摘要
为了解决传统特征分类软件缺陷检测方法存在检测开销大的问题,提出了基于K-means算法的电网用户标签特征分类缺陷检测方法。应用权值分配算法和对应方差的评估值,以权值定位标签的特征,然后基于数据标签值进行数据重新分配。对现有的标签数据进行K-means聚类,保证数据挖掘的最优影响性,然后进行分布式处理挖掘,进一步划分数据簇,确定簇内标签,最后通过加权幅值算法,重新定义当前用户特征标签,并与原标签进行对比,实现标签特征分类软件缺陷检测。实验研究表明,利用K-means算法可以有效降低软件分类算法的错误率,保证软件的检测性能。
In order to solve the problem of large detection overhead of traditional feature classification software defect detection methods,a K-means algorithm based method for grid user label feature classification defect detection is proposed.Apply the weight distribution algorithm and the evaluation value of the corresponding variance to locate the feature of the label with the weight,and then redistribute the data based on the data label value.Perform K-means clustering on the existing label data to ensure the optimal impact of data mining,and then perform distributed processing mining to further divide the data clusters,determine the tags in the cluster,and finally redefine the current user through the weighted amplitude algorithm Feature tag,and compare it with the original tag,to realize the defect detection of tag feature classification software.Experimental research shows that using K-means algorithm can effectively reduce the error rate of software classification algorithms and ensure the detection performance of software.
作者
王炼
陆惠惠
WANG Lian;LU Hui-hui(State Grid Shanghai Songjiang Electric Power Supply Company,Shanghai 201699,China)
出处
《电子设计工程》
2020年第18期122-126,共5页
Electronic Design Engineering
基金
国网上海松江供电公司2019年用电信息采集系统建设与更新改造项目(B20935190008)。
关键词
电网用户
分类软件
数据挖掘
缺陷检测
power users
classification software
data mining
defect detection
