联系我们

论文查询结果	返回搜索

论文编号：	9965
作者编号：	2120162672
上传时间：	2018/6/6 13:18:17
中文题目：	基于LDA和聚类方法的高校人文社科研究机构类别划分与评价研究
英文题目：	Research on the Classification Evaluation of Humanities and Social Sciences Institutions in Universities Based on LDA and Clustering Methods
指导老师：	石鉴
中文关键字：	LDA；层次聚类；K-Means；科研机构；分类评价
英文关键字：	LDA; Hierarchical Clustering; K-Means; Scientific Research Institutions; Classification Evaluation
中文摘要：	高校人文社科研究机构是我国人文社会科学研究事业的重镇，科学评价是高校人文社科研究机构管理的重要一环。目前科研机构评价的研究多集中于评价的内容和方法，关于分类评价的研究屈指可数。然而如果只用同一个评价标准来评价不同类别的研究机构，或者划分了类别但划分标准不合理，都不能客观、真实、准确地反映出不同科研机构的实际情况。本文就试图构建一个完整的科研机构类别划分与评价的方法，即分类评价，并以教育部全国高校人文社会科学重点研究基地为评价对象进行实证分析。 要进行高校人文社科研究机构的分类评价，首先要进行的是高校人文社科研究机构的分类。由于设立科研机构的目标大多是解决现实问题，将研究问题相似的科研机构放在一起评价是最合适不过的了。然而目前的评价实践中，科研机构的分类多是根据科研机构所属的学科。本文基于LDA主题挖掘对同学术片下科研机构的研究主题进行分析，发现研究机构的学科相似并不能代表研究主题相似，也即仅仅根据科研机构所属的学科来分类是不够的。为解决这个问题，本文通过基于科研机构研究主题聚类的方式来进行科研机构的类别划分，并选取了层次聚类方法。可以看到，这种方式可以有效的将研究主题相近的科研机构划分为一类，而层次聚类方法则可以让我们清晰的看到整个聚类过程，这样也便于决策者根据实际需要选取合适的聚类数目，真正起到辅助决策的作用。 在科研机构的绩效评价方面，目前大多是简单的对各种形式的科研成果进行加权求和。然而科研机构的产出成果形式各异，且各有各的价值，这样的评价并不能体现不出研究机构的特色。本文基于K-Means聚类算法，提出了一个通用的可以从整体层面确定科研机构绩效水平的方法，并选取一类研究主题的科研机构进行了研究。发现有的科研机构更侧重产出学术成果，有的则更侧重产出实践成果，都取得了不错的成绩。当然，还有一些科研机构这两方面做的都不够好，还需要继续努力。这也启示我们，“偏科生”其实也是“特长生”，评价的方式方法太死板，就不容易发现他们的价值。
英文摘要：	Humanities and social sciences research institutions in universities are the main force of the humanities and social sciences research in China. Scientific evaluation is an important part of the management of humanities and social sciences research institutions in universities. At present, the research on the evaluation of scientific research institutions focuses on the content and methods of evaluation, and there are only a few studies on the classification evaluation. However, if we use the same evaluation criteria to evaluate different types of research institutions or classify, or the standards are unreasonable, they cannot objectively, truly and accurately reflect the actual situation of these institutions. This paper attempts to build a classification and evaluation method and take the Ministry of Education's National Humanities and Social Sciences Research Base as an example. In order to carry out the classification and evaluation of the research institutions of humanities and social sciences, the first thing to do is to classify the humanities and social sciences research institutions in universities. As the goal of establishing scientific research institutions is mostly to solve practical problems, it is most appropriate to put together research institutions with similar problems to be evaluated together. However, in the current evaluation practice, the classification of scientific research institutions is based on the discipline, so what we should do first is to check whether this classification can achieve similar results. Based on the results of LDA topic mining, this paper analyzes the research themes of similar disciplines, and finds out that the similarity of disciplines does not represent the similarity of research questions, that is to say, it is not sufficient to classify them only according to the disciplines they belong to. In order to solve this problem, the classification of scientific research institutions is carried out by the method of Clustering based on the subject of these research institutions, and the hierarchical clustering method is selected. It can be seen that this method can effectively clustering the research institutions with similar topics into one class, and with the hierarchical clustering method, the whole cluster process can be seen clearly, so that the decision-makers can select the appropriate number of clusters according to the actual needs. In the aspect of performance evaluation of scientific research institutions, most of them are simply weighted summation of various forms of scientific research achievements. However, the output of scientific research institutions is different from each other, and each has its own value. Such an evaluation cannot reflect the characteristics of research institutions. Based on the K-Means clustering algorithm, this paper proposes a general method to determine the performance level of scientific research institutions from the overall level and selects a class of research institutions to study the subject. It is found that some research institutions are more focused on producing academic achievements, others are more focused on output practice results. Of course, there are still several scientific research institutions that are not good enough and need to continue their efforts. This also enlightens us that "biased undergraduates" are also "special students". And we cannot find their value if the way of evaluation is too rigid.
查看全文：	预览  下载(下载需要进行登录)