基于改进剪枝策略的专业大数据主特征提取方法研究  

Research on major feature extraction methods of professional big data based on improved pruning strategy

作　　者：苗立宁 张韫 王利军 李希喆 李美亮 MIAO Lining;ZHANG Yun;WANG Lijun;LI Xizhe;LI Meiliang(State Grid Zhongxing Co.,Ltd.,Beijing 100761,China;Beijing Guodian Tong Network Technology Co.,Ltd.,Beijing 100070,China)

机构地区：[1]国网中兴有限公司,北京100761 [2]北京国电通网络技术有限公司,北京100070

出　　处：《电子设计工程》2025年第5期22-25,30,共5页Electronic Design Engineering

基　　金：国网中兴有限公司电网数字化项目(733400230004)。

摘　　要：电力工程信息的智能化检测与分析大多是基于数字类型的数据来展开的,而对于工程建设中文本型数据的检测模型而言,在检测准确度与检测速度方面均存在较大的改进空间。针对这一问题,文中构建了一套基于改进剪枝策略的专业大数据主要特征提取模型,以提高对专业文本数据的处理能力,实现对电力工程项目的全面评估。该模型通过深度神经网络对电力工程数据进行特征提取,并对数据特征进行结构分析,通过数据特征与风险库信息的相似度匹配,实现了对电力工程信息的风险判别。以某电力工程项目文本数据进行的评估测试结果表明,文中所提方法能够准确地提取出电力工程文本型数据的关键特征,并且可以分析挖掘出数据所蕴含的风险。The intelligent detection and analysis of power engineering information is mostly based on digital data,and there is significant room for improvement in the detection accuracy and speed of text data detection models in engineering construction.In response to this issue,a set of professional big data main feature extraction models based on improved pruning strategies was constructed in the article to improve the processing ability of professional text data and achieve comprehensive evaluation of power engineering projects.This model extracts features from power engineering data through deep neural networks and analyzes the structure of data features.By matching the similarity between data features and risk database information,it achieves risk discrimination of power engineering information.The evaluation test results based on text data of a certain power engineering project show that the proposed method can accurately extract the key features of power engineering text data and analyze the risks contained in the data.

关 键 词：剪枝策略 电力工程 大数据分析 文本数据 

分 类 号：TN92[电子电信—通信与信息系统] TP311.13[电子电信—信息与通信工程]