【摘要】 作者合著关系的预测对于提高科研合作效率和有效的科研管理具有重要的意义。本文以中国知网中图书情报领域核心期刊作为信息来源,获取15年(2001-2015)的文献信息。通过计算指标方差和指标性质确定对合著关系预测的指标体系,同时对比基于单指标的无监督方法和基于分类算法的监督式机器学习方法 (逻辑回归、支持向量机和随机森林)的预测效果,本文最终确定基于随机森林和指标体系所构造的合著关系预测模型。通过实例应用证明该模型具有较好的准确性和稳定性。更多还原

【Abstract】 The prediction of co-authorship is of great significance to improve scientific research cooperation efficiency and manage scientific research more effectively. Using CNKI as the data resource,this paper selected co-authorship in the core journals between 2001 and 2015. The co-authorship prediction index system was determined by index properties and index variance. By comparing the prediction effect of the unsupervised method based on single index and supervised machine learning method based on classification algorithm which contained logistic regression,support vector machines and random forests,this paper finally confirmed the prediction model of the relationship based on the index system and random forests.更多还原