主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2023, Vol. 31 ›› Issue (11): 321-331.doi: 10.16381/j.cnki.issn1003-207x.2022.0286

• • 上一篇    

基于数据挖掘的电力装备企业多价值链协同数据预处理方法研究及应用

牛东晓1(),斯琴卓娅1,王董禹1,许晓敏1,张焕粉2   

  1. 1.华北电力大学经济与管理学院,北京 102206
    2.北京清畅电力技术股份有限公司,北京 100085
  • 收稿日期:2022-02-17 修回日期:2022-05-16 出版日期:2023-11-15 发布日期:2023-12-05
  • 通讯作者: 牛东晓 E-mail:niudx@126.com
  • 基金资助:
    国家重点研发计划资助项目(2020YFB1707801)

Method and Application of Multi-value Chain Collaborative Data Mining in Power Equipment Enterprises Based on Deep Learning

Dong-xiao NIU1(),Zhuo-ya SIQIN1,Dong-yu WANG1,Xiao-min XU1,Huan-fen ZHANG2   

  1. 1.School of Economy and Management, North China Electric Power University, Beijing 102206, China
    2.Beijing QingChang Power Technology Co. , Ltd, Beijing 100085, China
  • Received:2022-02-17 Revised:2022-05-16 Online:2023-11-15 Published:2023-12-05
  • Contact: Dong-xiao NIU E-mail:niudx@126.com

摘要:

在电力装备制造企业的数字化转型中,需要对数据空间中多价值链协同的高维数据进行挖掘与分析,本文针对电力装备制造业进销存大数据的预处理问题展开了研究。首先,给出了变点法和局部异常因子算法(local outlier factor method,LOF)组合的数据异常值检验校正预处理方法;其次,提出了基于LASSO(least absolute shrinkage and selection operator,LASSO)算法的栈式稀疏自编码器(stack sparse auto-encoder,SSAE)数据降噪降维组合机器学习处理方法(SSAE-LASSO),对特征进行压缩降维提取,去除严重干扰数据回归分析的噪声信息,并过滤影响度低的冗余数据,从而实现数据的降噪降维处理。最后,将本文提出的方法应用于不同的算法进行检验,通过对两种预处理的数据对比发现,本文提出的方法有效提高了电力产品销售量智能预测的精度。

关键词: 电力装备企业, 多价值链协同, 数据挖掘预处理方法, 机器学习

Abstract:

In the digital transformation of power equipment manufacturing enterprises, it is necessary to mine and analyze the high-dimensional data of multi-value chain collaboration in the data space. The preprocessing of purchase, sales and inventory big data in power equipment manufacturing industry is studied. Firstly, the preprocessing method of data outlier checking and correction based on the combination of change point method and Local Outlier Factor method (LOF method) is given. Secondly, the data dimensionality reduction processing method of Stack Sparse Auto Encoder (SSAE) based on LASSO (Least Absolute Shrinkage and Selection Operator) deep learning algorithm is proposed (SSAE-LASSO), which can compress and reduce the features, remove the noise information that seriously interferes with the data regression analysis, and filter the fault-tolerant redundant data with low influence, so as to realize the denoising and dimensionality reduction processing of the data. Finally, the method proposed in this paper is applied to different algorithms to test. By comparing the two preprocessed data, it is found that the method proposed in this paper can effectively improve the accuracy of the intelligent prediction of the sales volume of electric power products.

Key words: power equipment enterprises, multi value chain collaboration, data mining preprocessing method, deep learning

中图分类号: