高校熟人伤害情况的统计建模及关联规则挖掘

发布时间：2018-11-11 16:21

【摘要】：高校大学生群体是国家和社会极为关注的一个处在成长阶段的特殊群体,他们年轻活跃,有激情,好冲动,但心理不够成熟,伤害问题容易发生.对他们当中的熟人伤害问题的研究,可为预防、减少校园伤害事件提供相应的理论依据和对策建议。本文利用云南6所高校的6886个样本数据对高校熟人伤害问题进行分析,得到的基本结果是:熟人伤害约占大学生总伤害的58%,且熟人之间伤害的后果比自伤、陌生人伤害后果严重。进一步分别利用统计建模和数据挖掘方法分析熟人伤害的影响因素与伤害特征,获得了以下主要结果:首先,对熟人伤害数据进行描述性分析,得到了熟人伤害的一些基本特征,比如男生、家庭月供生活费低于1500元的学生、大二年级的学生、双亲家庭的孩子等这类人群更容易受到熟人伤害。其次,运用Logistic回归,逐步回归方法提取了熟人伤害的主要影响因素,并给出它们的重要性排序为:年级、学校、与辅导员接触情况、负面情绪、父母亲文化程度、是否恋爱、家庭月供、旷课情况、学校奖励、锻炼情况、性别、防范能力、是否赌博、安全意识淡薄、心智是否成熟、民族、与舍友相处情况、是否吸毒。将随机森林的三个版本(ntree分别等于200、500和700)、回归树两个版本(SE分别等于5.0和1)及Logistic模型共6个模型进行组合,且通过交叉验证的方法发现,随机森林ntree(28)700的版本为熟人伤害的最优统计模型。再次,运用关联规则,R编程计算揭示出多变量之间的关系,得到了熟人伤害的月份特征、时间段特征、伤害场所特征、人群特征。月份特征表明,熟人伤害主要发生在上半年,下半年中七月份最高。时间段特征表明,熟人伤害主要发生在学生空闲活动的时候,且大部分发生在17:30-00:30时间段,其中的时间段17:30-19:00是熟人伤害的高发时段;时间段00:00-00:30是熟人伤害发生的峰值时段。伤害场所特征表明,校内主要发生在宿舍和运动场;校外主要发生在实习见习地和往返途中。人群特征表明,大一至博士研究生,各年级都有各自的伤害特征。最后,对某高校发生的社会影响较大的一个实际案例进行了分析和验证。
[Abstract]:College students are a special group in the growing stage that the country and society pay close attention to, they are young and active, have passion, good impulse, but the psychology is not mature enough, the harm problem is easy to occur. The research on the injury problem of acquaintances among them can provide corresponding theoretical basis and countermeasures for preventing and reducing campus injury events. In this paper, 6886 sample data from 6 universities in Yunnan Province are used to analyze the injury of acquaintances in colleges and universities. The basic results are as follows: the injuries of acquaintances account for about 58% of the total injuries of college students, and the consequences of injuries among acquaintances are higher than those of themselves. The consequences of stranger injury are serious. Furthermore, by using statistical modeling and data mining methods to analyze the influence factors and injury characteristics of acquaintance injury, the following main results are obtained: firstly, descriptive analysis of acquaintance injury data is carried out. Some basic characteristics of acquaintance injury were obtained, such as male students, students whose monthly living expenses were less than 1500 yuan, sophomore students, children of two-parent families, and so on. Secondly, Logistic regression and stepwise regression were used to extract the main factors affecting the injury of acquaintances, and their importance was ranked as grade, school, contact with counselors, negative emotions, parents' education level, whether they were in love or not. Family monthly payments, truancy, school awards, exercise, gender, prevention, gambling, security awareness, mental maturity, nationality, getting along with roommates, whether drug abuse. Three versions of random forest (ntree = 200500 and 700), two versions of regression tree (SE = 5. 0 and 1) and six models of Logistic model were combined. The version of random forest ntree (28) 700 is an optimal statistical model for acquaintance injury. Thirdly, by using association rules and R programming calculation, the relationship between multivariable is revealed, and the month, time period, injury location and crowd characteristics of acquaintances are obtained. Monthly characteristics showed that acquaintance injuries mainly occurred in the first half of the year and the highest in July in the second half of the year. The characteristics of the time period indicate that the injury of acquaintances mainly occurs in the free activities of the students, and most of them occur in the period of 17: 30-0030: 30, in which the time period of 17: 30-19: 00 is the high frequency period of the injury of acquaintances. Time period 00: 00-00: 30 is the peak time when acquaintance damage occurs. The characteristics of the injury site indicate that the campus mainly takes place in the dormitory and sports ground, while the off-campus mainly occurs in the internship place and on the way back and forth. Crowd characteristics show that every grade has its own injury characteristics from freshman to doctoral graduate student. Finally, an actual case of social impact in a university is analyzed and verified.
【学位授予单位】：云南师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O212.1

【相似文献】