数据挖掘在图书馆读者忠诚度研究中的应用

Application of Data Mining in Library's Reader Loyalty Research

[目的/意义]分析数据挖掘在读者研究中应用的国内外现状及发展趋势,探讨图书馆读者忠诚度特点及读者流失预测问题。[方法/过程]设计一种新的基于后验概率的支持向量机模型(PPSVM),将读者数据的后验概率引入到支持向量机中来预测读者流失情况,并进行测试。[结果/结论]改进后的PPSVM模型能够避免类别模糊的样本对分类器的影响,获得比传统支持向量机更高的分类准确率,同时对于非确定性分类的问题有更好的稳健性;高校图书馆通过读者流失预测可以更好地对读者进行个性化服务,提高图书馆服务的能动性,因此模型具有很强的实用性。

[Purpose/significance]The paper is to analyze status quo and development trend of application of data mining in reader research at home and abroad, and to discuss characteristics of library＇s reader loyalty and the issue of reader loss prediction.[Method/process] The paper designs a new posterior probability-based SVM model（PPSVM）, which introduces posterior probability of readers＇ data into support vector machine to predict reader loss, and carry out a test. [Result/conclusion]The improved PPSVM can avoid the influence of the samples with fuzzy categories to classifier, to obtain higher classification accuracy than traditional support vector machine, while for non-deterministic classification problems, PPSVM model has better robustness; university library can better provide personalized service for readers through reader loss prediction, so as to improve the initiative of library service. Therefore the model is practicable.