摘要
互联网下,每时每刻产生的新闻报道堪称海量,用户很难从海量报道中获取有用信息,推荐是解决该问题的重要方案。协同过滤推荐算法常用于物品推荐,在新闻场景下,新闻推荐与物品推荐有些许不同,本文在传统协同过滤的基础上针对新闻场景的特性提出了一种个性化新闻推荐方式,通过分析用户活跃度和新闻时效性对推荐的影响,对协同过滤做出改进、从物品和用户的角度进行新闻召回,使用冷启动进行召回补充,筛选召回结果,选取TopK新闻报道产生个性化推荐结果。实验结果表明,本文针对新闻场景下设计的个性化推荐模型在数据稀疏和冷启动情况下推荐效果较优。
Under the Internet,the news reports generated at every moment can be called massive,and it is difficult for users to obtain useful information from the massive reports,and recommendation is an important solution to this problem.Collaborative filtering recommendation algorithm is often used in item recommendation,in the news scenario,news recommendation and item recommendation are slightly different.This paper proposes a personalized news recommendation method for the characteristics of news scene on the basis of traditional collaborative filtering,by analyzing the impact of user activity and news timeliness on recommendation,making improvements to collaborative filtering,news recall from the perspective of items and users,using cold start for recall supplementation,screening recall results,and selecting TopK news reports to generate personalized recommendation results.Experimental results show that the personalized recommendation model designed in this paper for news scenarios has a good recommendation effect under the condition of sparse data and cold start.
作者
王君威
余粟
WANG Junwei;YU Su(School of Mechanical and Automotive Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《智能计算机与应用》
2023年第11期197-201,共5页
Intelligent Computer and Applications
关键词
新闻推荐
个性化推荐
协同过滤
news recommendations
personalized recommendations
collaborative filtering algorithm
