%0 Journal Article
%T 基于合格邻居和异常检测的社区增强协同过滤<br>Community-Augmented Cosine Collaborative Filtering Based on All Qualified Neighbors and Abnormal Detection
%A 苏湛
%A 陈惠鑫
%A 艾均
%J Modeling and Simulation
%P 2133-2146
%@ 2324-870X
%D 2024
%I Hans Publishing
%R 10.12677/mos.2024.133196
%X 为了解决协同过滤推荐算法存在较大预测误差和推荐列表准确度不高的问题，提出一种结合异常检测和网络社区并基于所有合格邻居的协同过滤推荐算法。该算法使用修改的拉依达准则检测标记数据异常，在协同过滤相似度计算阶段降低与异常用户之间的相似性权重；使用得到的用户相似性建立网络模型，利用K核分解进行网络社区检测，在得到用户间的社区信息后对社区中用户进行相似性权重处理。基于MovieLens数据集并与五种同类型算法进行对比实验，结果表明，提出的算法可以有效降低预测误差以及提升推荐列表的排序准确度。<br />
In order to address the issue of the significant prediction errors and the low accuracy of recommendation lists in the collaborative filtering recommendation algorithm, a collaborative filtering recommendation algorithm that combines anomaly detection and network community based on all qualified neighbor is proposed. The algorithm uses the modified Pauta criterion for data anomaly detection and mark, and during the collaborative filtering similarity calculation stage reduces the similarity weight with abnormal users; Builds a network model using the obtained user similarity, and uses K-core decomposition for network community detection, and processes the similarity weights for users in the community after obtaining community information between users. Compared with five recommendation algorithms of the same type, and based on the MovieLens dataset, the experimental results showed that the proposed algorithm can effectively reduce prediction error and improve the accuracy of recommendation list rankings.
%K 协同过滤，异常检测，相似性网络，社区检测，K-Core分解<br>Collaborative Filtering
%K Abnormal Detection
%K Similarity Network
%K Community Detection
%K K-Core Decomposition
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=86549