CN114912074A - Intelligent calculation method for manufacturing production based on data space multi-value chain statistical analysis - Google Patents
Intelligent calculation method for manufacturing production based on data space multi-value chain statistical analysis Download PDFInfo
- Publication number
- CN114912074A CN114912074A CN202210359177.5A CN202210359177A CN114912074A CN 114912074 A CN114912074 A CN 114912074A CN 202210359177 A CN202210359177 A CN 202210359177A CN 114912074 A CN114912074 A CN 114912074A
- Authority
- CN
- China
- Prior art keywords
- value chain
- statistical analysis
- chain
- value
- influence factors
- 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
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 109
- 238000007619 statistical method Methods 0.000 title claims abstract description 57
- 238000004364 calculation method Methods 0.000 title claims abstract description 13
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 44
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 36
- 238000007637 random forest analysis Methods 0.000 claims abstract description 20
- 230000006870 function Effects 0.000 claims abstract description 18
- 230000004913 activation Effects 0.000 claims abstract description 14
- 241001481710 Cerambycidae Species 0.000 claims abstract description 8
- 238000000034 method Methods 0.000 claims description 63
- 239000002994 raw material Substances 0.000 claims description 25
- 238000012549 training Methods 0.000 claims description 19
- 238000005457 optimization Methods 0.000 claims description 11
- 238000012216 screening Methods 0.000 claims description 11
- 238000005070 sampling Methods 0.000 claims description 8
- 238000009499 grossing Methods 0.000 claims description 5
- 238000009825 accumulation Methods 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims 1
- 238000011160 research Methods 0.000 description 28
- 238000004458 analytical method Methods 0.000 description 24
- 238000011156 evaluation Methods 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 8
- 238000007726 management method Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 238000010276 construction Methods 0.000 description 6
- 238000000556 factor analysis Methods 0.000 description 5
- 238000012417 linear regression Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 238000000611 regression analysis Methods 0.000 description 4
- 238000010187 selection method Methods 0.000 description 4
- 238000013523 data management Methods 0.000 description 3
- 238000000513 principal component analysis Methods 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 238000012163 sequencing technique Methods 0.000 description 3
- 241000254173 Coleoptera Species 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000005094 computer simulation Methods 0.000 description 2
- 238000010219 correlation analysis Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000010146 3D printing Methods 0.000 description 1
- 238000012935 Averaging Methods 0.000 description 1
- 238000012952 Resampling Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000009133 cooperative interaction Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000013439 planning Methods 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression 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
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- 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
-
- 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
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Analysis (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Mathematical Optimization (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Pure & Applied Mathematics (AREA)
- Biomedical Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Computational Mathematics (AREA)
- Molecular Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Educational Administration (AREA)
- Databases & Information Systems (AREA)
- Game Theory and Decision Science (AREA)
- Algebra (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Computational Biology (AREA)
Abstract
The invention provides a data space-based intelligent calculation method for statistical analysis of a multi-value chain in manufacturing industry production, which reconstructs influence factors on an external value chain of the multi-value chain by using a random forest algorithm, calculates contribution degrees of the influence factors of all links of different value chains, and reconstructs the influence factors according to the contribution degrees. And then, optimizing the learning rate and the activation function of the convolutional neural network by using a longicorn beard algorithm, finding the optimal learning rate and the activation function of the convolutional neural network, and forming the basis of the BAS-optimized convolutional neural network as a statistical analysis decision intelligent model of the manufacturing enterprise. And finally, constructing an RF-BAS-CNN model, taking the comprehensive influence factors of the multi-value chain reconstruction as input, taking the internal value chain production factors as output, and performing statistical analysis decision of manufacturing enterprises.
Description
Technical Field
The invention belongs to the field of data space intelligent calculation, and particularly relates to a manufacturing production intelligent calculation method based on data space multi-value chain statistical analysis.
Background
The data management method of the data space is combined with the current statistical analysis decision problem of the intelligent manufacturing enterprise, so that the statistical analysis decision level in the intelligent manufacturing field can be effectively improved, and the high-quality rapid development of the manufacturing industry is promoted.
Today, the global industry value chain is being adjusted and upgraded, and how well each country participates in the global industry value chain reconfiguration. The establishment of the structure of the multivalent value chain is beneficial to fully integrating the value chains of various enterprises in the industrial chain and continuously designing and redesigning the value system of the industrial chain. Meanwhile, a multi-value chain cooperative statistical analysis decision model is created, the mutual relation among all business links in the manufacturing industry is effectively controlled, the cooperative interaction is realized, the influence of multi-dimensional factors such as supply, marketing and service on the manufacturing industry statistical analysis decision is accurately analyzed from the perspective of the multi-value chain, and an effective way is provided for the high-quality development of the manufacturing industry.
At present, a plurality of problems exist in the production and operation process of the manufacturing industry in China. First, manufacturing enterprises are under a double pressure of both internal challenges and external environmental changes in implementing smart manufacturing. From the interior of enterprises, the rising of production cost, the insufficient research and development investment and the more traditional production organization mode are all the specific problems to be solved urgently at present. From the external environment, consumers have greater dominance, the old manufacturing mode is overturned by the technical development of big data, cloud computing, 3D printing, robots and the like, and the trends of cross-border fusion and manufacturing service are increasingly remarkable. Second, current research on multiple value chains of manufacturing enterprises focuses more on the business scope of a single value chain, and research on the collaborative relationships between multiple value chains is less. The construction of a multivalence value chain cooperative architecture is basically blank for the manufacturing industry. Thirdly, the application of the intelligent data management technology of the data space in the multivalence value chain architecture is not realized, the selection method of the influence factors of the multivalence value chain focuses more on factor analysis, component analysis and the like, and the objective analysis is performed by less using an intelligent algorithm; in the research method of the synergy of the multivalent value chain, a regression analysis method is mostly used, and a deep learning algorithm is less used.
At present, a complete research system is not formed aiming at the cooperative relationship among the manufacturing industry multi-value chains, the selection method of the manufacturing industry multi-value chain influence factors focuses on factor analysis, component analysis and the like more, and an intelligent algorithm is less used for objective analysis; in the research method of the synergy of the multivalent value chain, a regression analysis method is mostly used, and a deep learning algorithm is less used. The deep learning algorithm is a mainstream method for solving the complex data driving problem in the field of artificial intelligence due to the characteristics of wide width, high upper limit, multiple neural network layers and the like, and is often applied to the field of prediction. And the manufacturing industry data space platform can better adapt to the characteristics of intelligent manufacturing, can better utilize distributed heterogeneous data of manufacturing enterprises, and associates the data with intelligent application, thereby better supporting the manufacturing industry statistical analysis decision.
Disclosure of Invention
Aiming at the problems, the invention provides an intelligent calculation method for manufacturing production based on data space multi-value chain statistical analysis. The method comprises the steps of firstly forming a manufacturing industry multivalued value chain cooperation mode by using a data space manufacturing industry multivalued value chain cooperation system architecture, then screening multivalued value chain factors by using a random forest algorithm, and finally constructing a Random Forest (RF) -longicorn whisker optimization (BAS) -Convolutional Neural Network (CNN) combination model to perform statistical analysis and decision analysis of manufacturing industry multivalued value chain cooperation. Therefore, the collaborative statistical analysis and decision analysis of the multi-value chain of the manufacturing enterprise based on the data space are realized, and a proposal is put forward for the production, operation and management of the manufacturing enterprise.
A manufacturing production intelligent computing method based on data space multi-value chain statistical analysis comprises the following steps:
step 1: determining influence factors of an internal value chain and an external value chain based on a data space multi-value chain cooperative system architecture; the internal value chain is a production link, and the external value chain comprises a supply link, a marketing link and a service link;
and 2, step: screening and analyzing the influence factors of the external value chain by adopting a random forest algorithm (RF) to obtain the contribution degrees of different influence factors in respective links, and reconstructing the influence factors according to the contribution degrees to obtain comprehensive influence factor indexes of the respective links;
and step 3: carrying out data accumulation smoothing treatment on the constructed external value chain comprehensive influence factor indexes and the internal value chain statistical analysis decision indexes to obtain accumulated comprehensive influence factor indexes of each link of the external value chain and internal value chain accumulated statistical analysis decision indexes;
and 4, step 4: taking the internal value chain accumulated statistical analysis decision index as an output variable, taking the accumulated comprehensive influence factor indexes of all links of the external value chain as input variables to establish an initial model, and training the initial model by using a longicorn stigma optimized convolutional neural network to obtain a statistical analysis decision model;
and 5: and calculating the production capacity of the manufacturing industry by using the obtained statistical analysis decision model to make decisions.
Further, the influencing factors of the supply chain include raw material purchase amount, raw material purchase cost, raw material inventory, ex-warehouse cost, raw material freight, raw material usage amount, raw material price, fund raising and borrowing interest; the marketing chain influence factors comprise sales volume, sales cost, sales income, sales gross profit and sales price; the service chain influence factors comprise overhaul cost, overhaul times, customer number and product percent of pass.
Further, the index of the comprehensive influence factor of each link of the external value chain is calculated by the following formula
Wherein i represents different links of the external value chain, g representing the second in a link of the external value chain g The number of the influencing factors is increased, m representing something in the middle of the external value chain m An influencing factor, d ig Represents the outer value chain link g Sequence data of an influencing factor, w ig Representing the external value chain i link g Degree of contribution of an influencing factor, D i Represents the comprehensive influence factor sequence data of the links of the external value chain i.
Further, when the initial model is trained, the selected activation function is SPReLu, and the mathematical form of the activation function is SPReLu
Wherein a is a random parameter, changes according to the real-time training of the model, and finally converges to a proper constant.
Further, a may take a constant of 0.3, 0.5, or 0.8.
Further, when the initial model is trained by using the activation function, before 10000 iterations, an optimal high learning rate is searched by using an optimization algorithm in an interval (0.0001, 0.01) to accelerate the speed of iterative update and find a global approximate optimal solution; and then, an optimal low learning rate is searched from (0.0001, 0.01) by using an optimization algorithm to obtain a global optimal solution.
Further, the specific steps of optimizing the convolutional neural network by using the longicorn whisker algorithm comprise:
step 1: initializing a weight value by the network;
step 2: the input data is transmitted forward through a convolution layer, a down-sampling layer and a full connection layer to obtain an output value;
step 3: calculating the error between the output value of the network and the target value;
step 4: when the error is larger than the expected value, the error is transmitted back to the network, and the errors of the full connection layer, the lower sampling layer and the convolution layer are sequentially obtained; when the error is equal to or less than the expected value, finishing the training;
step 5: and updating the weight according to the obtained error, optimizing the learning rate and the activation function, and returning to Step 2.
The invention has the beneficial effects that: by the provided intelligent calculation method for the manufacturing industry production based on the data space multi-value chain statistical analysis, an effective way is provided for the high-quality development of the manufacturing industry, the statistical analysis decision level of the manufacturing industry is improved, the enterprise competitiveness is improved, the opportunity brought by information resources in the new industrial era is fully mastered, and a manufacturing industry multi-value chain production management mode driven by data insight is constructed, so that the manufacturing industry intelligent statistical analysis decision management is promoted. The method can adapt to the specific requirements and changes of the market environment, accurately analyze various factors possibly appearing in the process of statistical analysis and decision, and analyze the cooperative relationship in the process from links of polyvalent values such as production, supply, marketing and service. Meanwhile, distributed heterogeneous data of a manufacturing enterprise is utilized, the data is associated with intelligent application, a mass data space technology is fully utilized, and a multi-value chain collaborative statistical analysis decision model is created by combining industrial chain fusion.
Drawings
FIG. 1 is a data space based multi-value chain statistical analysis manufacturing production intelligent computation method model framework diagram;
FIG. 2 is a reference diagram of the activation function SPReLU;
FIG. 3 is a detailed flow chart of the BAS-optimized CNN training algorithm;
FIG. 4 is a comparison of statistical analysis decision models under different training algorithms;
FIG. 5 is a graph comparing the error of the statistical analysis decision model under different training algorithms.
Detailed Description
The embodiments are described in detail below with reference to the accompanying drawings.
The current research methods for the architecture of the multivalent value chain system comprise system dynamics, a life cycle method, a comprehensive evaluation method and the like. System dynamics is a method of system modeling and dynamic simulation. It establishes a relatively effective model through the complex relation between the elements of the system. However, the system dynamics are not fine enough to describe the relevant problems with sufficient accuracy. The life cycle theory of an enterprise considers that the operation management mode and investment decision of the enterprise are different in different life cycle stages, and the value chain research in the enterprise is also influenced. The life cycle method can completely track the system and solve the global problem. However, the life cycle method is long in flow and needs much time. The comprehensive evaluation method is a method for evaluating a plurality of evaluation units by using a plurality of indexes, and can evaluate the related conditions of enterprises in a diversified manner. However, due to the limitations of availability of the index and subjectivity, it is difficult for the comprehensive evaluation method to completely and objectively reflect the characteristics of the object to be evaluated.
Currently, methods for researching value chains of enterprises and screening influence factors related to a plurality of business processes mainly include a factor analysis method, a grey correlation analysis method, a PCA method and the like. The factorial analysis is required for the number of samples. The excessive sample amount can cause the waste of manpower, material resources and financial resources; if the sample size is too small, sampling errors are increased, and the reliability of the analysis result is influenced. The grey correlation analysis method is also a common influence factor screening method and is commonly used for determining the production management characteristic analysis of enterprises and the sequencing of related influence factors. However, the correlation degree obtained by the gray correlation degree quantization model is a positive value, and this does not fully reflect the relationship between the objects. Meanwhile, the method cannot solve the problem of repeated evaluation information caused by correlation among evaluation indexes, so that the selection of the indexes has great influence on the evaluation result. Principal Component Analysis (PCA) is widely used for feature extraction and dimensionality reduction of pattern recognition and data analysis, and is commonly used for screening influence factors of various value chains of enterprises. Although the principal component analysis method has the advantages of simplicity, convenience and quickness, the positive and negative coefficients of each index variable in the first principal component may not accord with the actual evaluation meaning, and the naming definition is low.
Currently, the research on enterprise multi-value chains focuses more on the service range of a single value chain, and the research on the cooperative relationship among the multi-value chains is less. The method mainly adopts a research method for establishing a regression prediction model in the research of analyzing the correlation among the value chain factors of different business processes. There are many methods for establishing such a model, including a multiple linear regression method, a machine learning method, an elastic network method, and the like. Multiple Linear Regression (MLR) has the advantages of simplicity and interpretability. It can be used in the study of interrelationships. The linear regression model is simple in form and easy to model, but the assumption of the linear regression equation is strict, all factors of the explanatory variables causing the dependent variable change need to be known, otherwise problems such as pseudo regression and the like occur. The machine learning technology can help people to find rules from mass data and form a regression prediction model. However, the methods involved in machine learning are various, and it is necessary to select a method from a data set in different fields, and machine learning methods are often prone to problems such as overfitting, under-fitting, and time and labor consuming. The elastic network is a regularized least squares regression method, and is a convex combination of Ridge regression and Lasso regression. Elastic networks have found less application in enterprise-related research. When the elastic network falls into the local minimum, the network will lose the ability to continue learning and thus cannot get the optimal solution, which is an important drawback of this method.
According to the above researches, the current research on a single value chain of multiple value chains forms a relatively complete research system, but a complete research system is not formed yet for the cooperative relationship among the multiple value chains of manufacturing enterprises, the construction of the cooperative architecture of the multiple value chains of the manufacturing enterprises is basically blank, and the application of an intelligent data management technology of a data space in the multi-value chain architecture is also basically blank; analysis is rarely carried out from the perspective of the multi-value chain in the research of statistical analysis decision, the selection method of the multi-value chain influence factors focuses more on factor analysis, component analysis and the like, and objective analysis is carried out by less using an intelligent algorithm; in the research method of the synergy of the multivalent value chain, a regression analysis method is mostly used, and a deep learning algorithm is less used.
The planned data space multi-value chain cooperative system architecture carries out manufacturing industry statistical analysis decision analysis, the multi-value chain cooperative relationship in the multi-value chain cooperative system architecture is utilized, the influence factors of the multi-value chain are screened and reconstructed through a random forest algorithm, and statistical analysis decision is carried out through the optimized CNN of the longicorn silk, so that a three-stage (RF-BAS-CNN) statistical analysis decision model is constructed. FIG. 1 shows a data space multi-value chain based statistical analysis manufacturing production intelligent computation method model framework of the invention.
And constructing a data space multi-value chain cooperative system.
The method is based on the principles of integrity, hierarchy, expandability, manufacturing applicability, multi-value chain cooperativity, full-process property and multi-source heterogeneity, forms a data space multi-value chain cooperative system framework according to the characteristics of a manufacturing multi-value chain cooperative system of a data space, analyzes the mutual relation among production, supply, marketing and service value chains and the influence of an external value chain on an internal value chain, and determines a value chain angle analysis object of a manufacturing statistical analysis decision. Finally, in the production link of the multivalence value chain, the quantity (POQ) of production orders is determined as a research variable of statistical analysis decision; in the supply link of the multivalence value chain, 9 influence factors of raw material purchase amount (RMPV), Raw Material Purchase Cost (RMPC), Raw Material Inventory (RMI), ex-warehouse cost (OC), Raw Material Freight (RMF), raw material usage amount (RMU), Raw Material Price (RMP), fund raising (F) and borrowing interest (B) are finally selected, and the actual supply situation of the electric power manufacturing enterprise is reflected in the aspects of raw material supply, raw material use situation, fund supply and the like; in the marketing link of the multivalent value chain, 5 influence factors of Sales Volume (SV), sales cost (C), sales income (R), sales gross profit (G) and sales price (P) are selected as influence factors of the marketing link, and the marketing conditions of the power manufacturing enterprises can be reflected by the variables from the aspects of marketing income and expenditure conditions, profit conditions and the like; in the service link of the multivalence value chain, 4 influence factors of Maintenance Cost (MC), maintenance times (OT), customer Number (NC) and product qualification Rate (RQP) are selected to reflect the service condition of the power manufacturing industry to the customers and the influence of service to production, so that the service link in the multivalence value chain is formed.
The invention plans the manufacturing industry multi-value chain cooperative system architecture based on the data space, relates to the manufacturing industry multi-value chain data of a data service layer in the planning of the system architecture, and utilizes the data to carry out the statistical analysis and decision to obtain the basic internal relation of the manufacturing industry production and operation. From the construction of a system architecture, among multiple value chains, the supply, marketing and service links of an external value chain have important influence on the production and operation links inside an enterprise, and the enterprise can normally produce and operate and depends on the mutual cooperative relationship between the internal value chain and the external value chain.
And screening comprehensive influence factors of the multi-value chain.
Each value chain in the multiple value chains has multiple influence factors, the significance and influence of each value chain are different, and in order to better embody the collaborative relationship among the multiple value chains, the influence factors in the single value chain need to be analyzed. Therefore, before input variable data is determined, related influence factors need to be screened, and influence factors with the largest influence and the most significant influence are found out to serve as input variables, so that time and labor consumption can be avoided, and decision efficiency can be improved. The random forest algorithm is a method for comprehensively classifying data by utilizing a plurality of decision trees, can be used for screening influence factors, and has the advantages of high training speed, easiness in implementation of a parallelization method, capability of detecting mutual influence among the influence factors and the like. Based on the method, the influence factors are screened and analyzed in the supply, marketing and service links of the manufacturing industry multi-value chain by adopting a random forest algorithm to obtain the contribution degrees of different influence factors in the value chain, the influence factors are reconstructed according to the contribution degrees to obtain the multi-value chain comprehensive influence factors, and three comprehensive influence factor indexes of accumulated supply comprehensive influence factor index (C _ ICSU), accumulated marketing comprehensive influence factor index (C _ ICM) and accumulated service comprehensive influence factor index (C _ ICSE) are formed.
Step 1: and calculating the contribution degree of each influence factor in the supply, marketing and service links in the external value chain to the value chain by a random forest algorithm.
(1) When random forest
And (3) after the establishment is finished, assuming that k trees are in total in the random forest algorithm, randomly sampling data of a certain scale from the data set N as samples through random resampling aiming at each tree, forming a sample training subset, and forming p pieces of data outside bags by using the data which are not extracted. For n regression trees T in forest n And the corresponding data outside the bag is recorded as OOB n Then there is
When the raw data amount N is sufficiently large, k ≈ 0.368N. Using a regression Tree T n For OOB n Predicting to obtain OOB n Predicted mean square error of
In the formula: y is i Is OOB n The ith real measurement of the medium response variable;
is OOB n The ith predicted value of the re-responsive variable.
Thus, for the entire random forest, the predicted mean square error of ntree out-of-bag data is available, i.e.
[MSE 1 ...MSE n ...MSE ntree ] (3)
(2) For independent variable X j (1 ≦ j ≦ p), j representing the jth data among the p out-of-bag data that have not been extracted; keeping other lines of data unchanged in ntree OOB samples, randomly replacing jth line of data in each sample to form ntree new OOB samples, calculating the predicted mean square error of each new OOB sample to obtain [ MSE [ 1 ...MSE n ...MSE ntree ]Then for all independent variables X j (for each impact factor in each value chain) the mean square error matrix is predicted as follows
By [ MSE ] 1 ...MSE n ...MSE ntree ]MSE [ with respect to the above matrix 1 ...MSE n ...MSE ntree ]Subtracting the row vectors, averaging, and dividing by the standard error to obtain the independent variable X j The importance score (increment of mean squared error, IncMSE), i.e.
In the formula: t is t i The ith measured value of the response variable in the data outside the bag;
mean values of response variables in the out-of-bag data.
Step 2: the contribution degree of each influence factor in different links of the external value chain in the link is calculated through a random forest algorithm, the contribution degree can reflect the importance of the influence factor in the link in the external value chain, the influence factor of the external value chain can be reconstructed through the importance, and a comprehensive influence factor index capable of representing the different links of the external value chain is established. The specific calculation method is as follows:
wherein i represents different links of the external value chain, g representing the first in a link of the external value chain g The number of the influencing factors is increased, m representing something in the middle of the external value chain m An influencing factor, d ig Represents the outer value chain link g Sequence data of an influencing factor, w ig Representing the external value chain i link g Degree of contribution of an influencing factor, D i Represents the comprehensive influence factor sequence data of the links of the external value chain i.
And performing accumulated smoothing processing on the influence factor index data. Because the influence factors related by the method are numerous, and the data values of some influence factors have large fluctuation and are easy to influence the final result of the model, the method adopts an accumulative smooth mode to process the data of the influence factor sequence with large data fluctuation. The method can effectively smooth data, so that the data change has a basic rule changing along with time, thereby providing a good data basis for subsequent statistical analysis decision. The method can improve the research accuracy and realize effective data smoothing. Based on the method, the invention carries out data accumulation smoothing treatment on the constructed external value chain comprehensive influence factor index and the internal value chain statistical analysis decision index to form the input variable and the output variable of the statistical analysis decision model.
According to the invention, the contribution degree of each factor of different links is obtained through the research on the links such as the supply, marketing and service of an external value chain by a random forest algorithm. In the supply link, the influence degree of the raw material purchasing quantity and the raw material price on the whole supply link is higher, and the influence degree of the raw material inventory and the raw material purchasing cost on the supply link is weaker; in the marketing link, the marketing link is influenced by the sales volume and the sales income to a higher degree, and the marketing degree of the sales cost is smaller; in the service link, the contribution degree of several factors is basically kept at the same level, wherein the influence degree of the overhaul times is higher. Therefore, in order to ensure the effectiveness and rationality of the production operation of the internal value chain of the manufacturing enterprise, the purchase quantity and purchase price of raw materials, the sales quantity and sales income and the overhaul times of customer service should be effectively controlled in the external value chain link.
By constructing external comprehensive influence factor indexes, each index can be stably changed. In the production and operation process of enterprises, the external value chain indexes are controlled within a reasonable range, so that the production and operation can be effectively and stably carried out, and the stability of the value chain in the manufacturing enterprises is guaranteed.
And constructing a statistical analysis production decision model based on data space multi-value chain cooperation.
And taking the accumulated supply comprehensive influence factor index (C _ ICSU), the accumulated marketing comprehensive influence factor index (C _ ICM) and the accumulated service comprehensive influence factor index (C _ ICSE) which are selected and indicated by Random Forest (RF) screening as inputs, and taking the accumulated production order number (namely the accumulated statistical analysis decision index of the internal value chain) as an output to construct an initial model. The model is trained by using a Convolutional Neural Network (CNN) optimized by using the longicorn silk, and finally an RF-BAS-CNN model, namely the statistical analysis decision model of the invention, is obtained, and then the model is used for predicting the actual production condition.
Convolutional Neural Networks (CNN) are a class of feed forward Neural Networks (fed forward Neural Networks) that contain convolution computations and have a deep structure, and are one of the representative algorithms for deep learning (deep learning). Convolutional Neural Networks have a feature learning (rendering) capability, and can perform a Shift-Invariant classification on input information according to their hierarchical structure, and are therefore also called "Shift-Invariant Artificial Neural Networks (SIANN)". The invention uses the longicorn whisker algorithm to optimize the convolutional neural network. The Beatle antenna Search-BAS (Beetle antenna Search-BAS), also called Beetle antenna Search, is an efficient intelligent optimization algorithm proposed in 2017, and has the advantages of very small computation amount, very fast convergence and global optimization capability.
Combining the characteristics of several classes of activation functions, ReLU, PReLU and Softplus, the invention provides a new activation function, SPReLu, whose mathematical form is
Wherein a is a random parameter, changes according to the real-time training of the model, and finally converges to a proper constant. The value of a can be constant 0.3, 0.5, 0.8.
The function has the following characteristics that when x is larger than or equal to 0, the linear characteristic of Re-Lu is kept, and an output result and input data are kept unchanged; when x is less than 0, the curve of the Softplus function is shifted down by ln2 units, and the negative semi-axis curve is taken and multiplied by the parameter a.
The image of the sprellu function is shown in fig. 2. Where the parameter a effectively controls the saturation range of the function, it can be trained by back propagation and optimized simultaneously with other layers. A certain layer a i Has a gradient of
The momentum method is adopted when the gradient is updated:
where μ is the momentum coefficient and α is the learning rate.
For the CNN algorithm, the learning rate has a great influence on the iteration rate and the optimal solution. When the set learning rate is high, the network iteration is updated quickly, the optimal solution can be found from the global aspect, but the problem of difficult convergence exists. When the set learning rate is low, the network iterative update speed is slow, a local optimal value can be found, but sometimes the network falls into local optimization and loses the capability of obtaining a global optimal value. Therefore, when training the network, the two are often combined, a global approximately optimal solution is obtained by using a high learning rate, and a final optimal value is obtained by using a low learning rate.
The initial learning rate set in the BAS-CNN of the present invention is 0.01, and then the learning rate decreases to one tenth of the original one per 10000 iterations. However, the learning rate selected empirically may not be the optimal value, so the setting of the learning rate is replanned herein. Before 10000 iterations, an optimal high learning rate is searched in an interval (0.0001, 0.01) by using an optimization algorithm so as to accelerate the speed of iterative updating and find a global approximate optimal solution; therefore, the BAS optimization algorithm is adopted to set the learning rate of the CNN optimization algorithm as a search space, find the learning rate parameter value which enables the fitting effect to be optimal, and substitute the learning rate parameter value into the model training process to finish model training.
The specific steps for optimizing CNN based on BAS in model training are shown in fig. 3:
step 1: initializing a weight value by the network;
step 2: the input data is transmitted forwards through a convolution layer, a down-sampling layer and a full-connection layer to obtain an output value;
step 3: calculating the error between the output value of the network and the target value;
step 4: and when the error is larger than the expected value, the error is transmitted back to the network, and the errors of the full connection layer, the lower sampling layer and the convolution layer are sequentially obtained. The error of each layer can be understood as the total error of the network, and the network can bear the load; when the error is equal to or less than the desired value, the training is ended.
Step 5: and updating the weight according to the obtained error, optimizing the learning rate and the activation function, and returning to Step 2.
And after the RF-BAS-CNN model is established, the effectiveness of the statistical analysis decision model construction is judged. The invention carries out empirical analysis aiming at the model. The output result of the model is a trained decision value of the accumulated production order quantity, and the decision value is compared and analyzed with the actual production order quantity. The invention selects BPNN, BAS-BPNN, RNN, BAS-RNN and CNN training algorithms as comparison algorithms of the BAS-CNN algorithm, and selects MSE (Mean Squared error), MAE (Mean Squared error), MAPE (Mean Squared Percentage error), MSPE (Mean Squared error), RMSE (root Mean Squared error) and SSE (sum of Square for error) as objective index values to carry out error analysis on the prediction result, so as to judge the effectiveness of the statistical analysis decision model construction. The comparative results are shown in fig. 4 and 5 and table 1:
TABLE 1C _ POQ error analysis
From the above comparison, it can be seen that the models created by the present invention have high fitness and effectiveness. As can be seen from the error analysis of the model comparison graph and the prediction result of C _ POQ, the experimental result obtained by constructing the RF-BAS-CNN model is obviously superior to other results. Under six error indexes, the error values of the prediction results of the BAS-CNN algorithm are all the smallest, the matching effect of the prediction curve obtained by the model and the actual value curve is also good, the error is the prediction result of the BP model with the largest error, then the multiple linear regression prediction result, then the RNN prediction result, the BAS-BP prediction result and the CNN prediction result are the largest. Therefore, the model constructed by the invention has higher fitting performance on the cooperative relation between the internal value chain and the external value chain.
According to the method, the RF-BAS-CNN model is constructed, the decision analysis is effectively carried out on the production order of the internal value chain, the high fitting degree and the effectiveness of the model can be seen through the comparison of the decision value output by the model and the actual value, and the error of the decision result of the production order is 0.0058 in terms of MAPE (map algorithm). Therefore, the model constructed by the method has higher fitting performance on the cooperative relationship between the internal value chain and the external value chain, provides an effective analysis model for carrying out statistical analysis decision on the internal value chain under the consideration of the cooperation of multiple value chains by manufacturing enterprises in the future, and has important significance.
In conclusion, the invention has the following innovations:
the manufacturing industry multivalence value chain collaborative concept based on the data space is put forward for the first time, and a new idea is provided for management and management among value chains in the manufacturing industry. The former research is based on a single value chain, and does not consider the concept of the synergy of the multivalent value chain, but the invention systematically provides the concept of the synergy of the multivalent value chain, researches the synergy relationship between the value chains and forms a set of theoretical basis.
The invention constructs a statistical analysis decision model based on a data space multi-value chain cooperative system framework based on deep learning intelligent calculation. And taking the mutual relation among the links of production, supply, marketing, service and the like into consideration to form a statistical analysis decision model with mutual linkage, mutual cooperation, data sharing and safety management. In the research of predecessors, the application of an intelligent algorithm to value chain analysis is not a lot, analysis from the angle of a multi-value chain is rarely performed in the research of statistical analysis decision, the selection method of the multi-value chain influence factors focuses more on factor analysis, component analysis and the like, and the intelligent algorithm is rarely used for objective analysis; in the research method of the synergy of the multivalent value chain, a regression analysis method is frequently used, and a deep learning calculation method is less used. Based on the method, the three algorithms of the intelligent algorithm RF, the BAS and the CNN are combined to construct a statistical analysis decision model based on a data space multi-value chain cooperative system architecture. Wherein RF is used for screening and sequencing influencing factors of the multi-value chain, and the BAS-CNN model is used for statistical analysis and decision-making.
The invention firstly provides comprehensive influence factor indexes representing different links of the value chain, and the comprehensive influence factor indexes are used for representing the actual change conditions of the different links of the multivalence value chain. And macroscopically, a reference is provided for the manufacturing industry to analyze the actual conditions of different value chains. In the previous researches, a complete research system is not formed for the cooperative relationship among the manufacturing industry multivalent value chains, and the construction of the manufacturing industry multivalent value chain cooperative framework is basically blank. The invention summarizes the mutual relation among the production, supply, marketing and service value chains of the power enterprise and the influence of the external value chain on the internal value chain, pertinently provides the influence factors, and carries out sequencing and screening on the influence factors, establishes three comprehensive influence factor indexes for reflecting the change conditions of the value chains of different links, and lays a foundation for the related decision and evaluation of the future manufacturing industry.
The present invention is not limited to the above embodiments, and any changes or substitutions that can be easily made by those skilled in the art within the technical scope of the present invention are also within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (7)
1. A manufacturing production intelligent computing method based on data space multi-value chain statistical analysis comprises the following steps:
step 1: determining influence factors of an internal value chain and an external value chain based on a data space multi-value chain cooperative system architecture; the internal value chain is a production link, and the external value chain comprises a supply link, a marketing link and a service link;
step 2: screening and analyzing the influence factors of the external value chain by adopting a random forest algorithm (RF) to obtain the contribution degrees of different influence factors in respective links, and reconstructing the influence factors according to the contribution degrees to obtain comprehensive influence factor indexes of the respective links;
and 3, step 3: carrying out data accumulation smoothing treatment on the constructed external value chain comprehensive influence factor indexes and the internal value chain statistical analysis decision indexes to obtain the accumulated comprehensive influence factor indexes of each link of the external value chain and the internal value chain cumulative statistical analysis decision indexes;
and 4, step 4: taking the internal value chain accumulated statistical analysis decision index as an output variable, taking the accumulated comprehensive influence factor indexes of all links of the external value chain as input variables to establish an initial model, and training the initial model by using a longicorn stigma optimized convolutional neural network to obtain a statistical analysis decision model;
and 5: and calculating the production capacity of the manufacturing industry by using the obtained statistical analysis decision model to make decisions.
2. The method for statistically analyzing and intelligently calculating the production of the manufacturing industry according to the data space multiple value chains is characterized in that the influence factors of the supply chain comprise raw material purchase quantity, raw material purchase cost, raw material inventory, ex-warehouse cost, raw material freight, raw material usage quantity, raw material price, fund financing and borrowing interest; the marketing chain influence factors comprise sales volume, sales cost, sales income, sales gross profit and sales price; the service chain influence factors comprise overhaul cost, overhaul times, customer number and product percent of pass.
3. The method as claimed in claim 1, wherein the step2 comprises a step of analyzing the comprehensive influence factor indexes of the respective links of the external value chain
Calculated from the following equation
Wherein i represents different links of the external value chain, g represents the g-th influence factor in a certain link of the external value chain, m represents m influence factors in the certain link of the external value chain, and d ig Sequence data representing the g-th influencing factor of the i-link of the external value chain, w ig Representing the external value chain i linkDegree of contribution of the g-th influencing factor, D i Representing the comprehensive influence factor sequence data of the link of the external value chain i.
4. The method as claimed in claim 1, wherein the activation function SPReLu is selected during the training of the initial model in step3, and the mathematical form of the activation function SPReLu is SPReLu
Wherein a is a random parameter, changes according to the real-time training of the model, and finally converges to a proper constant.
5. The method of claim 4, wherein a is a constant of 0.3, 0.5, or 0.8.
6. The method for the intelligent calculation of the production of the manufacturing industry based on the statistical analysis of the data space multiple value chains as claimed in claim 4, wherein when the initial model is trained by using the activation function, before the iteration is carried out for 10000 times, an optimal high learning rate is found by using an optimization algorithm in an interval (0.0001, 0.01) so as to accelerate the updating speed of the iteration and find a global approximate optimal solution; and then, an optimal low learning rate is searched from (0.0001, 0.01) by using an optimization algorithm to obtain a global optimal solution.
7. The method of claim 4, wherein the step4 of optimizing the convolutional neural network by using the longicorn algorithm comprises the following specific steps:
step 1: initializing a weight value by the network;
step 2: the input data is transmitted forwards through a convolution layer, a down-sampling layer and a full-connection layer to obtain an output value;
step 3: calculating the error between the output value of the network and the target value;
step 4: when the error is larger than the expected value, the error is transmitted back to the network, and the errors of the full connection layer, the down sampling layer and the convolution layer are sequentially obtained; when the error is equal to or less than the expected value, finishing the training;
step 5: and updating the weight according to the obtained error, optimizing the learning rate and the activation function, and returning to Step 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210359177.5A CN114912074A (en) | 2022-04-07 | 2022-04-07 | Intelligent calculation method for manufacturing production based on data space multi-value chain statistical analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210359177.5A CN114912074A (en) | 2022-04-07 | 2022-04-07 | Intelligent calculation method for manufacturing production based on data space multi-value chain statistical analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114912074A true CN114912074A (en) | 2022-08-16 |
Family
ID=82763546
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210359177.5A Pending CN114912074A (en) | 2022-04-07 | 2022-04-07 | Intelligent calculation method for manufacturing production based on data space multi-value chain statistical analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114912074A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070230770A1 (en) * | 2005-11-18 | 2007-10-04 | Ashok Kulkarni | Methods and systems for determining a position of inspection data in design data space |
| CN109673034A (en) * | 2018-12-28 | 2019-04-23 | 中国科学院上海微***与信息技术研究所 | A kind of wireless sensor network cluster routing method that must be searched for based on longicorn |
| CN110378799A (en) * | 2019-07-16 | 2019-10-25 | 东北大学 | Aluminium oxide comprehensive production index decision-making technique based on multiple dimensioned depth convolutional network |
| CN114138766A (en) * | 2021-12-07 | 2022-03-04 | 华北电力大学 | Multi-value chain data system analysis architecture and integration cooperation method under data space |
-
2022
- 2022-04-07 CN CN202210359177.5A patent/CN114912074A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070230770A1 (en) * | 2005-11-18 | 2007-10-04 | Ashok Kulkarni | Methods and systems for determining a position of inspection data in design data space |
| CN109673034A (en) * | 2018-12-28 | 2019-04-23 | 中国科学院上海微***与信息技术研究所 | A kind of wireless sensor network cluster routing method that must be searched for based on longicorn |
| CN110378799A (en) * | 2019-07-16 | 2019-10-25 | 东北大学 | Aluminium oxide comprehensive production index decision-making technique based on multiple dimensioned depth convolutional network |
| CN114138766A (en) * | 2021-12-07 | 2022-03-04 | 华北电力大学 | Multi-value chain data system analysis architecture and integration cooperation method under data space |
Non-Patent Citations (3)
| Title |
|---|
| WENBO LI等: "Research on the network structure and organizational model of multi-value chain based on third-party platform", 《 2021 2ND INTERNATIONAL CONFERENCE ON E-COMMERCE AND INTERNET TECHNOLOGY (ECIT)》, 21 April 2021 (2021-04-21), pages 1 - 8 * |
| 吴庚奇等: "多价值链视角下基于深度学习算法的制造企业产品需求预测", 《 科学技术与工程 》, vol. 21, no. 31, 8 November 2021 (2021-11-08), pages 13413 - 13420 * |
| 李明钰等: "考虑多价值链协同的电力设备制造企业经营风险预测研究", 《工程管理科技前沿》, vol. 41, no. 3, 27 May 2022 (2022-05-27), pages 53 - 60 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mungle et al. | A fuzzy clustering-based genetic algorithm approach for time–cost–quality trade-off problems: A case study of highway construction project | |
| Li et al. | Optimal selection of heterogeneous ensemble strategies of time series forecasting with multi-objective programming | |
| US20210278825A1 (en) | Real-Time Production Scheduling with Deep Reinforcement Learning and Monte Carlo Tree Research | |
| US20180129961A1 (en) | System, method and computer-accessible medium for making a prediction from market data | |
| Alamsyah et al. | Artificial neural network for Indonesian tourism demand forecasting | |
| Taghizadeh | Utilizing artificial neural networks to predict demand for weather-sensitive products at retail stores | |
| CN113780667A (en) | Enterprise operation prediction method and system based on AI artificial intelligence | |
| Goyal et al. | A solution to forecast demand using long short-term memory recurrent neural networks for time series forecasting | |
| Zhu et al. | A fixed charge transportation problem with damageable items under uncertain environment | |
| Lee et al. | Automobile parts reliability prediction based on claim data: The comparison of predictive effects with deep learning | |
| Jiang et al. | Hybrid model for profit-driven churn prediction based on cost minimization and return maximization | |
| Klocker et al. | A machine learning approach for automated cost estimation of plastic injection molding parts | |
| Noorul Haq et al. | Effect of forecasting on the multi-echelon distribution inventory supply chain cost using neural network, genetic algorithm and particle swarm optimisation | |
| CN114912074A (en) | Intelligent calculation method for manufacturing production based on data space multi-value chain statistical analysis | |
| Fiorini et al. | A simple reinforcement learning algorithm for stock trading | |
| Zhang et al. | Decision making methods for formulating the reserve scheme of spare parts based on deep neural network | |
| Kotenko et al. | Supervised Machine Learning Models for Forecasting Fuel Consumption by Vehicles During the Grain Crops Delivery | |
| Khandelwal et al. | Financial Data Time Series Forecasting Using Neural Networks and a Comparative Study | |
| Yoon et al. | Methodology for improving the performance of demand forecasting through machine learning | |
| Polupanov et al. | Improving the neural network mathematical model of corporate bankruptcy | |
| Hu | Intelligent Financial Decision Support System Using Artificial Intelligence | |
| Yunianto et al. | Comparison of Stock Price Predictions Using Support Vector Regression and Recurrent Neural Network Methods | |
| Toluevs | ILYA JACKSON | |
| UmaDevi et al. | Fostering a Predictive Model with Neural Networks Trading in the Forex Market using AI Technology | |
| Salinas-Guerra et al. | An Evolutionary Bilevel Optimization Approach for Neuroevolution |
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 |