CN109118122A - Outsourcing supplier's evaluation method based on mixing PSO-Adam neural network - Google Patents
Outsourcing supplier's evaluation method based on mixing PSO-Adam neural network Download PDFInfo
- Publication number
- CN109118122A CN109118122A CN201811083184.7A CN201811083184A CN109118122A CN 109118122 A CN109118122 A CN 109118122A CN 201811083184 A CN201811083184 A CN 201811083184A CN 109118122 A CN109118122 A CN 109118122A
- Authority
- CN
- China
- Prior art keywords
- neural network
- pso
- outsourcing
- particle
- adam
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 54
- 238000012946 outsourcing Methods 0.000 title claims abstract description 47
- 238000011156 evaluation Methods 0.000 title claims abstract description 38
- 238000000034 method Methods 0.000 claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000012360 testing method Methods 0.000 claims abstract description 9
- 238000003062 neural network model Methods 0.000 claims abstract description 5
- 239000002245 particle Substances 0.000 claims description 54
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 13
- 238000004519 manufacturing process Methods 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 10
- 230000008859 change Effects 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 6
- 230000001537 neural effect Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 claims description 3
- 230000015556 catabolic process Effects 0.000 claims description 2
- 238000004891 communication Methods 0.000 claims description 2
- 238000012797 qualification Methods 0.000 claims description 2
- 230000011218 segmentation Effects 0.000 claims description 2
- 238000013068 supply chain management Methods 0.000 abstract 1
- 238000004364 calculation method Methods 0.000 description 3
- 238000012937 correction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明公开了一种基于混合PSO‑Adam神经网络的外协供应商评价方法，包括以下步骤：建立外协供应商评价数据集；使用留出法将数据集划分成k个大小相似的互斥子集；每次用k‑1个子集的并集作为训练集，余下的子集作为测试集，得到k组训练集与测试集；建立神经网络，神经网络的模型结构包括输入层、隐含层、输出层；神经网络训练，获得训练后的神经网络模型；完成对神经网络的训练后，将外协供应商数据集输入神经网络的输入层，神经网络输出层的输出即为外协供应商的评价得分。本发明方法能客观高效地进行外协供应商评价决策，减少对个人经验的依赖，降低供应商评价难度，减少供应链管理成本。
The invention discloses an outsourcing supplier evaluation method based on a hybrid PSO-Adam neural network, comprising the following steps: establishing an outsourcing supplier evaluation data set; dividing the data set into k mutually exclusive sets of similar sizes by using a set aside method Subset; each time the union of k-1 subsets is used as the training set, the remaining subsets are used as the test set, and k groups of training sets and test sets are obtained; the neural network is established, and the model structure of the neural network includes the input layer, hidden layer and output layer; neural network training, to obtain the trained neural network model; after completing the training of the neural network, input the outsourcing supplier data set into the input layer of the neural network, and the output of the neural network output layer is the outsourcing supply business evaluation score. The method of the invention can objectively and efficiently make outsourcing supplier evaluation decisions, reduce dependence on personal experience, reduce supplier evaluation difficulty, and reduce supply chain management costs.
Description
Technical field
The present invention relates to assessment techniques more particularly to a kind of outsourcing supplier based on mixing PSO-Adam neural network to comment
Valence method.
Background technique
As the emerging technologies such as artificial intelligence, big data, Internet of Things and cloud computing deepen continuously manufacturing, cause
The new round Industrial Revolution of developing direction, the manufacture ring of enterprise are turned to globalization, informationization, intelligence, wisdom and green
The change of essence also has occurred in border.More and more enterprise's selections carry out the Collaborative Manufacturing of networking with other enterprises, and will
Some non-core process requirements give more professional outsourcing supplier, to be absorbed in the raising of itself core competitiveness.When
The external coordination environment of preceding complexity provides more diversified cooperation processing resource to manufacturing enterprise, however the dispersion of cooperation processing resource
Property, also to the evaluation of the outsourcing supplier of manufacturing enterprise, more stringent requirements are proposed for diversity, dynamic, associativity, this makes
Traditional method is difficult to meet the needs of outsourcing supplier evaluates under current environment.Therefore, it is necessary to develop a kind of new external coordination
Supplier's decision-making technique is further reduced dependence of the conventional method to personal experience, reduces the difficulty of supplier evaluation, reduces and supplies
Chain management cost is answered, to adapt to the external coordination environment of current complexity.
Summary of the invention
The technical problem to be solved in the present invention is that for the defects in the prior art, providing a kind of based on mixing PSO-
The outsourcing supplier of Adam neural network evaluates decision-making technique, makes up the too strong defect of Multiobjective Decision Making Method subjectivity, thus
It is objective efficiently to carry out outsourcing supplier's evaluation decision, the dependence to personal experience is reduced, supplier evaluation difficulty is reduced, is reduced
Overhead cost for supply chain.
The technical solution adopted by the present invention to solve the technical problems is: a kind of based on mixing PSO-Adam neural network
Outsourcing supplier's evaluation method, comprising the following steps:
1) outsourcing supplier's evaluation data set is established, includes: from qualification rate, the matter for examining equipment to acquire in the data set
Measure stability and rate of breakdown data；The production and processing efficiency acquired from process equipment；The information acquired from MES system passes
Pass timeliness and accuracy of information communication data；The average cross ambiguity segmentation being calculated according to existing assessment indicator system is poor
Volume, the rate that delivers goods on schedule deliver accuracy rate, manufacture processing staff's quality, sophisticated equipment ratio, processing request change completion rate, work
Phase changes completion rate, processing quotation difference and transportation cost difference data；According to expert opinion obtain technology standardization degree,
The level of IT application, service level data；
2) use reserves method and data set D is divided into the similar exclusive subsets of k size, i.e. D=D1∪D2∪…∪Dk,
3) use the union of k-1 subset as training set every time, remaining subset obtains k group training set as test set
With test set, the normalization of sample data is finally carried out using min-max standardization；
4) neural network is established, the model structure of neural network includes input layer, hidden layer, output layer；Input layer and defeated
The neuron number of layer is determined by actual sample out, the network of the model selection list hidden layer of this neural network；
5) neural metwork training, the neural network model after being trained；
6) after completing to the training of neural network, by the input layer of outsourcing supplier's data set input neural network, nerve
The output of network output layer is the evaluation score of outsourcing supplier.
According to the above scheme, neural metwork training in the step 5), the neural network model after being trained, using PSO-
Adam hybrid optimization algorithm completes the training to neural network,
It is specific as follows:
5.1) PSO algorithm parameter and coding structure: particle length, population scale and the number of iterations of PSO are determined；
5.2) PSO algorithm initialization: the Position And Velocity of each particle in initialization particle populations；
5.3) global optimization: carrying out particle iteration, and global search is carried out in solution space, finds globally optimal solution, every time
After the completion of iteration, according to formula (10) (11) more new particle optimal location PiWith group optimal location PG；
5.4) neural network local optimum is carried out using Adam.
According to the above scheme, when the step 5.3) carries out particle iteration, in the latter stage of PSO iteration, it will accumulate in same point
Extra particle carry out randomization resetting.
The beneficial effect comprise that: outsourcing supplier's evaluation method energy based on mixing PSO-Adam neural network
It is enough for current complicated external coordination environment, it is objective and efficiently solve outsourcing supplier's evaluation problem, be further reduced to individual
The dependence of experience reduces supplier evaluation difficulty, reduces overhead cost for supply chain.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples, in attached drawing:
Fig. 1 is the algorithm flow chart of the embodiment of the present invention；
Fig. 2 is that the outsourcing supplier of the embodiment of the present invention evaluates network；
Fig. 3 is the PSO algorithm fitness curve of the embodiment of the present invention；
Fig. 4 is the Adam neural metwork training collection goodness of fit of the embodiment of the present invention；
Fig. 5 is the Adam neural network test set goodness of fit of the embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, is not used to limit
The fixed present invention.
As shown in Figure 1, outsourcing supplier's evaluation method based on mixing PSO-Adam neural network, this method are divided into data
Acquisition and processing, the global optimization of PSO algorithm, Adam neural network local optimum three phases, detailed step are as follows:
(1) data acquisition and procession
Step 1 data processing
The characteristics of present invention combination outsourcing supplier and actual conditions are established based on destination layer, rule layer, indicator layer
Three-level assessment indicator system.Destination layer is the comprehensive ability evaluation of outsourcing supplier；Rule layer is divided into eight classes, respectively quality
Level, delivery capability, manufacture working ability, information transfer capacity, exception processes ability, cost level, enterprise's internal and external environment, clothes
Business is horizontal；And with these eighth types of 22 indicator layers to be unfolded.Outsourcing supplier's assessment indicator system is as shown in Table 1.
1 outsourcing supplier's assessment indicator system of table
According to outsourcing supplier's assessment indicator system, data set is by being divided into acquisition, quantifying, qualitative three kinds of data.Acquisition
Index examines the data of equipment, process equipment and MES system to obtain by acquisition, and quantitative data is according to related data according to existing
Assessment indicator system is calculated, and for qualitative index, by expert system or establishes expert opinion group and score
It arrives.
Data set D is divided into the similar exclusive subsets of k size, i.e. D=D using the method that reserves (hold-out)1∪D2
∪…∪Dk,Each subset DiThe consistency of data distribution, i.e. D are all kept as far as possibleiFrom D
It is obtained by stratified sampling.Then, use the union of k-1 subset as training set every time, remaining subset is obtained as test set
To k group training set and test set, sample data finally is realized using min-max standardization (Min-Max Normalization)
Normalization, formula is such as shown in (1).
Wherein, X is not normalized sample data, X*It is a sample data after normalization, XminIt is sample number
According to minimum value, XmaxIt is the maximum value of sample data.
The design of step 2) Artificial Neural Network Structures
The model structure of neural network includes input layer, hidden layer, output layer.The neuron number of input layer and output layer
It is to be determined by the sample of realistic model.The present invention selects the network of single hidden layer, and the neural network of single hidden layer is not limiting mind
While through number of network node, arbitrary nonlinear mapping may be implemented, and the training time is relatively short, precision, which can reach, to be wanted
It asks.The selection of node in hidden layer k is according to following four kinds of empirical equations:
(1)N is input layer number in formula, and k is node in hidden layer, and N is sample number, as i > k
Provide number of combinations
(2)N is input layer number in formula, and k is node in hidden layer, and m is output node layer
Number, constant of the t between [1,10].
(3) k=log2N, n is input layer number in formula, and k is node in hidden layer.
(4)N is input layer number in formula, and k is node in hidden layer, and m is output layer number of nodes.
The evaluation problem of outsourcing supplier is substantially a kind of Generalized Multivariate regression problem, therefore by the target letter of neural network
Number is set as MSE, and formula is such as shown in (2):
Wherein, m refers to the number of output node, and N is training sample number,It is network desired output,It is network reality
Border output valve.
In order to allow network to obtain powerful nonlinear fitting ability, the present invention is added in the active coating and output layer of network
Tanh activation primitive maps the output of neuron, and tanh formula is such as shown in (3).
Step 3) neural metwork training
3.1) PSO algorithm parameter and coding structure are determined
PSO in the present invention in global search space for finding optimal neural network initial weight threshold value, it may be assumed that net
Connection weight w of the network input layer to hidden layerji, the threshold value b of hidden layer1j, the connection weight w of hidden layer to output layerkj, output
Layer threshold value b2k.In standard three-layer neural network structure, network input layer number of nodes is m, node in hidden layer n, output layer
Number of nodes is q, then the connection weight w of input layer to hidden layerjiNumber be n × m.The threshold value b of hidden layer1jNumber be n,
Connection weight w of the hidden layer to output layerkjNumber be q × n.Output layer threshold value b2kNumber be q.The particle length of PSO is
Parameter summation to be optimized, calculation formula is such as shown in (4).
N=m × n+n × q+n+q (4)
3.2) PSO algorithm global optimization
PSO algorithm initialization
Before PSO starting, the Position And Velocity of each particle in random initializtion particle populations is needed.Each particle
Massless is considered without the point in the D dimension space of volume, wherein D is the number of parameter.Assuming that the population of a N number of particle,
The original position of its population can indicate as follows:
w0=[w1,0,w2,0…wn,0] (5)
v0=[v1,0,v2,0…vn,0] (6)
Wherein, W0And V0Respectively refer to the initial position collection and initial velocity collection of particle.W is the D n dimensional vector n of particle coordinate.And V
It is the D n dimensional vector n of particle rapidity.
PSO algorithm iteration
In the iteration searching process of PSO, the coordinate of particle is a parameter set, represents the possibility of optimization problem
Solution.The purpose of particle is to be moved to a more excellent position from current location in search space.Particle movement speed is worked as by particle
Preceding speed, personal best particle, population optimal location determine that position is determined by current location and movement speed after particle updates.
Particle rapidity of the invention and position iterative formula are as follows:
vi,t+1=x [vi,t+c1r1(Pi-wi,t)+c2r2(PG-wi,t)] (7)
wi,t+1=wi,t+vi,t+1 (8)
Wherein, vI, t+1It is the pace of change vector of particle i, vI, tIt is the current velocity vector of particle i；wI, t+1It is particle i
Updated coordinate, wI, tIt is the changing coordinates of particle i；c1, c2It is acceleration constant；r1, r2It is two groups to be uniformly distributed at random
Number；Pi, PGRespectively refer to the personal best particle of particle i and group's optimal location of particle populations.Wherein x is for stable particle
The compressibility factor that movement speed is added, is defined as follows:
Wherein, φ=c1+c2>4.Particle optimal location PiWith group optimal location PGIt is updated according to following formula:
ifξ(Pi) > ξ (wi)then Pi=wi,t,1≤i≤N (10)
ifξ(PG) > ξ (wi)then PG=wi,t,1≤i≤N (11)
Wherein, ξ is fitness function.
Particle TSP question
The particle inside population reaches unanimity too early in order to prevent, so that the search space of algorithm reduces, herein original
PSO on the basis of, joined TSP question and particle reset mechanism, prevent network over-fitting, guarantee the generalization ability of network.
In PSO iteration mid-term, with some value in random chance resetting random particles coordinate vector.In the latter stage of PSO iteration, will assemble
Randomization resetting is carried out in the extra particle of same point.
3.3) Adam neural network local optimum
Step 4Adam neural network iteration
The present invention carries out the Feedback error of neural network, the method ratio SGD of this autoadapted learning rate using Adam
Convergence rate is faster, it is easier to converge to global extremum.Adam needs to calculate the band weight average m of gradienttThere is partial variance with cum rights
vt, calculation formula is as follows:
mt=β1mt-1+(1-β1)gt (12)
Due to mtAnd vtInitialization is 0 vector, therefore is easy to be biased to 0 vector, so needing to mtAnd vtCarry out deviation school
Just, the gradient zone weight average after correction isGradient cum rights after correction has the partial variance to beUpdating formula is as follows:
Therefore, Adam final more new formula is as follows:
Wherein β1Take default value 0.9, β2Default value be 0.999, ε value be 10-8。
Step 5) outsourcing supplier evaluation
After PSO-Adam hybrid optimization algorithm is completed to the training of neural network, the evaluation of outsourcing supplier can be carried out.
By the input layer of outsourcing supplier's data set input neural network, the output of neural network output layer is commenting for outsourcing supplier
Valence score.
One specific embodiment:
(1) data acquisition and procession
Step 1 data processing
The method of reserving is used to extract 24 groups of outsourcing supplier's data of certain equipment for building materiaIs manufacturing company as sample data, specific number
According to as shown in table 2, wherein S1To S20To have flag data collection, S21To S24For data set to be evaluated, finally marked using min-max
The normalization of standardization (Min-Max Normalization) realization sample data.
2 outsourcing supplier of table evaluates sample data table (1)
2 outsourcing supplier of table evaluates sample data table (2)
Step 2 Neural Network Structure Design
The data of analytical table 2 can determine that neural network input layer number of nodes is 22 according to this method, output layer number of nodes
It is 1, the hidden layer number of plies is 1, number of nodes 12, and it is as shown in Figure 2 to obtain outsourcing supplier's evaluation network.
Step 3 determines PSO algorithm parameter and coding structure
Particle length press according to formula (4) calculate 289.Particle inertia weight enables algorithm from 0.9 linear decrease to 0.4
There is stronger ability of searching optimum at the initial stage of operation, and being capable of fast convergence in the later period of operation.Minimum change is set
Step-length and minimum change fitness are 1e-6, and the result that algorithm obtains when less than threshold value tends towards stability, and algorithm is enabled to terminate, to reduce
Calculation amount.The population scale of PSO algorithm is usually determined by experiment with the number of iterations, and it is as shown in table 3 to obtain PSO major parameter.
3 PSO major parameter table of table
PSO algorithm global optimization
Step 4 PSO algorithm initialization
The PSO particle length known to PSO coding structure is 289, therefore initializes the position of each particle in particle populations
It is as follows with speed:
w0=[w1,0,w2,0…w289,0]
v0=[v1,0,v2,0…v289,0]
Step 5 PSO algorithm iteration
Particle iteration is carried out according to formula (7) (8), global search is carried out in solution space, finds globally optimal solution, every time
After the completion of iteration, according to formula (10) (11) more new particle optimal location PiWith group optimal location PG。
Step 6 particle TSP question
The particle inside population reaches unanimity too early in order to prevent, so that the search space of algorithm reduces, in PSO iteration
Phase, with some value in random chance resetting random particles coordinate vector.In the latter stage of PSO iteration, same point will accumulate in
Extra particle carries out randomization resetting.
Fitness curve is obtained by PSO algorithm initialization and iteration as shown in figure 3, PSO is by 15 generations as seen from the figure
The overall situation simultaneously optimizes, and the MSE of neural network is down to 0.1289 from 0.7653, error significantly reduces.
Adam neural network local optimum
Step 7 neural network iteration
Local optimum, the mean square error and fitting of neural network are carried out using Adam neural network described in this method respectively
Goodness is as shown in table 4, and training set, the goodness of fit figure difference of test set are as shown in Figure 4, Figure 5.
4 algorithm performance table of table
It can be seen that very high precision and the goodness of fit can be obtained using method of the invention, and the method was using
In journey can the parameters such as adaptive polo placement learning rate, influence evaluation result by initial parameters such as learning rates, reduce method
Enforcement difficulty, enable supplier evaluation result more objective credible.
Step 8 outsourcing supplier evaluation
4 groups of data samples to be evaluated are evaluated using the neural network of PSO-Adam algorithm optimization, evaluation knot
Fruit is as shown in table 5.
5 sample to be tested evaluation result of table
It is found that outsourcing supplier S21Evaluation score highest, S22Score is lower；S21And S23It can be used as outstanding outsourcing supplier
It is included in resources bank；S24It can be used as the outsourcing supplier commonly cooperated, S22Cooperation Risk is higher, need to cooperate with caution.Evaluation result is handed over
It is examined by panel of expert of company, has obtained the generally approval of associate.Application example proves, based on mixing PSO-Adam nerve
Outsourcing supplier's evaluation method of network can be objective and efficiently solve outsourcing supplier for current complicated external coordination environment
Evaluation problem is further reduced the dependence to personal experience, reduces supplier evaluation difficulty, reduces overhead cost for supply chain.
It should be understood that for those of ordinary skills, it can be modified or changed according to the above description,
And all these modifications and variations should all belong to the protection domain of appended claims of the present invention.
Claims (3)
1. a kind of outsourcing supplier's evaluation method based on mixing PSO-Adam neural network, which is characterized in that including following step
It is rapid:
1) outsourcing supplier's evaluation data set is established, includes: steady from the qualification rate, quality for examining equipment to acquire in the data set
Qualitative and rate of breakdown data；The production and processing efficiency acquired from process equipment；From MES system acquire information transmitting and
When property and accuracy of information communication data；The average cross ambiguity segmentation difference that is calculated according to existing assessment indicator system, by
Phase delivery ratio delivers accuracy rate, manufacture processing staff's quality, sophisticated equipment ratio, processing request change completion rate, duration change
Completion rate, processing quotation difference and transportation cost difference data；According to the technology standardization degree of expert opinion acquisition, informationization
Horizontal, service level data；
2) use reserves method and data set D is divided into the similar exclusive subsets of k size, i.e. D=D1∪D2∪…∪Dk,
3) use the union of k-1 subset as training set every time, remaining subset obtains k group training set and survey as test set
Examination collection finally carries out the normalization of sample data using min-max standardization；
4) neural network is established, the model structure of neural network includes input layer, hidden layer, output layer；Input layer and output layer
Neuron number determined by actual sample, the network of the model selection list hidden layer of this neural network；
5) neural metwork training, the neural network model after being trained；
6) after completing to the training of neural network, by the input layer of outsourcing supplier's data set input neural network, neural network
The output of output layer is the evaluation score of outsourcing supplier.
2. outsourcing supplier's evaluation method according to claim 1 based on mixing PSO-Adam neural network, feature
It is, neural metwork training in the step 5), the neural network model after being trained, is calculated using PSO-Adam hybrid optimization
Method completes the training to neural network,
It is specific as follows:
5.1) PSO algorithm parameter and coding structure: particle length, population scale and the number of iterations of PSO are determined；
5.2) PSO algorithm initialization: the Position And Velocity of each particle in initialization particle populations；
5.3) global optimization: carrying out particle iteration, and global search is carried out in solution space, finds globally optimal solution, each iteration
After the completion, according to formula (10) (11) more new particle optimal location PiWith group optimal location PG；
5.4) neural network local optimum is carried out using Adam.
3. outsourcing supplier's evaluation method according to claim 2 based on mixing PSO-Adam neural network, feature
It is, when the step 5.3) carries out particle iteration, in the latter stage of PSO iteration, the extra particle that will accumulate in same point is carried out
Randomization resetting.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811083184.7A CN109118122A (en) | 2018-09-17 | 2018-09-17 | Outsourcing supplier's evaluation method based on mixing PSO-Adam neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811083184.7A CN109118122A (en) | 2018-09-17 | 2018-09-17 | Outsourcing supplier's evaluation method based on mixing PSO-Adam neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109118122A true CN109118122A (en) | 2019-01-01 |
Family
ID=64859530
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811083184.7A Pending CN109118122A (en) | 2018-09-17 | 2018-09-17 | Outsourcing supplier's evaluation method based on mixing PSO-Adam neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109118122A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111105127A (en) * | 2019-11-06 | 2020-05-05 | 武汉理工大学 | Modular product design evaluation method based on data driving |
| CN112417759A (en) * | 2020-11-19 | 2021-02-26 | 天津大学 | A Method for Solving Inverse Problem of Thermal Conduction Based on Dynamic Neural Network |
| CN112508378A (en) * | 2020-11-30 | 2021-03-16 | 国网北京市电力公司 | Processing method and device for screening power equipment production manufacturers |
| CN114287018A (en) * | 2020-02-12 | 2022-04-05 | 株式会社日立制作所 | Purchasing service assisting system, purchasing service assisting device, and purchasing service assisting method |
| CN116028864A (en) * | 2023-01-06 | 2023-04-28 | 国网江苏省电力有限公司物资分公司 | A supplier classification method using GA parameter optimization based on knowledge graph |
-
2018
- 2018-09-17 CN CN201811083184.7A patent/CN109118122A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111105127A (en) * | 2019-11-06 | 2020-05-05 | 武汉理工大学 | Modular product design evaluation method based on data driving |
| CN111105127B (en) * | 2019-11-06 | 2023-04-07 | 武汉理工大学 | Modular product design evaluation method based on data driving |
| CN114287018A (en) * | 2020-02-12 | 2022-04-05 | 株式会社日立制作所 | Purchasing service assisting system, purchasing service assisting device, and purchasing service assisting method |
| CN112417759A (en) * | 2020-11-19 | 2021-02-26 | 天津大学 | A Method for Solving Inverse Problem of Thermal Conduction Based on Dynamic Neural Network |
| CN112417759B (en) * | 2020-11-19 | 2022-09-23 | 天津大学 | A method for solving the inverse problem of heat conduction based on dynamic neural network |
| CN112508378A (en) * | 2020-11-30 | 2021-03-16 | 国网北京市电力公司 | Processing method and device for screening power equipment production manufacturers |
| CN116028864A (en) * | 2023-01-06 | 2023-04-28 | 国网江苏省电力有限公司物资分公司 | A supplier classification method using GA parameter optimization based on knowledge graph |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109118122A (en) | Outsourcing supplier's evaluation method based on mixing PSO-Adam neural network | |
| CN104536412B (en) | Photoetching procedure dynamic scheduling method based on index forecasting and solution similarity analysis | |
| CN107862970B (en) | Teaching quality evaluation model for turnover classroom | |
| Xiong et al. | AHP fuzzy comprehensive method of supplier evaluation in social manufacturing mode | |
| Tsyganov | Designing adaptive information models for production management | |
| Azarenko et al. | The model of human capital development with innovative characteristics in digital economy | |
| CN106779346A (en) | A kind of Forecasting Methodology of monthly power consumption | |
| Shi et al. | Analytics for IoT‐enabled human–robot hybrid sortation: an online optimization approach | |
| Sun | Study on application of data mining technology in university computer network educational administration management system | |
| Foroozesh et al. | A hesitant fuzzy extension of VIKOR method for evaluation and selection problems under uncertainty | |
| Santos et al. | Fuzzy systems for multicriteria decision making | |
| CN114841371A (en) | Characteristic-based box separation processing method, device, equipment and medium | |
| Ocampo et al. | An integrated sustainable manufacturing strategy framework using fuzzy analytic network process | |
| Shi et al. | A dynamic novel approach for bid/no-bid decision-making | |
| Eraslan et al. | A decision making method by combining TOPSIS and grey relation method under fuzzy soft sets | |
| CN109919219A (en) | A Xgboost Multi-View Portrait Construction Method Based on Granular Computing ML-kNN | |
| López et al. | A comparison of forecasting methods for ro-ro traffic: A case study in the strait of Gibraltar | |
| Liu et al. | [Retracted] Analysis of Efficiency of Human Resource Management Evaluation Model Based on SOM Neural Network | |
| Chen et al. | A two-phase decision-making method for handling personnel selection problem | |
| Sudakov et al. | Use of fuzzy areas of preference in tasks of multi–criteria evaluation of freight air transport | |
| Murni et al. | The effect of entrepreneurial orientation on the management capability and sustainable innovation (study on Batik small and medium enterprises in East Java Indonesia) | |
| Zhiyuan et al. | Research on the evaluation of enterprise competitiveness based on the wavelet neural network forecasting system | |
| Tu | Economic benefits of flexible production in green manufacturing enterprises based on fuzzy scheduling model algorithm | |
| Zhao et al. | An improved Pythagorean fuzzy hybrid aggregation method based on grey correlation for knowledge collaborative innovation partner selection | |
| Baušys et al. | Multicriteria decision making approach by vikor under interval neutrosophic set environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190101 |
|
| RJ01 | Rejection of invention patent application after publication |