摘要
为了将HMM的Baum-welch算法应用到高校家庭贫困生认定过程中,首先将学生的经济状态依据客观情况设置为5个状态,然后将得到的观测数据依据外部低成本变量进行加权处理,再将加权处理后的数据按照一定的比例划分为7个等级,对不同等级进行分段统计,并在此基础上提出了使用HMM的Baum-Welch算法解决这个问题时构建初始化参数的方法,最后将迭代的结果依据学生贫困状态期望百分比由高到低顺序进行排序,并将结果与直接计算方法及通过实际调研得到的结论进行对比,通过对比得到了HMM算法在解决此类问题中存在的局限性,同时给出了提高预测准确性的新模型建立的建议.然后将这种方法在其它班级进行了验证,以检验结论的可靠性.
In order to apply the HMM Baum-welch algorithm to the identification of poverty stricken students in colleges, students’ economic situation is set into five levels according to the objective situation. Then, the observed data are weighted based on the external low-cost variables. The weighted data are divided into seven grades in appropriate proportions, and the different grades are counted. On this basis, the method of constructing initialization parameters, A, B when using Baum-Welch algorithm of HMM to solve this problem is put forward. Finally, the iterative results are sorted from high to low according to the expected percentage of students’ poverty state, and the results are compared with the direct calculation method and the conclusions obtained through practical investigation, and the limitations of HMM algorithm in solving this kind of problems are obtained by comparison. At the same time, some suggestions for establishing a new model to improve the accuracy of prediction are given. Then this method is verified in other classes to test the reliability of the conclusion.
作者
谢颖
朱远胜
马维聪
张英
XIE Ying;ZHU Yuan-sheng;MA Wei-cong;ZHANG Ying(Dean's Office,Zhejiang Fashion Institute of Technology,Ningbo,Zhejiang 315211,China)
出处
《浙江纺织服装职业技术学院学报》
2019年第3期75-83,106,共10页
Journal of Zhejiang Fashion Institute of Technology
基金
浙江省教育科学规划课题[编号:2018SCG226]
