基于改进SVM的电力工程标签数据挖掘与分类技术  

Tag Data Mining and Classification Technology of Power Engineering Based on Improved SVM

作　　者：韩立芝 刘明红 左雅 刘灵爽 李香平 HAN Lizhi;LIU Minghong;ZUO Ya;LIU Lingshuang;LI Xiangping(State Grid Xinjiang Electric Power Co.,Ltd.,Urumqi 830017,China;Economic and Technical Research Institute of State Grid Xinjiang Electric Power Co.,Ltd.,Urumqi 830063,China)

机构地区：[1]国网新疆电力有限公司,新疆乌鲁木齐830017 [2]国网新疆电力有限公司经济技术研究院,新疆乌鲁木齐830063

出　　处：《微型电脑应用》2025年第1期83-86,共4页Microcomputer Applications

基　　金：国网新疆电力有限公司研究项目(BD30JY220001)。

摘　　要：为了给电力工程知识图谱的建设提供理论支撑,对标签挖掘技术进行研究,通过支持向量机(SVM)算法实现文本标签的分类。基于电力工程标签场景的稀疏特性对SVM算法加以优化,并采用一种双超平面的孪生SVM(TWSVM)算法提升泛化性能。利用改进后的粒子群优化(PSO)算法来解决TWSVM超参数取值困难的问题,一方面引入适应值增益反馈机制提升算法的迭代速度,另一方面通过渐变随机扰动机制避免训练的过早收敛。基于集成学习的思想,以实际电力工程数据为样本进行模型训练。仿真结果表明,所提改进算法的各项指标显著提升,F_(1)值提升了9.89个百分点,优化效果明显。In order to provide theoretical support for the construction of knowledge atlas of power engineering,the label mining technology is studied,and text labels are classified by support vector machine(SVM)algorithm.The SVM algorithm is improved based on the sparsity of label scenes in power engineering,and a twin SVM(TWSVM)algorithm with double hyperplanes is adopted to improve the generalization performance.The improved particle swarm optimization(PSO)algorithm is used to solve the problem in obtaining the hyperparameters of TWSVM.On the one hand,the fitness gain feedback mechanism is introduced to improve the iteration speed of the algorithm,and on the other hand,the gradual random perturbation mechanism is used to avoid premature convergence of training.Based on the idea of integrated learning,the model is trained with the actual power engineering data as samples.The simulation results show that the indicators of the improved algorithm are significantly improved,F_(1)value is increased by 9.89 percentage points,and the optimization effect is obvious.

关 键 词：标签分类 标签提取 粒子群优化算法 SVM 工程管理 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TN929.5[自动化与计算机技术—计算机科学与技术]