基于用户隐性反馈与协同过滤相结合的电子书籍推荐服务

Recommending Books Based on the Combination of Implicit Feedback and Collaborative Filtering

随着电子设备的普及,越来越多的人更愿意在他们的手机或者平板上阅读电子书籍.推荐服务的出现是为了从海量电子书籍中找到符合读者兴趣的书籍,其中协同过滤(Collaborative Filtering,CF)作为推荐系统的主流方法,也被应用在书籍推荐服务中.传统基于CF的书籍推荐在解决用户显性评分缺失问题时,仅考虑了用户对书籍的喜好程度与阅读时长和阅读频次等隐性反馈内容有关,忽略了在阅读书籍时不同用户间阅读速度可能存在差异.从阅读速度出发展开研究,提出阅读速度感知模型(Reading Speed-aware Model,RSA)和书籍阅读权重模型(Reading Book-weight Model,RBW),把用户的阅读时长转换为阅读速度,最后结合上述两个模型提出一个混合的速度-权重模型(Speed-Weight Model),将用户的隐性反馈转换为喜好程度的评分来补全CF评分矩阵.通过对现有方法的实验对比分析,本文所提方法能够在一定程度上提高书籍推荐的准确度.

With the popularization of electronic devices, an increasing number of people prefer to read E-books on their phones and pads. The advent of the recommendation system helps people to find their interesting books easily from the massive E-books. Collaborative Filtering （ CF ）, as a main technique used in the recommendation system, could also be applied in book recommendation. To deal with the lack of readers＇ rating on books,previous CF based method only takes into account the implicit behavior of user reading time and frequencies that can capture the readers＇ interests to a book. They neglect the fact that different users may have different speed for reading ,which can be further used for improving the recommendation accuracy. In this paper, starting from the user reading speed, we propose Reading Speed-aware Model （ RSA } and Reading Book-weight Model （ RBW ）, which convert user reading time to reading speed. Finally,combining with the two models, we propose a Speed-Weight Model which transforms readers＇ implicit feedback to readers＇ rating and completes the CF rating matrix. Through the experimental evaluations, our model outperforms the baseline algorithm and improves the accuracy of the book recommendation.