基于用户偏好的文献推荐系统  

The literature recommendation system based on users’ preference

作　　者：刘韵毅[1] 梁樑[1] 

机构地区：[1]中国科学技术大学管理学院,安徽合肥230026

出　　处：《情报理论与实践》2007年第1期61-63,25,共4页Information Studies:Theory & Application

基　　金：国家自然科学基金(项目号:70371023);国家教育部博士点基金(项目号:20030358052);国家杰出青年科学基金(项目号:70525001)资助项目研究成果。

摘　　要：本系统基于模糊联想记忆神经网络,建立偏好评价模型,根据用户偏好对搜索引擎搜索到的候选文献进行评级,为用户推荐偏好值高的文献。本系统的学习模块采用PCA-CG算法和误差反向传播算法,以用户阅读过的基准文献和其对应评级作为训练样本,对用户偏好进行学习;推理模块根据学习到的模糊规则和隶属函数来计算候选文献的偏好值,并以偏好值排序,把偏好值高的文献推荐给用户。把该模型应用于信息技术类文献的检索,实验表明系统提供的推荐文献具有较高可信度。A preference evaluation model is constructed based on Fuzzy Associative Memory Neural Network. It＇ s proposed to recommend literatures with high preference values to users among literatures retrieved by search engines and ranked according to user＇ s preference. The system＇ s learning module uses PCA - CG algorithm and error hack propagation algorithm to learn user＇ s preference with the reference literatures read by users and their corresponding appraisal as training samples. The reasoning module computes the preference values according to fuzzy rules and membership functions, ranks the result from high preference to low preference and recommends the literatures with high preference to users. The model is applied in the retrieval of literatures of information technology category, and the experimental result proves that the literatures recommended by the system is reliable.

关 键 词：模糊联想记忆 神经网络 隶属函数 文献推荐 

分 类 号：G252[文化科学—图书馆学]