摘要
目前电子商务推荐系统中存在客户评分稀疏性问题,隐性评分是解决该问题的有效方法,但现有方法只考虑客户对单个网页的兴趣度.客户浏览行为和网页之间语义相似度可以综合客观反映客户对商品隐性的评分.建立网页商品概念-属性矩阵CA(Goods Concept-Attribute Matrix),综合考虑商品属性、相关商品及其在网页中的分布等因素,基于客户浏览路径和时间的统计分析,通过语义词林WordNet计算网页语义相似度,综合计算商品语义隐性评分.通过算例和实证研究说明该算法的有效性.
On the basis of analyzing the recommendation system of e-commerce and algorithm existing problem at present, recommend algorithms, and put forward a kind of accurate implicit Rating approach based on Web Pages Semantic Similarity. By exploiting goods knowledge distributed in different Web pages, a CA( Goods Concpet-Attribute Matrix) model was structured which combines goods attributes, related goods and those distribution. Using thesaurus (e.g., WordNet) to calculate the Semantic Similarity of Web pages of the user's current access path, a semanticsbased implicit rating model was presented in this paper. An example of calculating process was showed. Experimental evaluation demonstrated this approach was more accurate than traditional Implicit Rating algorithms.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2006年第11期98-102,共5页
Systems Engineering-Theory & Practice
基金
国家自然科学基金(70572079)
湖北省自然科学基金(2006ABA303)
关键词
隐性评分
语义相似度
商品属性
协同过滤
电子商务
implicit rating
semantic similarity
goods attributes
collaborative filtering
E-commerce