基于双重扰动与核ELM融合的大学生贫困认定模型研究  

Research on College Students’Poverty Identification Model Based on Fusion of Dual Perturbation and Kernel ELM(DP_KELM)

作　　者：郑建华 朱蓉 刘双印 贺超波 ZHENG Jianhua;ZHU Rong;LIU Shuangyin;HE Chaobo(College of Information Science and Technology,Zhongkai University of Agriculture and Engineering,Guangzhou 510225,China;Guangdong Engineering&Technology Research Center for Smart Agriculture,Guangzhou 510225,China)

机构地区：[1]仲恺农业工程学院信息科学与技术学院,广州510225 [2]广东省高校智慧农业工程技术研究中心,广州510225

出　　处：《重庆理工大学学报（自然科学）》2021年第5期243-252,共10页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金项目(61471133,61871475);广东省科技计划项目(2017A070712019,2020A1414050062);广东省教育厅项目(2016KZDXM001,2017GCZX001,2020KZDZX1121);广州市科技计划项目(201704030098)。

摘　　要：精准实现大学生贫困认定是实现高校精准扶贫的重要前提,但是因为贫困认定需要提交的材料涉及隐私和数据非客观性原因,当前贫困认定存在部分学生做假和自卑学生不敢申请的不公平现象。针对该问题,以学生一卡通消费数据和个人基本信息等客观数据为基础,构建贫困特征分箱和特征交叉算法,形成一套大学生贫困认定特征参数。随后,为处理贫困认定数据集不平衡性,提出了数据样本和输入属性双重扰动方法,并与核ELM算法融合,构建了大学生贫困认定DP_KELM算法。实验结果表明:构建的特征在随机森林和KELM算法的准确率方面都超过0.82,而双重扰动模式对提高算法的G-mean值有较好作用,DP_KELM算法在G-mean和AUC上均优于对比的9种算法。DP_KELM算法能够有效识别贫困大学生,为实现校园精准扶贫提供辅助决策工具。Accurate poverty recognition of College Students is an essential prerequisite for the realization of targeted poverty alleviation in universities.However,due to the privacy and nonobjectivity of data involved in the materials required for poverty identification,there are some unfair phenomena in the current identification of poverty,such as some fake students’poverty and students who feel inferior dare not apply.To solve this problem,based on the objective data such as the consumption data of student card and personal basic information,this paper constructs the poverty feature box cutting and feature cross algorithm,and forms a set of poverty identification features of college students.Then,in order to deal with the imbalance of poverty identification data sets,a dual perturbation method of data sample and input attributes is proposed,and the DPKELM algorithm for poverty identification of college students is constructed by merging dual perturbation with the kernel ELM(Extreme Learning Machine)algorithm.The experimental results show that the accuracy of the features constructed in this paper exceeds 0.8 in both random forest and KELM algorithm,while the double disturbance mode has a good effect on improving the G-mean value of the algorithm,and DPKELM algorithm is better than the other nine algorithms in G-mean and AUC.DPKELM algorithm can effectively identify poor college students,and provides an auxiliary decision-making tool for achieving accurate poverty alleviation on campus.

关 键 词：双重扰动 不平衡数据 核极限学习机 贫困认定 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]