基于情感分析和情感遗忘的协同过滤推荐策略  

A Collaborative Filtering Recommendation Strategy Based on Emotional Analysis and Forgotten Rule

作　　者：卢竹兵 马小琴 吴汶娟 李玉州 LU Zhubing;MA Xiaoqin;WU Wenjuan;LI Yuzhou(College of Economy Management,Southwest University,Chongqing 400715;School of Big Data and Artificial Intelligence,Chizhou University,Chizhou Anhui 247000;Library of Southwest University,Chongqing 400715;College of Computer Science and Information,Southwest University,Chongqing 400715,China)

机构地区：[1]西南大学经济管理学院,重庆400715 [2]池州学院大数据与人工智能学院,安徽池州247000 [3]西南大学图书馆,重庆400715 [4]西南大学计算机与信息科学学院,重庆400715

出　　处：《重庆师范大学学报（自然科学版）》2020年第5期103-108,共6页Journal of Chongqing Normal University:Natural Science

基　　金：中央高校基本业务费项目基金(No.XDJK2017C087);西南大学科研基金项目(No.SWU1908036);安徽省高校优秀青年人才支持计划项目(No.gxyq2019110);安徽省高校优秀青年骨干教师国内访问研修项目(No.gnfx2020113)。

摘　　要：【目的】探索弥补现有协同过滤推荐系统中存在的诸如无法有效地识别用户兴趣的迁移、数据稀疏情况下推荐准确率低等缺陷的方法。【方法】通过对网络用户在线评论信息进行基于方面级的情感分析,利用改进的DP算法提取出评论信息中用户潜在的情感倾向,并对它进行有效地量化,有效地将用户的情感因素引入到用户兴趣建模过程中。同时,引入艾宾浩斯遗忘曲线理论,解决因时间变化而导致的用户兴趣迁移的问题。【结果】模拟实验在所选的两套数据集上进行,分别针对平均绝对误差(MAE)和覆盖率(Coverage)两个常用的性能评价标准,与选定的对比算法进行了对比实验。实验结果显示提出的引入情感分析和遗忘的协同过滤推荐算法能够有效地降低MAE,并能有效地提升Coverage。【结论】提出的算法有效地弥补了兴趣迁移对推荐准确率的影响,提高了系统对商品长尾的发掘能力。[Purposes]In traditional collaborative filtering recommendation system,drawbacks such as interest in migration and sparse data always exist,which has weakened performance of the system.[Methods]Through analyzing the online comment information of users emotions,and extracting the emotional tendency of the users,some emotional information from their comments can be mined.The emotional information is then get quantified,which is combined in interest modeling process.At the same time,the Ebbinghaus forgetting curve was also introduced to overcome the problem that caused by the change of interest,which can effectively compensate the influence of interest transfer to recommendation accuracy.[Findings]The simulation experiment was carried out on the selected data set,and the MAE and coverage were compared with the corresponding algorithm respectively.The data showed that the collaborative filtering recommendation strategy of emotional analysis could reduce MAE and improve coverage effectively.[Conclusions]The proposed algorithm effectively compensates for the influence of interest migration on recommendation accuracy and improve the system’s ability to explore the long tail of commodities.

关 键 词：推荐系统 协同过滤 兴趣迁移 数据稀疏 遗忘曲线 情感分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]