一种改进List-wise的科技论文推荐方法研究

发布时间：2019-07-27 08:17

【摘要】：近些年,科研社交网站中的科技论文数量呈现出爆炸式增长的趋势,用户很难发现符合自己要求的科技论文,而科技论文推荐正是解决这个问题的有效方法之一。但是现有科技论文推荐方法大多专注于评分预测的准确性,忽视了推荐科技论文之间的排序问题,并且现有的科技论文推荐方法没有充分利用科研社交网站中的社会化信息。为此,提出了一种改进List-wise的科技论文推荐方法,系统地分析了科研社交网站中的好友关系,科技论文的标题、摘要和标签等社会化信息,并将其融入到List-wise方法中。为了验证提出方法的有效性,抓取了科研社交网站Cite ULike上的数据进行验证,实验结果表明,与其他传统的推荐方法相比,该方法取得了较好的实验结果,具有良好的可扩展性。
[Abstract]:In recent years, the number of scientific and technological papers in scientific and social networking sites has shown an explosive growth trend, it is difficult for users to find scientific and technological papers that meet their own requirements, and the recommendation of scientific and technological papers is one of the effective ways to solve this problem. However, most of the existing scientific and technological paper recommendation methods focus on the accuracy of scoring and prediction, ignoring the ranking problem between recommended scientific and technological papers, and the existing scientific and technological paper recommendation methods do not make full use of the social information in scientific research social networking sites. Therefore, this paper puts forward a method to improve the recommendation of scientific and technological papers in List-wise, systematically analyzes the social information such as friend relationship, title, abstract and label of scientific and technological papers in scientific research social networking sites, and integrates them into List-wise method. In order to verify the effectiveness of the proposed method, the data on the scientific research social networking site Cite ULike are grasped for verification. The experimental results show that compared with other traditional recommended methods, this method has achieved better experimental results and has good scalability.
【作者单位】： 合肥工业大学科学技术研究院;合肥工业大学管理学院;
【基金】：国家自然科学基金资助项目(71101042,71471054,91646111) 安徽省自然科学基金资助项目(1608085MG150) 合肥工业大学应用科技成果培育计划资助项目(JZ2017YYPY0235)
【分类号】：TP391.3
，

本文编号：2519876