基于数据挖掘的贫困助学金认定方法研究

发布时间：2018-03-14 16:29

本文选题：贫困生　切入点：精准资助　出处：《华中师范大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着高等院校的不断扩招,其中一类特殊群体也在不断扩大——贫困生。贫困生群体成为国家重视、社会关注、学校关心的社会问题。教育公平是社会公平的基础,为促进教育公平,国家、高校及社会各界共同努力研究,建立了较为完善的学生资助政策体系,从多方面保障每个学生顺利入学、顺利完成学业。然而在贫困生认定、助学金发放上仍存在着不足:助学金未能覆盖全部贫困生、非贫困生却受到国家资助。社会各界不断为精准资助献力献策,旨在达到国家精准资助的目标。本文运用统计学方法及数据挖掘理论,结合某高校一卡通及贫困助学金发放数据,从大数据的角度研究贫困助学金的认定方法:(1)采用箱线图分析该校贫困生分布情况,粗略判断该校在精准资助贫困生上还有完善的空间;(2)通过单因素方差分析及列联表分析,得到了以下结论:学生消费能力与学生学业成绩是影响贫困生认定的因素之一,认为资助贫困生不仅要资助家庭经济困难的生存型困难学生也要资助没有能力支付其学习发展费用的发展型困难学生;(3)结合K-means聚类模型将学生根据其各消费方式的消费金额聚类,比较得到消费最低人群占总人数比例为25.07%,比较符合我国贫困生比例,然而该校的资助比例却未达到此,由此构建贫困生预测模型,以进一步帮助学校、社会及政府做好贫困生精准资助工作,保证贫困生们顺利完成学业。本文选取三种分类预测模型:Logistic回归模型、Naive Bayes算法模型、K近邻模型,结合三种分类模型评价标准(查全率、查准率及F1值),经综合比较发现K近邻模型能更好的判别出学生是否是贫困生,该模型的准确率达75.28%,查全率达85.35%。
[Abstract]:With the increasing enrollment of colleges and universities, one of the special groups is also expanding the poor students. The poor students have become the social problems that the state attaches importance to, the society pays close attention to, and the schools are concerned about the social problems. Education equity is the foundation of social equity. In order to promote educational fairness, the country, colleges and universities and all walks of life have worked together to establish a more perfect policy system for student support, which in many ways ensures the smooth admission of each student and the smooth completion of his studies. However, in the case of poor students, There are still deficiencies in the distribution of grants: grants do not cover all poor students, but non-poor students are funded by the state. The purpose of this paper is to achieve the goal of national precision funding. This paper applies the statistical method and data mining theory to combine the data of one card and poverty grant in a certain university. From big data's point of view, we study the identification method of poverty grant. (1) using box diagram to analyze the distribution of poor students in this school, roughly judging that there is a perfect space for the poor students in this school to subsidize the poor students with precision.) through single factor variance analysis and column table analysis, The following conclusions are drawn: students' consumption ability and students' academic achievement are one of the factors that influence the identification of poor students. It is believed that financial aid for poor students should not only subsidize students with financial difficulties in their families, but also those students with developmental difficulties who cannot afford to pay for their study and development.) combining K-means clustering model, the students can be used according to their different consumption patterns. The amount of money spent is clustered, The comparison shows that the proportion of the lowest consuming population to the total population is 25.07, which is more in line with the proportion of poor students in China, but the proportion of financial aid in this school has not reached this level. Therefore, a forecast model for poor students is constructed to further help the school. In order to ensure the poor students to finish their studies successfully, the society and the government have done a good job of providing accurate financial assistance to poor students. This paper selects three kinds of classification prediction models: logistic regression model and naive Bayes algorithm model as well as K nearest neighbor model, and combines the evaluation criteria of three classification models (recall rate, recall rate). By comprehensive comparison, it was found that K-nearest neighbor model could better judge whether the students were poor students. The accuracy of the model was 75.28 and the recall rate was 85.35.
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：G647

【参考文献】