基于用户偏好与商品属性情感匹配的图书个性化推荐研究

发布时间：2018-11-07 17:57

【摘要】：【目的】识别并获取细粒度的用户偏好信息,优化图书个性化推荐的效果。【方法】使用情感分析方法对用户图书评论进行属性层文本挖掘,通过用户本身的图书评论获取用户对图书属性的偏好;基于每本图书的所有评论的情感计算获得其属性评分;将用户偏好矩阵、图书属性得分矩阵进行匹配,从而实现用户对图书属性情感偏好的个性化推荐。【结果】利用亚马逊图书评论数据作为数据来源分别对传统的协同过滤方法与本文提出的推荐方法进行实验对比。结果表明,本文提出的方法在准确性、召回率、覆盖率上分别提高了0.030、0.097、0.2812。【局限】未考虑时间因素对用户偏好的影响,并且属性类型的全面程度受亚马逊图书评论数量和质量的限制。【结论】本文计算用户对图书属性的情感得分,得到细粒度的用户偏好信息,并通过与图书属性的得分进行匹配,提升了图书个性化推荐的效果。
[Abstract]:[objective] to identify and obtain fine-grained user preference information and optimize the effect of personalized book recommendation. The user's preference for the book attribute is obtained through the user's own book review. The attribute score was obtained based on the emotional calculation of all the comments in each book; Matching user preference matrix, book attribute score matrix, [results] the traditional collaborative filtering method is used as the data source to compare the traditional collaborative filtering method with the recommendation method proposed in this paper. [results] using Amazon book review data as the data source, we can make a comparison between the traditional collaborative filtering method and the recommendation method proposed in this paper. The results show that the accuracy, recall rate and coverage rate of the proposed method are increased by 0.030 / 0.097 / 0.2812 respectively. [limitation] the influence of time factors on user preference is not considered. And the comprehensive degree of attribute type is limited by the quantity and quality of Amazon book review. [conclusion] this paper calculates the user's emotion score to the book attribute, and obtains the fine granularity user preference information. And by matching with the score of book attributes, the effect of personalized book recommendation is improved.
【作者单位】：华中师范大学信息管理学院;华中师范大学青少年网络心理与行为教育部重点实验室;
【基金】：国家自然科学基金项目“基于用户偏好感知的Saa S服务选择优化研究”(项目编号:71271099),国家自然科学基金项目“基于屏幕视觉热区的网络用户偏好提取及交互式个性化推荐研究”(项目编号:71571084)的研究成果之一
【分类号】：TP391.3

【相似文献】