摘要
电力公司在运行时会产生大量的电力营销信息,为保证电力公司的正常运行,需要对电力营销信息实施有效的异常信息检测,为此提出基于分群算法与关联规则的电力营销信息异常检测方法。运用关联规则中的Apriori算法对电力营销信息展开挖掘处理,获取电力营销信息数据的频繁项集,并对项集展开支持度和置信度的计算,通过计算结果对频繁项集实施剪枝处理,去除其中的无关项集,完成电力营销信息异常特征项集的获取;基于群分算法中的k-means算法,根据电力营销信息异常特征项集找出聚类中心,引入轮廓系数描述初始聚类数量,对电力营销信息异常特征进行聚类分群处理,依据聚类结果输出不同电力营销信息异常类型,实现电力营销信息异常检测。实验表明,应用所提方法后,置信度最高接近1.000,最低值也达到了0.954,保证了生成的关联规则的有效性,改善了电力营销信息异常检测效果。
During the operation of power companies,a large amount of electricity marketing information is generated.To ensure the normal operation of power companies,effective anomaly detection of electricity marketing information is required.There-fore,a method for anomaly detection of electricity marketing information based on clustering algorithm and association rules is proposed.This method uses the Apriori algorithm in association rules to mine and process electricity marketing information,obtains frequent itemset of power marketing information data,and calculates support and confidence for the itemset.Based on the calculation results,pruning is applied to the frequent itemset to remove irrelevant itemset and complete the acquisition of anomaly feature itemset of electricity marketing information.Re-based on the k-means algorithm in the cluster partitioning algo-rithm,the clustering center is identified according to the anomaly feature itemset of power marketing information.The contour coefficient is introduced to describe the initial number of clusters,and the anomaly features of power marketing information are clustered and grouped.Based on the clustering results,different anomaly types of power marketing information are output to achieve anomaly detection of power marketing information.The experiment shows that after the application of the proposed method,the maximum confidence is close to 1.000,and even the lowest value reaches around 0.954,ensuring the effectiveness of the generated association rules,and improving the anomaly detection effect of power marketing information.
作者
李凯
苏华权
温游
李俊伟
卢妍倩
周昉昉
LI Kai;SU Huaquan;WEN You;LI Junwei;LU Yanqian;ZHOU Fangfang(Information Center,Guangdong Power Grid Corporation,Guangzhou 51o080,China)
出处
《微型电脑应用》
2025年第5期156-159,共4页
Microcomputer Applications
关键词
分群算法
关联规则
电力营销信息
异常检测
置信度
K-MEANS算法
clustering algorithm
association rule
electricity marketing information
anomaly detection
confidence
k-means algorithm
