情报科学 ›› 2022, Vol. 40 ›› Issue (11): 133-138.

• 业务研究 • 上一篇    下一篇

基于时间序列算法的高校图书馆借阅数据预测及分析 

  

  • 出版日期:2022-11-01 发布日期:2022-12-13

  • Online:2022-11-01 Published:2022-12-13

摘要: 【目的/意义】图书借阅数据的预测对于图书馆的资源建设和精准服务具有重要的指导意义。本文收集了
中国东北地区某双一流高校图书馆管理系统十年的借阅数据,并分别按图书类别、借阅者所属学院分类,对未来的
借阅趋势进行了预测。【方法
/过程】本文使用一种基于时间序列的混合预测模型进行图书借阅数量的预测,其中混
合预测是一元时间序列预测与多元时间序列预测的结合。【结果
/结论】实验结果表明,时间序列算法用于高校图书
馆借阅数据预测,
2008 年到 2017 年借书数量由 300 左右增加到近 40002018 年到 2021 CDGJSUI类图书
中,
D类、S类图书的借阅数量下降幅度最大,U类图书借阅数量下降幅度最小,T类、E类借阅量的上升幅度最大,Q
类、X类的误差率较大,研究结论供高校图书馆管理工作参考。【创新/局限】学界上针对图书馆馆藏资源建设和服务
创新研究较多,但以一元时间序列与多元时间序列预测角度进行研究的相对较少,本文弥补了此方面的不足。

Abstract: Purpose/significanceThe prediction of book borrowing data has an important guiding significance for library resource construction and accurate service. This paper collects the ten years' borrowing data of the library management system of a double firstclass university in Northeast China, and forecasts the borrowing trend in the future according to the categories of books and the col⁃leges to which the borrower belongs.Method/processThis paper uses a hybrid prediction model based on time series to predict the number of books borrowed, in which the hybrid prediction is the combination of monistic time series prediction and multivariate time series prediction.Results/conclusionThe experimental results show that the time series algorithm is used to predict the borrowing
data of university libraries. From 2008 to 2017, the number of books borrowed has increased from about 300 to nearly 4000. From 2018 to 2021, the number of books borrowed by category D and S has decreased the most, the number of books borrowed by category U has decreased the least, the number of books borrowed by category T and E has increased the most, and the error rate of category Q and X is large, The research conclusion can be used as a reference for the management of university libraries.
Innovation/limitationThere are many researches on the construction of library collection resources and service innovation in the academic circle, but there are relatively few researches from the perspective of monistic time series and multivariate time series prediction. This paper makes up
for the deficiencies in this regard.