基于用户阅读行为的图书自动评测算法

User behavior based automatic book rating algorithm

针对图书质量较难量化的问题,以用户在阅读过程中的流失和留存数据为基础,对各个关键行为节点设置评测权重,通过用户阅读图书的深度来量化图书质量.在分析在线阅读平台特点的基础上,根据用户阅读图书深度到达的难度来量化用户的阅读行为,提出基于用户阅读行为的图书自动评测算法.图书的最终综合评测值考虑图书半衰期和用户分群对图书评分的影响,能够根据图书的多个评测指标对图书进行自动评测,帮助用户选择更符合自身喜好的图书,提高用户满意度和用户体验.

In order to tackle with the difficulty of book quality quantification, evaluation weight was set for each critical behavior node based on user＇s churn/retention data in the reading process, and an automatic book rating method was presented based on users＇ reading behavior. User segmentation was considered due to the difference of users＇ behavior before and after the half-life of books. The daily score of each book was used to fine-tune the comprehensive score of books in order to differentiate the preference of user segmentations. The effectiveness of the algorithm was tested on a mobile reading platform which has tens of millions of readers. Results show that multiple indexes can automatically model the quality of books and help the users to find the book they like.