基于GBDT算法的供电企业信息化会计核算数据异常挖掘  

Power supply enterprise information accounting data anomaly mining based on GBDT algorithm

作　　者：杨磊 YANG Lei(Audit Center,State Grid Shanghai Electric Power Company,Shanghai 200135,China)

机构地区：[1]国网上海市电力公司审计中心,上海200135

出　　处：《计算机应用文摘》2025年第2期85-87,90,共4页

摘　　要：对于会计核算数据异常的挖掘,供电企业主要依靠基于规则的专家系统。针对数据准确性不高的问题,文章提出了基于GBDT算法的供电企业信息化会计核算数据异常挖掘方法。首先,根据获取的特征分块函数和熵权法来提取异常数据特征;其次,采用独立成分分析法结合引入的拉格朗日乘子对特征矩阵进行约束,从而确定数据异常阈值,并使用GBDT数据分类算法对异常数据进行迭代处理与分类;最后,计算异常数据的最大均值差异,使用模糊神经网络对其进行识别与训练,从而输出异常挖掘结果。在实验过程中,通过与其他2种常规方法进行对比,发现该方法对异常数据挖掘的准确率为99.8%,说明其最为准确。For mining abnormal accounting data,power supply enterprises mainly rely on rule-based expert system.At the same time,aiming at the problem of low accuracy,this paper proposes a GBDT algorithm based method for mining abnormal accounting data of power supply enterprises.Firstly,the abnormal data features are extracted according to the obtained feature block function and entropy weight method.Secondly,the eigenmatrix is constrained by the independent component analysis method combined with the Lagrange multiplier,so as to determine the data anomaly threshold,and the GBDT data classification algorithm is used to iteratively process and classify the abnormal data.Finally,the maximum mean difference of the abnormal data is calculated,and the fuzzy neural network is used to identify and train the abnormal data,so as to output the anomaly mining results.Compared with other two conventional methods,the accuracy of this method is 99.8%,indicating that it is the most accurate method for abnormal data mining.

关 键 词：GBDT算法 企业信息化 数据异常 数据挖掘 异常挖掘 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]