基于GAIFWFCM和TFNs的协同过滤算法  

COLLABORATIVE FILTERING HYBRID RECOMMENDATION ALGORITHM INCORPORATING GAIFWFCM AND TFNS

作　　者：金玉 徐新卫[1] 陶飞 韩业 陈荣凯 JIN Yu;XU Xin-wei;TAO Fei;HAN Ye;CHEN Rong-kai(College of Management Science and Engineering,Anhui University of Technology,Ma′anshan 243032,China;Engineering Research Institute,Anhui University of Technology,Ma′anshan 243032,China)

机构地区：[1]安徽工业大学管理科学与工程学院,安徽马鞍山243032 [2]安徽工业大学工程研究院,安徽马鞍山243032

出　　处：《南阳理工学院学报》2024年第4期49-57,共9页Journal of Nanyang Institute of Technology

基　　金：安徽省自然科学基金项目(2108085MF217);安徽省教育厅高校重点基金项目(KJ2021A0374)。

摘　　要：针对传统推荐系统中使用离散评分未能合理表达用户偏好的问题,将遗传算法引入优化特征加权模糊C均值,通过梯形模糊聚合相似目标用户,提出基于遗传算法的优化加权模糊C均值聚类融合梯形模糊数的协同过滤模型。通过遗传算法进行增强初始聚类中心,利用优化加权模糊C均值聚类融合梯形模糊数,分析类内与类间属性特征关系,实现用户精细划分,合理过滤推荐。在两组数据集中以MAE和RMSE作为评估指标进行实验,实验结果表明,所提算法在与其余6种算法对比中分类误差更低,对用户意愿识别更加清晰。In response to the issue of traditional recommendation systems′inability to effectively capture user preferences using discrete ratings,this study introduces a genetic algorithm to optimize feature-weighted fuzzy C-means.By employing trapezoidal fuzzy aggregation to cluster similar target users,a collaborative filtering model based on the genetic algorithm optimized weighted fuzzy C-means clustering and trapezoidal fuzzy numbers is proposed.The genetic algorithm enhances initial clustering centers and utilizes weighted fuzzy C-means clustering and trapezoidal fuzzy numbers for fusing intra-class and inter-class attribute features,enabling fine-grained user segmentation and rational recommendation filtering.Experimental evaluations on two datasets utilizing MAE and RMSE as evaluation metrics demonstrate that the proposed algorithm exhibits lower classification errors compared to six other algorithms while providing clearer identification of user preferences.

关 键 词：协同过滤 梯形模糊数 模糊C均值 遗传算法 特征加权 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]