基于离群检测的学生学习状态分析
发布时间：2018-07-23 15:58
【摘要】：教育数据挖掘在我国目前尚处在理论探索阶段,大部分研究都是理论描述和可行性分析,应用研究很少。本文立足于所研究课题,从教育数据挖掘的应用研究入手,探索数据挖掘技术在学生学习状态分析中的应用。在我国高校存在学生工作者与学生比例失衡的问题,导致教育管理资源得不到有效利用,本文旨在解决一个实际的问题:给哪些学生分配稀缺的个性化学习指导的教育管理资源,以助其顺利升学毕业,让有限的教育管理资源发挥更大的价值。本文使用基于离群点检测的学生学习状态分析找出学习状态异常的学生,为教育管理资源的分配提供依据。为获取离群学生,针对本问题设计了混合两阶段离群点检测算法,首先使用基于密度的局部离群点检测算法计算出每名学生的局部离群点因子,再使用基于统计的离群点检测算法进行二元判断以获取离群学生,该算法获取的离群学生中绝大部分是学习状态异常的学生,但学习状态异常的学生的总数偏少。针对该算法存在的不足从多方面进行改进,从数据角度,添加人工属性扩大知识库;从算法角度,使用缩减迭代对不断减小的数据集进行迭代离群点检测;最后将添加人工属性和缩减迭代融合形成一种综合的改进。在本问题中,丢失学习状态异常的学生的代价很大,相比混合两阶段离群点检测算法,改进后的算法找出了更多学习状态异常的学生,在覆盖人数上实现了对原始算法的优化。实例分析表明,使用混合两阶段离群点检测算法或其改进算法,基于离群点检测的学生学习状态分析都能有效地找出学习状态异常的学生,其结果可以辅助高校学生工作者更加科学、高效地管理。
[Abstract]:Educational data mining in China is still in the stage of theoretical exploration, most of the research is theoretical description and feasibility analysis, the application of research is rare. Based on the research topic, this paper explores the application of data mining technology in the analysis of students' learning state from the perspective of the application of educational data mining. In our country, there exists the problem of imbalance between student worker and student ratio in colleges and universities, which leads to the lack of effective utilization of educational management resources. The purpose of this paper is to solve a practical problem: which students are allocated scarce personalized learning guidance of educational management resources in order to help them graduate smoothly, so that the limited educational management resources play a greater value. In this paper, the students with abnormal learning state are found out by using the analysis of students' learning state based on outlier detection, which provides the basis for the allocation of educational management resources. In order to obtain outliers, a hybrid two-stage outlier detection algorithm is designed. Firstly, the local outlier detection algorithm based on density is used to calculate the local outlier factor of each student. Then the outlier detection algorithm based on statistics is used for binary judgment to obtain outliers. Most of the outliers obtained by the algorithm are students with abnormal learning state, but the total number of students with abnormal learning state is small. In view of the shortcomings of the algorithm, the algorithm is improved from many aspects, from the point of view of data, the artificial attributes are added to expand the knowledge base, and from the point of view of the algorithm, the reduced iteration is used to detect the outliers in the decreasing data set. Finally, a comprehensive improvement will be formed by adding artificial attributes and reducing iterative fusion. In this problem, the cost of missing students with abnormal learning state is very high. Compared with the mixed two-stage outlier detection algorithm, the improved algorithm finds out more students with abnormal learning state, and optimizes the original algorithm in terms of coverage. Examples show that using mixed two-stage outlier detection algorithm or its improved algorithm, the learning state analysis of students based on outlier detection can effectively find out the students with abnormal learning state. The results can help college student workers to manage more scientifically and efficiently.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：G642;TP311.13
[Abstract]:Educational data mining in China is still in the stage of theoretical exploration, most of the research is theoretical description and feasibility analysis, the application of research is rare. Based on the research topic, this paper explores the application of data mining technology in the analysis of students' learning state from the perspective of the application of educational data mining. In our country, there exists the problem of imbalance between student worker and student ratio in colleges and universities, which leads to the lack of effective utilization of educational management resources. The purpose of this paper is to solve a practical problem: which students are allocated scarce personalized learning guidance of educational management resources in order to help them graduate smoothly, so that the limited educational management resources play a greater value. In this paper, the students with abnormal learning state are found out by using the analysis of students' learning state based on outlier detection, which provides the basis for the allocation of educational management resources. In order to obtain outliers, a hybrid two-stage outlier detection algorithm is designed. Firstly, the local outlier detection algorithm based on density is used to calculate the local outlier factor of each student. Then the outlier detection algorithm based on statistics is used for binary judgment to obtain outliers. Most of the outliers obtained by the algorithm are students with abnormal learning state, but the total number of students with abnormal learning state is small. In view of the shortcomings of the algorithm, the algorithm is improved from many aspects, from the point of view of data, the artificial attributes are added to expand the knowledge base, and from the point of view of the algorithm, the reduced iteration is used to detect the outliers in the decreasing data set. Finally, a comprehensive improvement will be formed by adding artificial attributes and reducing iterative fusion. In this problem, the cost of missing students with abnormal learning state is very high. Compared with the mixed two-stage outlier detection algorithm, the improved algorithm finds out more students with abnormal learning state, and optimizes the original algorithm in terms of coverage. Examples show that using mixed two-stage outlier detection algorithm or its improved algorithm, the learning state analysis of students based on outlier detection can effectively find out the students with abnormal learning state. The results can help college student workers to manage more scientifically and efficiently.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：G642;TP311.13
【参考文献】
中国期刊全文数据库 前10条
1 NMC地平线项目;龚志武;吴迪;陈阳键;苏宏;黄淑敏;陈木朝;吴杰锋;焦建利;;新媒体联盟2015地平线报告高等教育版[J];现代远程教育研究;2015年02期
2 乔慧君;周筠s，
本文编号：2139907
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2139907.html
