论文部分内容阅读
针对高校图书馆新读者、新图书因没有借阅历史记录导致的冷启动问题,借鉴电子商务推荐系统冷启动问题的一些解决办法,提出利用改进的K-medoids分别对已有读者、已有图书进行基于借还时间间隔的聚类,根据已有读者和已有图书的特征分别建立决策树分类模型。然后,通过目标新读者、新图书的特征分别匹配对应的已有读者类、已有图书类。依据所归属的类内借阅偏好情况,向目标新读者推荐其可能喜好并且在架的图书,并把目标新图书推荐给相应的读者。在高校图书馆借阅记录数据集上,提出的方法得到有效的验证。
For the new readers of the university library, the new books because of the lack of history of borrowing cold start problems, draw lessons from some solutions to the cold start problem of e-commerce recommendation system, put forward the use of improved K-medoids, respectively, existing readers, existing books Based on the cluster of borrowing and repaying time interval, the decision tree classification model is established according to the features of the existing readers and the existing books. Then, through the target new reader, the characteristics of the new book match the corresponding existing reader categories and the existing book categories, respectively. Depending on the preferred in-class lending preferences, recommend to target new readers their likely favorite and on-shelf books, and refer new target books to their readers. The method proposed in the lending record dataset of university library is validated.