摘要
为了解决期刊多属性评价中评价方法众多，评价结果不唯一问题，本文提出了一种新的多属性评价方法筛选方法——指标数据综合拟合法。其原理是首先提出两个标准：第一是评价结果数据分布必须遵循评价指标；第二是任何评价方法评价结果必须和评价指标较好拟合。关于数据分布，在评价方法权重和模拟权重的基础上计算评价结果的理论偏度、理论峰度，再用实际评价结果的偏度和峰度除以理论结果，得到数据分布贴近度；关于数据拟合，则基于评价指标与评价结果的关系采用拟合优度R^2确定。最后结合数据分布贴近度和拟合优度R^2计算指标数据综合拟合度，用它作为筛选评价方法的依据。本文以熵权法、TOPSIS、主成分分析、因子分析4种评价方法为例，选取JCR光学期刊2011年数据为例进行实证。结果表明，指标数据综合拟合法是一种有效的多属性评价方法筛选方法，具有较广的应用前景，线性评价类方法具有一定的筛选优势，该方法也可以广泛应用到其他领域。
In order to solve the problem which evaluation methods for journal of multi-attribute evaluation are numerous, and the evaluation results are not unique, a new screening method called comprehensive fitting method of indicator data ,which is for multi-attribute evaluation methods, is proposed in this paper. Its principle gives two axioms at the beginning： the first one is that evaluation result data distribution must follow evaluation indicators, and any evaluation result must be fit in with evaluation indicator is the second one. Then calculate theoretical skewness and theoretical kurtosis on the basis of evaluation weights and analog weights. Next, using the skewness and kurtosis of actual evaluation results divided by theoretical results to get the close degree of data distribution. Finally the close degree of data distribution and goodness of fit are combined to compute the comprehensive fitting degree of indicator data, which is used as the basis to select the best evaluation method. In the end, illustrated by the examples of four evaluation methods included entropy weight method, TOPSIS, PCA and factor analysis with data of JCR optical periodicals in 2011 for demonstration. The results indicate that comprehensive fitting method of indicator data is an effective screening method for multi-attribute evaluation methods and has expansive application prospect. Linear evaluation method possesses certain screening advantages. It can also be applied to other areas widely.
出处
《情报学报》
CSSCI
北大核心
2014年第3期296-304,共9页
Journal of the China Society for Scientific and Technical Information
基金
宁波大学区域经济与社会发展研究院重点课题(QYJYND1301)
关键词
多属性评价
方法筛选
指标数据综合拟合度
光学期刊
multi-attribute evaluation, methods selection, comprehensive fitting degree of indicator data, optical periodicals
