基于主题概率分布模型的个性化信息推荐系统研究  

Research on Personalized Information Recommendation System Based on Subject Probability Distribution Model

作　　者：豆志磊[1] 金洁洁[2] DOU Zhilei;JIN Jiejie(Luohe Vocational Technology College,Luohe 462000,China;Luohe Medical College,Luohe 462000,China)

机构地区：[1]漯河职业技术学院,河南漯河462000 [2]漯河医学高等专科学校,河南漯河462000

出　　处：《河南科技》2024年第15期12-16,共5页Henan Science and Technology

基　　金：漯河医学高等专科学校2022年度科技创新项目“新媒体环境下大学生信息素养影响因素及提升策略研究”(2022SKZD21)。

摘　　要：【目的】传统基于相似度计算的个性化信息推荐系统因算力要求过高、推荐时滞过长等问题,致使其无法在中小型新闻图情领域得到广泛的普及。为了帮助中小型新闻图情机构以较低的成本开展个性化信息的精准推荐服务,构建了一套基于主题概率分布模型的个性化信息推荐系统。【方法】通过数据采集技术实现原始数据语料的采集;通过LDA模型训练实现原始文本信息的分类;通过将用户信息代入LDA模型训练获取用户主题画像;将用户主题画像与文本信息分类相结合实现个性化信息推荐。【结果】经试验验证,该系统推荐时效强,可达毫秒级。通过与用户阅读记录进行比较,该系统的推荐结果均符合用户兴趣主题,具有较高的推荐精准度。【结论】该基于主题概率分布模型的个性化信息推荐系统,能够帮助中小型新闻图情机构以较低的成本开展个性化信息精准推荐服务,具有一定的应用价值。[Purposes]The traditional personalized information recommendation system based on similar-ity calculation can not be widely used in the field of small and medium-sized news picture because of the high requirement of computing power and time delay of recommendation.This paper constructs a per-sonalized information recommendation system based on subject probability distribution model to help small and medium-sized news organizations to carry out personalized information accurate recommenda-tion service with lower cost.[Methods]The data collection technology was used to collect the original data corpus;LDA model training was used to classify the original text information;the user's subject por-trait was obtained by substituting user's information into LDA model training results;the personalized in-formation recommendation is realized by combining user subject portrait with text information classifica-tion.[Findings]the experimental results showed that the system had a strong recommendation time,which could reach the millisecond level.Compared with the reading records of users,the recommenda-tion results were in accordance with user's interest topics,and had a high recommendation accuracy.[Conclusions]The personalized information recommendation system based on topic probability distribu-tion model can help small and medium-sized news picture and information organizations to develop per-sonalized information accurate recommendation service with lower cost,which has certain application value.

关 键 词：LDA主题模型 主题概率分布模型 个性化信息推荐 系统设计与实现 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]