摘要
在互联网高速发展的今天,网络新闻已成为人们获得信息的主要途径,如何准确地为用户提供个性化的新闻推荐已成为业内人士日益关注的问题。为解决这一问题,出现了很多基于LDA的新闻推荐,但它们只进行新闻内容的分析,没有考虑用户兴趣的变化。针对此问题,本文提出了一种基于主题兴趣变化的热点新闻推荐算法。首先,用固定时间窗大小划分用户的阅读历史,并在每个阶段根据用户的阅读历史,利用LDA得到用户兴趣的概率分布。其次,利用时间惩罚加权函数和用户在每个阶段的新闻主题分布预测用户下一阶段的可能兴趣。最后,根据用户兴趣概率分布利用基于用户的协同过滤和待推荐新闻的主题分布完成热点新闻推荐。通过实际数据集上的实验表明,该方法提高了推荐的性能。
In the rapid development of the Internet today, network news has become the main way for people to get information, how to accurately provide personalized news recommendation for users has become an increasingly concerned problem in the industry. To solve this problem, many news recommendations based on LDA have appeared, but they only analyze the news content without considering the changes of users’ interests. To solve this problem, this paper proposed a hot news recommendation algorithm based on topic interest change. Firstly, users’ reading history is divided by the fixed time window size, and the probability distribution of users’ interest is obtained by LDA according to users’ reading history in each stage. Secondly, the time penalty weight function and the news topic distribution of users in each stage are used to predict the possible interest of users in the next stage. Finally, according to the user interest probability distribution, user-based collaborative filtering and topic distribution of news are used to complete hot news recommendation. Experiments on real data sets show that the proposed method improves the recommended performance.
出处
《计算机科学与应用》
2019年第10期1831-1838,共8页
Computer Science and Application
