。经过多组实验的对比分析,各类推荐算法没有绝对优势,不同应用场景选择不一样的推荐算法可以获得最佳的推荐效果。The collaborative filtering based on user attributes and the collaborative filtering recommendation algorithm based on model were summarized and studied, and the experimental results were reproduced. It is found that adding the user trust value and reducing the number of users’ neighbors when calculating the similarity between users can enhance the scalability of the recommendation system;At the same time, the collaborative filtering based on user attributes and the collaborative filtering recommendation combined with density clustering are more suitable for high-dimensional data sets. Collaborative filtering recommendation that introduces C-means is not affected by the number of neighbors, and the recommendation effect is stable;K-means clustering collaboration Filtering may fall into a local optimum when selecting cluster centers. After comparative analysis of multiple sets of experiments, various recommendation algorithms have no absolute advantages, and different recommendation algorithms can be selected for different application scenarios to obtain the best recommendation effect.

关 键 词：协同过滤 用户相似度计算 聚类 推荐评价标准 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] TP391.9[自动化与计算机技术—计算机科学与技术]