面向学生体质测量的数据异常可视化检测方法  

Visualized detection method of abnormal data for students’physical fitness measurement

作　　者：何佳佳[1] 安大庆[1] 叶美丽[1] HE Jia-jia;AN Da-qing;YE Mei-li(Shangluo University,Shangluo 726000,Shaanxi Province,China)

机构地区：[1]商洛学院,陕西商洛726000

出　　处：《信息技术》2023年第6期83-87,93,共6页Information Technology

基　　金：陕西省教育厅项目(21JK0113);陕西省体育局项目(2020233)。

摘　　要：单一的学生体质测量数据异常很难形成标准化判断方案,存在预测偏差准确性差、异常数据检测率低、异常数据误报率高的问题。提出面向学生体质测量的数据异常可视化检测方法,标准化处理数据,从而降低数据维度,清洗数据后预测出异常数据量最多的残缺数据。其次利用信息增益完成学生体侧数据特征的筛选以及排序,结合信息增益和正常数据点降低数据复杂度并识别出异常数据,实现学生体质测量异常数据检测。实验结果表明,所提方法的预测偏差准确性较高、异常数据检测率较高、异常数据误报率较低、检测运行时间较短。It is difficult for a single student to form a standard judgment scheme for abnormal data in physical fitness measurement,which has the problems of poor accuracy,low detection rate and high false alarm rate.Thus,an abnormal data visualization detection method for students’constitution measurement is proposed to process data in a standardized way,so as to reduce data dimension.After cleaning the data,the most abnormal data is forecasted.Secondly,information gain is used to select and sort the features of students’side data,and information gain and normal data points are combined to reduce the data complexity and identify abnormal data to achieve abnormal data detection.Experiment results show that the proposed method has high prediction bias accuracy,high detection rate of abnormal data,low false alarm rate of abnormal data,and short detection time.

关 键 词：可视化 数据异常检测 信息增益 预处理 偏差预测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]