基于BPNN回归预测算法的专家数量预测模型

doi:10.11959/j.issn.1000-0801.2023252

摘要/Abstract

摘要：

为迎合新型电力系统对数智化技术的发展需求，提升在集中招标采购数字化、智慧化过程中对数据价值提升、数据挖掘和数据分析的能力，站在招标人的角度，研究并提出了一种基于反向传播神经网络（back propagation neural network，BPNN）回归预测算法的预测模型。首先，深入探讨了电力行业的招标采购专家抽取现状，发现在集中招标采购过程中，需要大量专家参与其中，但专家数量的预测十分困难；其次，分析和学习历史数据，基于现有采购需求的情况及采购策略，有效定位专家需求基本情况，快速准确地预测专家数量；最后，验证了模型在资格预审与资格后审环节中的预测精度，为实际应用提供了参考，达到了降本增效的效果。

关键词: 预测模型, BPNN, 资格预审, 资格后审, 专家数量

Abstract:

To meet the development demands of digital technology for the new power system and enhance the data value, mining, and analytics capabilities for power industry in the process of centralized bidding and procurement digitalization and intelligence, a prediction model based on BPNN regression algorithm from the perspective of the bidder was proposed.Firstly, an in-depth investigation of the current situation of expert selection for power industry bidding and procurement was conducted, and it was discovered that predicting the required number of experts is challenging.Secondly, expert demand was efficiently located based on the existing procurement needs and strategies, and the required number of experts was rapidly predicted by analyzing and learning from historical data.Finally, the prediction accuracy of the model in the pre-qualification and post-qualification review stages was validated, providing practical reference and achieving the goal of reducing costs and increasing efficiency.

Key words: prediction model, BPNN, pre-qualification, post-qualification, expert number

中图分类号:

TP183

商天文, 仉健维, 黄昊晨. 基于BPNN回归预测算法的专家数量预测模型[J]. 电信科学, 2023, 39(12): 152-160.

Tianwen SHANG, Jianwei ZHANG, Haochen HUANG. Expert number prediction model based on BPNN regression algorithm[J]. Telecommunications Science, 2023, 39(12): 152-160.

图/表 10

表1

图1

表2

数据评估模型"

	专家工作效率	公司经济指标	权重
样本1	$f_{1}^{WE*}$	$f_{1}^{EF*}$	$f_{1}^{WE} + f_{1}^{EF}$
样本2	$f_{2}^{WE*}$	$f_{2}^{EF*}$	$f_{2}^{WE} + f_{2}^{EF}$
...	...	...	...
样本M	$f_{m}^{WE*}$	$f_{m}^{EF*}$	$f_{m}^{WE} + f_{m}^{EF}$

表2

图2

图3

表3

3种常用的激活函数"

函数名	表达式	图像
sigmoid函数	$f_{1} (s) = \frac{1}{1 + e^{- s}}$
tanh函数	$f_{2} (s) = \frac{e^{s} - e^{- s}}{e^{s} + e^{- s}}$
ReLU函数	$f_{3} (s) = max (0, s)$

表3

表4

图4

表5

表6

参考文献 17

[1]	李攀攀, 谢正霞, 乐光学 ,等. 基于深度学习的无线通信接收方法研究进展与趋势[J]. 电信科学, 2022,38(2): 1-17.
	LI P P , XIE Z X , YUE G X ,et al. Research progress and trends of deep learning based wireless communication receiving method[J]. Telecommunications Science, 2022,38(2): 1-17.
[2]	郭泽华, 朱昊文, 徐同文 . 面向分布式机器学习的网络模态创新[J]. 电信科学, 2023,39(6): 44-51.
	GUO Z H , ZHU H W , XU T W . Network modal innovation for distributed machine learning[J]. Telecommunications Science, 2023,39(6): 44-51.
[3]	陈靓, 钱亚冠, 何志强 ,等. 深度卷积神经网络的柔性剪枝策略[J]. 电信科学, 2022,38(1): 83-94.
	CHEN L , QIAN Y G , HE Z Q ,et al. A flexible pruning on deep convolutional neural networks[J]. Telecommunications Science, 2022,38(1): 83-94.
[4]	林瑾 . 电力物资招标采购管理大数据应用规划研究[J]. 营销界, 2019(46): 89,91.
	LIN J . Big data application planning for power supply materials bidding and procurement management[J]. Marketing Circles, 2019(46): 89,91.
[5]	李岩, 韩秋, 郑乃仁 . BP神经网络在电力需求决策中的应用[J]. 现代经济信息, 2009(22): 2.
	LI Y , HAN Q , ZHENG N R . Application of BP neural network in power demand decision-making[J]. Modern Economic Information, 2009(22): 2.
[6]	周正信 . 物资招标采购管理中的大数据应用规划[J]. 铁路采购与物流, 2020,15(11): 2.
	ZHOU Z X . Application planning of big data in material bidding and purchasing management[J]. Railway Purchase and Logistics, 2020,15(11): 2.
[7]	徐杰 . 面向电力物资多阶段迭代采购的供应商动态选择方法及系统[D]. 广州:广东工业大学, 2015.
	XU J . Dynamic supplier selection method and system for multi-stage iterative procurement of power materials[D]. Guangzhou:Guangdong University of Technology, 2015.
[8]	张浩 . 对如何加强电力设备招标采购管理工作的思考[J]. 中国管理信息化, 2020,23(19): 116-117.
	ZHANG H . Thoughts on how to strengthen the bidding and purchasing management of power equipment[J]. China Management Informationization, 2020,23(19): 116-117.
[9]	李良 . 遗传神经网络在股票市场预测中的应用[J]. 中国证券期货, 2013(2X): 47-48.
	LI L . Application of BP network in stock market forecasting[J]. Securities ＆ Futures of China, 2013(2X): 47-48.
[10]	王海燕, 刘佳康, 邓亚平 . 基于预估—校正综合BP神经网络的短期光伏功率预测[J]. 智慧电力, 2023,51(3): 46-52.
	WANG H Y , LIU J K , DENG Y P . Short-term photovoltaic power forecasting based on predict-correct combination BP neural network[J]. Smart Power, 2023,51(3): 46-52.
[11]	张宇晨, 张宇霖, 封春菲 ,等. 基于BP神经网络模型的输电线路造价预测模型研究[J]. 电力大数据, 2020,23(6): 35-42.
	ZHANG Y C , ZHANG Y L , FENG C F ,et al. Research on transmission line cost prediction model based on BP neural network[J]. Power Systems and Big Data, 2020,23(6): 35-42.
[12]	徐艳华, 周荣亚 . 基于大数据神经网络算法的学生成绩分析与预测模型仿真[J]. 电子设计工程, 2022,30(11): 108-112.
	XU Y H , ZHOU R Y . Simulation of student performance analysis and prediction model based on big data neural network algorithm[J]. Electronic Design Engineering, 2022,30(11): 108-112.
[13]	包研科, 陈然, 郑宏杰 ,等. 指数平滑与自回归融合预测模型及实证[J]. 计算机工程与应用, 2022,58(14): 269-281.
	BAO Y K , CHEN R , ZHENG H J ,et al. Index smoothing and autoregressive fusion forecasting model and its empirical study[J]. Computer Engineering and Applications, 2022,58(14): 269-281.
[14]	孙建国, 肖颂红 . 评标专家量化管理研究及应用[J]. 企业管理, 2016(S2): 454-455.
	SUN J G , XIAO S H . Quantitative management research and application of evaluation experts[J]. Enterprise Management, 2016(S2): 454-455.
[15]	陈祖军, 杨柳松, 陈泳 . 以电力行业为例探究如何提高招评标效率[J]. 中国招标, 2023(4): 141-144.
	CHEN Z J , YANG L S , CHEN Y . Exploring ways to improve bidding and evaluation efficiency in the power industry[J]. China Tendering, 2023(4): 141-144.
[16]	郭慧敏 . 评标专家违法违规行为成因及治理对策[J]. 招标采购管理, 2023(2): 42-43.
	GUO H M . Causes and countermeasures of illegal behavior of bid evaluation experts[J]. Tendering ＆ Purchasing Management, 2023(2): 42-43.
[17]	李百新 . 远程异地评标模式探索[J]. 建设监理, 2023(1): 39-41.
	LI B X . Exploration of remote bid evaluation mode in different places[J]. Project Management, 2023(1): 39-41.

评标批次	专业种类	投标人数/个	专家人数/个	评标天数/日	单日费用/元	产生费用/元
A1	材料A	157	30	4	500	60 000
	材料B	80	18	4	500	36 000
	材料C	76	12	4	500	24 000
	材料D	40	8	4	500	16 000
	材料E	37	10	4	500	20 000
	材料F	170	32	4	500	64 000
A2	材料A	172	35	5	500	87 500
	材料B	101	25	5	500	62 500
	材料C	130	21	5	500	52 500
	材料D	23	10	5	500	25 000
	材料E	70	12	5	500	30 000
	材料F	237	37	5	500	92 500
A3	材料A	148	28	4	500	56 000
	材料B	82	15	4	500	30 000
	材料C	98	16	4	500	32 000
	材料D	32	6	4	500	12 000
	材料E	43	13	4	500	26 000
	材料F	186	30	4	500	60 000

神经网络	神经元个数/个	激活函数	参数
输入层	120	sigmoid	840
隐藏层	60	ReLU	7 260
输出层	6	linear	366

组别	投标人数/个	拟邀请专家数量/个	实际邀请专家数量/个
材料A	154	11	10
材料B	55	10	10
材料C	51	10	10
材料D	29	4	5
材料E	35	8	7
材料F	324	23	24
合计	648	66	66

组别	投标人数/个	拟邀请专家数量/个	实际邀请专家数量/个
材料A	170	20	21
材料B	69	15	15
材料C	55	13	13
材料D	29	7	7
材料E	37	10	9
材料F	360	24	24
合计	720	89	89