基于高考录取成绩的院校竞争网络实证研究

平行志愿录取规则与优质生源的紧缺使得院校在高考招生时存在激烈的竞争关系。从网络科学视角正确理解院校竞争关系，能够合理地指导考生进行志愿填报。利用2019年山西省高考理科录取成绩，通过改进的Jaccard相似度计算方法，构建院校竞争网络的邻接矩阵。基于复杂网络度量方法，实证研究院校竞争网络得到:1) 度分布存在一定的异质性，且具有较大的平均度；2) 簇度负相关表明院校竞争网络存在层次结构；3) 富人俱乐部连通性表明度数大于375的院校完全连通构成富人俱乐部；4) 利用基于节点标签与网络结构的社团划分算法，通过基尼不纯度评估社团内院校的同质性，进一步利用Louvain算法识别社团内院校存在层次化差异结构。这些实证结果刻画了院校竞争关系，在此基础上归纳了一些指导考生进行合理、分梯度志愿填报的建议和策略。

Empirical Research on College Competition Network Based on the Admission Scores of Colleges in Chinese Gaokao

There is a very fierce competitive relationship among colleges because of the parallel application and the shortage of high-equality students in Chinese Gaokao. From the perspective of network science, correctly understanding such competitive relationship is helpful to guide candidates to apply for admission. This paper uses the 2019-year colleges’ admission scores in Shanxi Province to construct the adjacency matrix of the competition network via the improved Jaccard similarity calculation method. Based on the measurements of complex network, the empirical research on the competition network shows that 1) there is the proper heterogeneity in the degree distribution and the larger average degree; 2) the negative correlation between the clustering coefficients and degrees of nodes suggests that the competitive network has a hierarchical structure; 3) the connectivity of rich club suggests that the colleges with a degree greater than 375 are fully connected and constitute the rich club; 4) using the community detection algorithm based on the colleges’ metadata and network structure, the homogeneity of colleges in the community is evaluated through Gini Impurity. On this basis, The Louvain algorithm is further used to identify the hierarchical structure of colleges in the community. These empirical results clearly portray the competitive relationship among colleges, which enables us to bring great enlightenment for guiding candidates to apply for admission.