简介
此次发布的排行榜与其他机构以往发布的排名有所不同，主要聚焦于计算机学科的多个研究领域，旨在反映现今全球计算机研究机构的客观实力，用于帮助人们了解世界计算机科研机构的学术水平及分布情况。
排名主要依据《中国计算机学会推荐国际学术会议和期刊目录》（第三版）|（http://www.ccf.org.cn/sites/ccf/paiming.jsp）中的A类会议近10年（2003－2013）所发表的全部论文数据，在区分长文（regular paper）、短文（short paper）和海报论文（poster）以及考虑论文引用数的基础上，计算的全球科研机构（包括大学、研究院/所和公司）的科研实力排名。
整个排名趋于简单、明了、客观、公平，人们可以很直观的看出研究机构的成果及其近几年的学术影响。
数据来源
这些数据全部是来自于世界权威数据机构的第三方数据，包括：ACM Digital Library、DBLP、Google Scholar等机构，再经由我课题组经过集成和分析而生成。我们在排名中使用的基础数据包括发表论文数（区分长文、短文和海报论文）、论文引用数等。
算法
分数计算综合考虑了论文引用数、作者数、第几作者等因素，忽略短论文和海报论文数据
对于一个具体的机构A，定义集合P为至少有一个作者的机构信息为机构A的所有论文集合，则其各项分数计算方法如下：
- $\#Paper(A)$ : P中的所有论文数
- $\#FirstAuthor(A)$: 第一作者为机构$A$的所有论文数
- $Credit$: 定义 $af(p,k)$为P中特定论文$p$的第$i$作者的机构， $C(n)$为 $$C(n)=\sum_{k=1}^n1\big(af(p,k)=A\big)*\frac{\frac{1}{k}}{\sum_{i=1}^n\frac{1}{i}}$$ span 则$Credit$为： $$Credit=\sum_{p\,in\,P}C(n)\;where\,p\,has\,n\,author(s)$$
- $\#Citation(A)$: P中所有论文引用次数求和 $$\#Citation=\sum_{p\,in\,P}NCitation(p)$$
- $Weighted\_Credit(A)$: 对于$P$中的每篇论文$p$ $$Weighted\_Credit(A)=\sum_{p\,in\,P}C(n)*NCitation(p)\;其中n为p的作者数目$$
- $AM-Score$: 一般来说年代久远的论文引用数越大，所以我们把每年的$Weighted\_Credit(A)$用当年所有有效引用数的均值进行归一化，于是对于$P$中的每篇论文$p$ $$AMScore=\sum_{y\,in\,years}\frac{\sum_{p\,in\,P}C(n)*NCitation(p)}{avg(nCitations(y))}\;其中n为p的作者数目$$
Brief Introduction
The purpose of this ranking is to determine scientific impact of organizations from big data at AMiner, on Internet, and possible from other sources. Our primary goal is to rank academic organizations according to bibliometric data using various metrics.
Data Source
We use all papers published at thirty-nine CCF A Class conferences (Cf.http://www.ccf.org.cn/sites/ccf/paiming.jsp）in recent ten years (2003-2013). For each paper, we extract organizations from the authors’ affiliation information and also extract citation information from Google Scholar for each paper.
Metrics
We report the rank according to five metrics: paper number, first-authored paper number, scientific credit, citation number, AM-score. We distinguish different type of paper such as regular paper, short paper and poster. Basically, in this rank, we only consider long paper published at those conferences.
Below, we provide details on how we calculate the different scores. Let $P$ denote the set of papers published by an organization $A$.
- $\#Paper(A)$ : number of papers published by each organization $$\#paper=\|P\|$$
- $\#FirstAuthor(A)$ : number of papers with first author associated with the corresponding organization. Thus we have $\#FirstAuthor(A)\leq \#Paper(A)$
- $Credit$: considers the author order for each paper. The basic idea is that the first author is credited more and the following authors will be credited with gradually decreased scores. $$Credit=\sum_{p\,in\,P}C(n)\;where\,p\,has\,n\,author(s)$$ $$C(n)=\sum_{k=1}^n1\big(af(p,k)=A\big)*\frac{\frac{1}{k}}{\sum_{i=1}^n\frac{1}{i}}$$ where $1(af(p, k)=A)$ is an indicator function which returns 1, if the affiliation of paper p’s $k^{th}$ author is $A$; 0 otherwise; $C(n)$ is the sum of normalized credits of all authors associated with organization $A$. If all authors are associated with the same organization $A$, then the score is 1. Finally Credit of organization $A$ in this rank is defined as the sum of credit of all papers associated with the organization.
- $\#Citation(A)$: sum of all the citation number of papers in $P$ $$\#Citation=\sum_{p\,in\,P}NCitation(p)$$
- $Weighted\_Credit(A)$: the difference of this metric from the #Citation(A) is that we consider the credit score of the organization for each paper. $$Weighted\_Credit(A)=\sum_{p\,in\,P}C(n)*NCitation(p)\;where\,p\,has\,n\,author(s)$$
- $AM-Score$: In general, the older paper trends to obtain more citation numbers. Thus, we normalize the $Weighted\_Credit(A)$ above for each year by the avarage of all valid citation numbers that year. $$AMScore=\sum_{y\,in\,years}\frac{\sum_{p\,in\,P}C(n)*NCitation(p)}{avg(nCitations(y))}\;where\,p\,has\,n\,author(s)$$ $$AMScore=\sum_{y\,in\,years}\frac{\sum_{p\,in\,P}C(n)*NCitation(p)}{avg(nCitations(y))}\;where\,p\,has\,n\,author(s)$$