基于用户特征和评分的精准推荐策略研究基于用户特征和评分的精准推荐策略研究
Accurate recommendation strategy based on user characteristics and ratings
  
DOI:
中文关键词:  协同过滤推荐;用户冷启动;K means聚类算法
英文关键词:collaborative filtering recommendation algorithm; user cold start; K means clustering algorithm
基金项目:国家自然科学基金(61572260,61872196)资助项目
作者单位
傅金京 南京邮电大学 计算机学院,江苏南京210023 
李玲娟 南京邮电大学 计算机学院,江苏南京210023 
摘要点击次数: 184
全文下载次数: 67
中文摘要:
      个性化推荐系统是帮助用户发现内容,克服信息过载的重要工具。为了提高推荐算法的准确率和效率,综合协同过滤推荐算法和K means聚类算法,设计了一种基于用户特征和评分的精准推荐策略。该策略一方面针对新用户冷启动问题,引入K means聚类算法对全体用户特征进行聚类,将新用户所属类中其他用户喜好的物品中的Top N个推荐给新用户;另一方面根据物品数和用户数的大小关系,或者不同推荐算法所得F1值的大小关系,来决定选择将哪种推荐算法产生的结果推荐给老用户。在Movielens和FilmTrust数据集上的实验结果表明,这种基于用户特征和评分的精准推荐策略能够有效地针对新用户和老用户做出准确的最佳推荐。
英文摘要:
      A personalized recommendation system is an important tool to help users discover the content and overcome the information overload. To improve the accuracy and the efficiency of the recommendation algorithm, a collaborative filtering recommendation algorithm and a K means clustering algorithm are combined to design an accurate recommendation strategy based on user characteristics and ratings. Aiming at the cold start problem of new users, K means clustering algorithm is adopted to cluster all user characteristics, and the Top N items preferred by other users in the class to which the new user belongs are recommended to the new users. According to the size relationship between the number of items and the number of users, or the size relationship between the F1 values obtained by different recommendation algorithms, results with the appropriate recommendation algorithm are chosen for the existing users to solve the problem,that is, a fixed recommendation algorithm is not used properly for existing users. Experimental results on Movielens and FilmTrust datasets show that the strategy is effective in making accurate and best recommendations for both new and existing users.
查看全文  查看/发表评论  下载PDF阅读器

你是第2143162访问者
版权所有《南京邮电大学学报(自然科学版)》编辑部
Tel:86-25-85866913 E-mail:xb@njupt.edu.cn
技术支持:本系统由北京勤云科技发展有限公司设计