神经网络与Logistic回归在基础研究成果综合评价中的应用比较  

Comparative Study of Artificial Neural Network and Logistic Regression in Evaluating Basic Research Achievement

作　　者：何钦成[1] 黄亚明[1] 何苗[2] 

机构地区：[1]中国医科大学信息管理与信息系统(医学)系 [2]中国医科大学附属第一医院计算机室,辽宁沈阳110001

出　　处：《科技进步与对策》2005年第9期102-104,共3页Science & Technology Progress and Policy

基　　金：辽宁省教育厅高等学校科学研究项目计划(20124346)

摘　　要：应用现代数理统计、系统综合评价及科学计量学方法构建对基础性研究成果的综合评价系统。在以科学计量学和数理统计方法处理综合评价指标集输入数据的基础上,通过Logistic回归和神经网络(网络类型为多层感知器,训练算法采用共轭梯度下降法)实现对基础研究成果多指标的综合评价并比较二法的评价效果。To construct a comprehensive ev which multiple indexes should be comprise aluating system of basic scientific and technologic achievements in d. The input and output data were dealt with methods of scientometrics, mathematics and statistics. And the systemic differentiation was implemented with Logistic Regression and artificial neural network（Muhilayer Perceptrons as network type and Conjugate Gradient Decent as training algorithm）.The results of evaluation with the two methods were compared and analyzed.

关 键 词：神经网络 共轭梯度下降法 LOGISTIC回归 基础研究 成果评价 

分 类 号：G311[文化科学]