diff options
author | Norbert Preining <norbert@preining.info> | 2021-05-30 03:01:26 +0000 |
---|---|---|
committer | Norbert Preining <norbert@preining.info> | 2021-05-30 03:01:26 +0000 |
commit | fd9ad0362e70874436480f729796d81aec6912a2 (patch) | |
tree | 52d51b2cd19abb4cf99078a424697946c443046b /macros/latex/contrib/bjfuthesis/demo/contents/abstract.tex | |
parent | c735cad430ba6a619963ad7248020b781a0a45f3 (diff) |
CTAN sync 202105300301
Diffstat (limited to 'macros/latex/contrib/bjfuthesis/demo/contents/abstract.tex')
-rw-r--r-- | macros/latex/contrib/bjfuthesis/demo/contents/abstract.tex | 14 |
1 files changed, 0 insertions, 14 deletions
diff --git a/macros/latex/contrib/bjfuthesis/demo/contents/abstract.tex b/macros/latex/contrib/bjfuthesis/demo/contents/abstract.tex deleted file mode 100644 index 71627e7e3f..0000000000 --- a/macros/latex/contrib/bjfuthesis/demo/contents/abstract.tex +++ /dev/null @@ -1,14 +0,0 @@ -\chapter*{摘要} -\begin{abstract} - 随着在线电影数量不断增加,用户选择电影的时间成本不断上升,准确的推荐算法成为了必然要求。为解决协同过滤推荐算法中的稀缺性问题与冷启动问题,研究人员用商品属性或社交网络等信息来辅助推荐算法。现有的将知识图谱作为辅助信息的推荐算法包括基于嵌入的方法和基于路径的方法,但这两种方法均存在一些缺陷,没有充分有效地利用知识图谱中的相关信息,推荐的准确度较低。 - - 本文设计并实现了基于“涟漪网络”知识图谱的推荐算法。“涟漪网络”算法的核心是利用现实生活中的雨滴产生的涟漪在水面上不断扩散的思路,来模拟用户偏好的扩散。对于每一个用户,涟漪网络将其过往偏好作为知识图谱中的一个种子集,然后沿知识图谱中的关系路径不断地拓展用户的偏好,进而发现基于等级划分的某个候选物品的潜在兴趣,其中多个“涟漪”重叠形成知识图谱中的用户偏好分布。该算法的实验结果和以往的CKE、DKN、PER等模型结果相比,表现出更优的性能。利用该算法,本文设计并实现了一个基于电影知识图谱的推荐系统,该系统包括管理员用户和普通用户,管理员能够新增、编辑和删除电影与用户,普通用户能够浏览、收藏与购买电影。该系统可以为用户提供一个高效的电影推荐功能,方便用户选择满足自己偏好的电影。 -\end{abstract} -\keywordscn{知识图谱,推荐系统,涟漪网络,用户偏好,电影商店} -\chapter*{Abstract} -\begin{abstract} - As the number of online movies continues to increase and the time cost for users to choose movies continues to rise, accurate recommendation algorithms have become an inevitable requirement. In order to address the scarcity and cold start problem of collaborative filtering, researchers usually make use of side information, such as product attributes or social networks as side information to assist the recommendation. The existing recommendation algorithms that use knowledge graph as side information include embedding-based methods and path-based methods, but both methods have some shortcomings. They do not make full and effective use of the relevant information in the knowledge graph, and the accuracy of recommendation is relatively low. - - This paper designs and implements a recommendation algorithm based on Ripple Network. The core of the Ripple Network algorithm is to use the idea that the ripples produced by raindrops in real life continue to spread on the water surface to stimulate the spread of user preferences. For each user, Ripple Network uses its past preference as a seed set in the knowledge graph, and then continuously expands the user’s preferences along the relationship path in the knowledge graph, and then discovers his hierarchical potential interests concerning a certain candidate item. Multiple ``ripples" overlap to form the user preference distribution in the knowledge graph. Compared with previous model results of CKE, DKN, PER, etc., the experimental results of this algorithm show better performance. Using this algorithm, this paper designs and implements a recommendation system based on the movie knowledge graph. The system includes administrator users and general users. The administrator can add, edit and delete movies and users, and general users can browse, collect and purchase films. The system can provide users with an efficient movie recommendation function, which is convenient for users to choose movies that suit their preferences. -\end{abstract} -\keywordsen{Knowledge graph, recommendation system, Ripple Network, user preferences, movie store} |