高校教师科研能力评估的贝叶斯网络聚类方法  

The Bayesian Network Clustering Method for Research Capacity Assessment of University Teachers

作　　者：李兰春[1] 王双成[2] 王婧[3] 

机构地区：[1]上海立信会计学院外语学院,上海201620 [2]上海立信会计学院数学与信息学院,上海201620 [3]上海立信会计学院人事处,上海201620

出　　处：《科技管理研究》2011年第12期114-116,共3页Science and Technology Management Research

基　　金：国家自然科学基金项目(60675036);上海市教委科研创新重点项目(09zz202)

摘　　要：针对高校教师科研能力特点和现有评估方法存在的问题,建立高校教师科研能力评估的朴素贝叶斯网络聚类方法,这种方法不需要许多例子,甚至在没有类标签的情况下也能够进行规则提炼和预测。为提高聚类预测的可靠性,在高斯核函数中引入形状参数,并通过形状参数的优化来提高聚类的泛化能力。实验结果显示,经过优化的朴素贝叶斯网络聚类能够有效提高预测的准确性。A naive Bayesian network clustering method is developed for research capability assessment of university teachers based on the features and the existing problems in assessment methods.This method does not need many examples.Even if no example,it can also extract rules and do prediction.For improving prediction reliability,a shape parameter is put into Gaussian kernel function.And clustering generalization capacity can be improved by optimizing the shape parameter.Experimental results show that optimized clustering model is more accurate for prediction.

关 键 词：科研能力评估 指标体系 贝叶斯网络 聚类 

分 类 号：G645.1[文化科学—高等教育学]