CN116050600A - CNN-GA-BP-based combined model spare part demand prediction method and system - Google Patents

CNN-GA-BP-based combined model spare part demand prediction method and system Download PDF

Info

Publication number
CN116050600A
CN116050600A CN202211729757.5A CN202211729757A CN116050600A CN 116050600 A CN116050600 A CN 116050600A CN 202211729757 A CN202211729757 A CN 202211729757A CN 116050600 A CN116050600 A CN 116050600A
Authority
CN
China
Prior art keywords
cnn
prediction
layer
model
network
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.)
Granted
Application number
CN202211729757.5A
Other languages
Chinese (zh)
Other versions
CN116050600B (en
Inventor
张红旗
曹锐
黄国兴
曹先怀
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
CETC 38 Research Institute
Original Assignee
CETC 38 Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by CETC 38 Research Institute filed Critical CETC 38 Research Institute
Publication of CN116050600A publication Critical patent/CN116050600A/en
Priority to PCT/CN2023/094438 priority Critical patent/WO2024139014A1/en
Application granted granted Critical
Publication of CN116050600B publication Critical patent/CN116050600B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • 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
    • G06N3/08—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
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • 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
    • G06Q10/20—Administration of product repair or maintenance
    • 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)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Software Systems (AREA)
  • Human Resources & Organizations (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computational Linguistics (AREA)
  • Economics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Strategic Management (AREA)
  • Operations Research (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Quality & Reliability (AREA)
  • Marketing (AREA)
  • Physiology (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Computational Mathematics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Tourism & Hospitality (AREA)
  • Development Economics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Algebra (AREA)
  • Game Theory and Decision Science (AREA)

Abstract

The invention provides a CNN-GA-BP-based combined model spare part demand prediction method, a CNN-GA-BP-based combined model spare part demand prediction system, a CNN-GA-BP-based combined model spare part demand prediction storage medium and electronic equipment, and relates to the technical field of electronic equipment spare part prediction. In the invention, a combined prediction model of a CNN (CNN) network and a BP (back propagation) neural network is combined, wherein the BP neural network is optimized by adopting a GA (genetic algorithm), and a regression model for predicting future spare part demands of equipment by using historical demand data of the spare parts is established. Comprising the following steps: firstly, three types of individual direct prediction models are predicted, then a CNN (computer numerical network) is used for convolving a direct prediction result, and then a back propagation BP neural network is optimized based on a GA (genetic algorithm) to form a three-layer combined prediction model which is superior to the individual combined model and the individual direct prediction model, and finally, the prediction performance is estimated by utilizing historical demand data of electronic equipment spare parts. The prediction result proves the superiority of the combined prediction method; the method can effectively provide a new way for predicting the needs of the medium-and-long-term spare parts, so that the management of the internal inventory of the enterprise is optimized.

Description

CNN-GA-BP-based combined model spare part demand prediction method and system
Technical Field
The invention relates to the technical field of electronic equipment spare part prediction, in particular to a CNN-GA-BP-based combined model spare part demand prediction method, a CNN-GA-BP-based combined model spare part demand prediction system, a CNN-GA-BP-based combined model spare part demand prediction storage medium and electronic equipment.
Background
Since the 21 st century, the electronic industry has been developing day-to-day, and the proportion of the electronic manufacturing industry in China in the market has been significantly improved. Thus, the problem of electronic related equipment spare part inventory control is related, one basic element of spare part inventory control is accurate spare part demand prediction, and the quality of electronic spare part demand prediction can have a significant effect on equipment spare part maintenance and guarantee cost and combat readiness.
The quantification of spare part consumption of electronic equipment is very complex, especially due to the non-linear, gray nature of the requirements. Electronic equipment spare parts suppliers typically empirically determine spare parts requirements; however, this is a random, blind process that can lead to a significant backlog of spare parts if the consumption is lower than expected and the critical spare parts are in shortage. The key to solving this problem is to accurately predict spare part requirements.
The accuracy of different prediction theory and methods is often directly affected by the manner in which they are applied. Thus, current demand forecast research and applications often encounter two major challenges: data is inconsistent with actual demands, and the prediction method is inaccurate to apply.
Disclosure of Invention
(one) solving the technical problems
Aiming at the defects of the prior art, the invention provides a CNN-GA-BP-based combined model spare part demand prediction method, a CNN-GA-BP-based combined model spare part demand prediction system, a CNN-GA-BP-based combined model spare part demand storage medium and electronic equipment, and solves the technical problem that spare part demands cannot be predicted accurately.
(II) technical scheme
In order to achieve the above purpose, the invention is realized by the following technical scheme:
a combined model spare part demand prediction method based on CNN-GA-BP is used for predicting the demand quantity of middle-long-term equipment spare parts, and in the training process of the combined model based on CNN-GA-BP:
s1, acquiring and vectorizing historical demand data of electronic equipment spare parts;
s2, acquiring a direct prediction result by adopting a bottom layer prediction model according to the historical demand data vector;
s3, splicing the historical demand data vector and the direct prediction result, and acquiring a convolution result by adopting a middle layer prediction model CNN network;
s4, splicing the historical demand data vector, the direct prediction result and the convolution result, and dividing the spliced result into a training set and a testing set; the training set is used as input of a top layer prediction model GA-BP network, and input data are trained;
s5, taking the test set as input of the trained GA-BP network, and obtaining a spare part demand prediction result.
Preferably, the bottom layer prediction model in S2 includes an exponential smoothing method and a differential autoregressive moving average model.
Preferably, the exponential smoothing method and the differential autoregressive moving average model include:
(1) Exponential smoothing method
Figure BDA0004031154410000031
wherein ,yt For the actual value of the t-th period, the time series consists of historical demand data of spare parts, 1<t<T, T is the total period number; alpha is a smoothing coefficient, 0<α<1；
Figure BDA0004031154410000032
The predicted value of the t-th phase i method is shown as a predicted value, wherein i=1 and 2 respectively represent a primary exponential smoothing method and a secondary exponential smoothing method;
is provided with
Figure BDA0004031154410000033
For the actual value of the i method at phase t+1, then the two exponential smoothing modes become:
Figure BDA0004031154410000034
Figure BDA0004031154410000035
wherein ,
Figure BDA0004031154410000036
the actual values of the T+1st stage of the primary and secondary exponential smoothing methods are respectively;
(2) Assume { X t And a group of weak stationary time sequences, which are subjected to d-level difference:
Figure BDA0004031154410000037
the differential autoregressive moving average model is expressed as:
W t ＝α 1 W t-1 +α 2 W t-2 …+α p W t-p -θ 1 β t-1 -θ 2 β t-2 …-θ p β t-p
wherein ,Wt The sequence is subjected to d-order differential processing; b is a backward shift operator and satisfies the condition B k X t ＝X t-k ；
Figure BDA0004031154410000038
Representing differential calculation, and having ∈>
Figure BDA0004031154410000039
X t Is the time sequence value at the t time; beta t Mean 0 and variance sigma a White noise of (a); alpha t 、θ t Is a model parameter.
Preferably, in the step S3, after the historical demand data vector and the direct prediction result are spliced, zero filling is performed first, and then a CNN network is adopted for convolution.
Preferably, the training process of the GA-BP network in S4 includes:
s10, initializing a network structure of a BP neural network and learning parameters; the network structure comprises an input layer, an implicit layer and an output layer, wherein the GA algorithm is adopted to execute learning parameter optimization, and the optimized weight and threshold value are obtained;
s20, inputting samples in a training set, and calculating the output of an hidden layer;
Figure BDA0004031154410000041
wherein ,Hj For the j-th node of the hidden layer, f is the excitation function, l is the node number of the input layer, omega ij For the weight value between the i node of the input layer and the j node of the hidden layer, x i An input value of an i-th node of the input layer, a j A threshold value for the j-th node of the hidden layer;
s30, forward propagation, calculating output O of an output layer k ；
Figure BDA0004031154410000042
wherein ,Ok For the output of the kth node of the output layer, q is the number of nodes of the hidden layer, omega jk B is the weight between the jth node of the hidden layer and the kth node of the output layer k A threshold value for the kth node of the output layer;
s40, calculating the actual output of the current network through the training set, comparing the actual output with the expected output of the network,
e k ＝Y k -O k
wherein ,ek For the output error of the kth node of the output layer, Y k A desired output for a kth node of the output layer;
s50, back propagation, correcting the weight of the neural network layer by layer according to the error,
Figure BDA0004031154410000043
wherein eta is the learning rate and m is the node number of the output layer;
s60, updating the threshold value, and correcting the threshold value layer by layer according to the error;
Figure BDA0004031154410000051
s70, judging whether iteration of the BP neural network algorithm is finished, if the number of times of iteration is specified or the difference value between two adjacent errors is smaller than a specified value, and outputting a predicted value after the iteration is finished
Figure BDA0004031154410000055
Otherwise, the process returns to S20.
Preferably, in the step S100, the learning parameter optimizing is performed by using a GA algorithm, which specifically includes:
s100, initializing a population;
s200, taking the training error of the BP neural network as an fitness function;
s300, repeatedly executing the operations of selection, crossing and mutation;
s400, stopping iteration until the individual fitness meets the condition or the iteration times reach the designated times, and outputting the optimal weight and the threshold value.
Preferably, four evaluation indexes are adopted to verify the prediction accuracy of the combined model based on CNN-GA-BP, and the method specifically comprises the following steps:
(1) Mean absolute error MAE:
Figure BDA0004031154410000052
(2) Mean square error MSE:
Figure BDA0004031154410000053
(3) Root mean square error RMSE:
Figure BDA0004031154410000054
(4) R2 determines the coefficient:
Figure BDA0004031154410000061
wherein n is the number of samples;
Figure BDA0004031154410000062
is a predicted value; y is t Is an actual value; />
Figure BDA0004031154410000063
Is the average of the actual values.
A CNN-GA-BP-based combined model spare part demand prediction system, which is used for predicting the demand quantity of middle-long-term equipment spare parts, and is used for training a CNN-GA-BP-based combined model in the process of training:
the vectorization module is used for acquiring and vectorizing historical demand data of the electronic equipment spare parts;
the direct prediction module is used for acquiring a direct prediction result by adopting a bottom layer prediction model according to the historical demand data vector;
the convolution module is used for splicing the historical demand data vector and the direct prediction result, and a middle layer prediction model CNN network is adopted to obtain a convolution result;
the training module is used for splicing the historical demand data vector, the direct prediction result and the convolution result, and dividing the spliced result into a training set and a testing set; the training set is used as input of a top layer prediction model GA-BP network, and input data are trained;
and the test module is used for taking the test set as the input of the trained GA-BP network and obtaining the spare part demand prediction result.
A storage medium storing a computer program for CNN-GA-BP-based combined model spare part demand prediction, wherein the computer program causes a computer to execute the CNN-GA-BP-based combined model spare part demand prediction method as described above.
An electronic device, comprising:
one or more processors;
a memory; and
one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs comprising means for performing a CNN-GA-BP based combined model spare part demand prediction method as described above.
(III) beneficial effects
The invention provides a CNN-GA-BP-based combined model spare part demand prediction method, a CNN-GA-BP-based combined model spare part demand prediction system, a CNN-GA-BP-based combined model spare part demand prediction storage medium and electronic equipment. Compared with the prior art, the method has the following beneficial effects:
in the invention, a combined prediction model of a CNN (CNN) network and a BP (back propagation) neural network is combined, wherein the BP neural network is optimized by adopting a GA (genetic algorithm), and a regression model for predicting future spare part demands of equipment by using historical demand data of the spare parts is established. Comprising the following steps: firstly, predicting three types of individual direct prediction models, secondly, convoluting direct prediction results by using a CNN (computer numerical network), and optimizing back propagation BP neural networks based on a GA (genetic algorithm) algorithm to form a three-layer combined prediction model which is superior to the individual combined model and the individual direct prediction model, and finally, evaluating the prediction performance by using historical demand data of equipment spare parts. The prediction result proves the superiority of the combined prediction method; the method can effectively provide a new way for predicting the needs of the medium-and-long-term spare parts, so that the management of the internal inventory of the enterprise is optimized.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is an overall framework diagram of a combined model spare part demand prediction method based on CNN-GA-BP provided by an embodiment of the invention;
FIG. 2 is a block diagram of a method for predicting the demand of a spare part of a combined model based on CNN-GA-BP according to an embodiment of the present invention;
FIG. 3 is a schematic overall flow chart of a method for predicting the demand of a spare part of a combined model based on CNN-GA-BP according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of direct prediction based on an exponential smoothing method and a differential autoregressive moving average model according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of splicing the convolved results according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a convolution process of a CNN network according to an embodiment of the present invention;
fig. 7 is a schematic diagram of a training flow of a BP neural network according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of a BP neural network according to an embodiment of the present invention;
FIG. 9 is a schematic diagram of a GA algorithm according to an embodiment of the present invention;
FIG. 10 is a plot of fitness function changes during three method iterations provided by an embodiment of the present invention;
fig. 11 shows predicted values and actual values of three methods according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention are clearly and completely described, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The embodiment of the application solves the technical problem that the spare part demand cannot be accurately predicted by providing the CNN-GA-BP-based combined model spare part demand prediction method, the CNN-GA-BP-based combined model spare part demand prediction system, the CNN-GA-BP-based combined model spare part demand storage medium and the CNN-GA-BP-based combined model spare part demand electronic equipment.
The technical scheme in the embodiment of the application aims to solve the technical problems, and the overall thought is as follows:
in the embodiment of the invention, a convolutional neural network (CNN, convolutional Neural Network) and a BP neural network (Back Propagation Neural Network) are combined, wherein the BP neural network is optimized by adopting a genetic algorithm (GA, genetic Algorithm), and a regression model for predicting future spare part demands of electronic equipment by using historical demand data of the spare parts is established. Comprising the following steps: firstly, predicting three types of individual direct prediction models, secondly, convoluting direct prediction results by using a CNN (computer numerical network), and optimizing back propagation BP neural networks based on a GA (genetic algorithm) algorithm to form a three-layer combined prediction model which is superior to the individual combined model and the individual direct prediction model, and finally, evaluating the prediction performance by using historical demand data of equipment spare parts. The prediction result proves the superiority of the combined prediction method; the method can effectively provide a new way for predicting the needs of the medium-and-long-term spare parts, so that the management of the internal inventory of the enterprise is optimized.
In order to better understand the above technical solutions, the following detailed description will refer to the accompanying drawings and specific embodiments.
Examples:
as shown in fig. 1-2, the embodiment of the invention provides a method for predicting the demand quantity of spare parts of a combined model based on CNN-GA-BP, which is used for predicting the demand quantity of spare parts of medium-long-term equipment, in the training process of the combined model based on CNN-GA-BP:
s1, acquiring and vectorizing historical demand data of equipment parts;
s2, acquiring a direct prediction result by adopting a bottom layer prediction model according to the historical demand data vector;
s3, splicing the historical demand data vector and the direct prediction result, and acquiring a convolution result by adopting a middle layer prediction model CNN network;
s4, splicing the historical demand data vector, the direct prediction result and the convolution result, and dividing the spliced result into a training set and a testing set; the training set is used as input of a top layer prediction model GA-BP network, and input data are trained;
s5, taking the test set as input of the trained GA-BP network, and obtaining a spare part demand prediction result.
The prediction result obtained by the prediction method proves the superiority of the combined prediction method; the method can effectively provide a new way for predicting the medium-and-long-term spare part demands, so that the internal inventory management of the enterprise is optimized.
As shown in fig. 3, the following will describe each step of the above technical solution in detail:
in step S1, historical demand data for the equipment piece is acquired and vectorized.
In the step, historical demand data (historical consumption data) of electronic equipment spare parts in nearly five years are collected, and corresponding vectors are obtained after preprocessing the data.
In step S2, according to the historical demand data vector, a bottom layer prediction model is adopted to obtain a direct prediction result.
The bottom layer prediction model comprises an exponential smoothing method (a primary and a secondary exponential smoothing method) and a differential autoregressive moving average model; comprising the following steps:
(1) Exponential smoothing method
Figure BDA0004031154410000101
wherein ,yt For the actual value of the t-th period, the time series consists of historical demand data of spare parts, 1<t<T, T is the total period number; alpha is a smoothing coefficient, 0<α<1；
Figure BDA0004031154410000111
The predicted value of the t-th phase i method is shown as a predicted value, wherein i=1 and 2 respectively represent a primary exponential smoothing method and a secondary exponential smoothing method;
is provided with
Figure BDA0004031154410000112
For the actual value of the i method at phase t+1, then the two exponential smoothing modes become:
Figure BDA0004031154410000113
Figure BDA0004031154410000114
wherein ,
Figure BDA0004031154410000115
the actual values of the T+1st stage of the primary and secondary exponential smoothing methods are respectively;
(2) Assume { X t And a group of weak stationary time sequences, which are subjected to d-level difference:
Figure BDA0004031154410000116
the differential autoregressive moving average model is expressed as:
W t ＝α 1 W t-1 +α 2 W t-2 …+α p W t-p -θ 1 β t-1 -θ 2 β t-2 …-θ p β t-p
wherein ,Wt The sequence is subjected to d-order differential processing; b is a backward shift operator and satisfies the condition B k X t ＝X t-k ；
Figure BDA0004031154410000117
Representing differential calculation, and having ∈>
Figure BDA0004031154410000118
X t Is the time sequence value at the t time; beta t Mean 0 and variance sigma a White noise of (a); alpha t 、θ t Is a model parameter.
And respectively predicting the time sequence by using a primary exponential smoothing method, a secondary exponential smoothing method and a differential autoregressive moving average model through a bottom layer prediction model, and respectively predicting to obtain a prediction result as shown in fig. 4.
In fig. 4: predicting the first twelve data of the initial data row by using an exponential smoothing method to obtain first data of the LES row in figure 2, and the like to obtain the whole LES row data; similarly, two to thirteen data and three to fourteen data of the initial data line are respectively predicted by a secondary exponential smoothing method and a differential autoregressive moving average model (Autoregressive Integrated Moving Averagemodel, ARIMA) to obtain first data of the SES line and the ARIMA line of the graph, and the like to obtain data of the whole SES line and the whole ARIMA line.
In step S3, the historical demand data vector and the direct prediction result are spliced, and a middle layer prediction model CNN network is adopted to obtain a convolution result.
Particularly, in the step, zero filling is performed first after the historical demand data vector and the direct prediction result are spliced, and then a CNN network is adopted for convolution, so that the defect of edge data characteristics can be avoided. The convolution process is shown in fig. 5, and the result of the convolution and splicing is shown in fig. 6.
In step S4, the historical demand data vector, the direct prediction result and the convolution result are spliced, and the spliced result is divided into a training set and a testing set; and taking the training set as the input of the top layer prediction model GA-BP network, and training the input data.
The training process of the GA-BP network is shown in fig. 7, and comprises the following steps:
s10, initializing a network structure and learning parameters of the BP neural network.
As shown in fig. 8, the network structure includes an input layer, an implicit layer and an output layer, and performs learning parameter optimization by using GA algorithm to obtain optimized weights and thresholds;
s20, inputting samples in a training set, and calculating the output of an hidden layer;
Figure BDA0004031154410000121
wherein ,Hj For the j-th node of the hidden layer, f is the excitation function, l is the node number of the input layer, omega ij For the weight value between the i node of the input layer and the j node of the hidden layer, x i An input value of an i-th node of the input layer, a j A threshold value for the j-th node of the hidden layer;
s30, forward propagation, calculating output O of an output layer k ；
Figure BDA0004031154410000131
wherein ,Ok For the output of the kth node of the output layer, q is the number of nodes of the hidden layer, omega jk B is the weight between the jth node of the hidden layer and the kth node of the output layer k A threshold value for the kth node of the output layer;
s40, calculating the actual output of the current network through the training set, comparing the actual output with the expected output of the network,
e k ＝Y k -O k
wherein ,ek For the output error of the kth node of the output layer, Y k A desired output for a kth node of the output layer;
s50, back propagation, correcting the weight of the neural network layer by layer according to the error,
Figure BDA0004031154410000132
wherein eta is the learning rate and m is the node number of the output layer;
s60, updating the threshold value, and correcting the threshold value layer by layer according to the error;
Figure BDA0004031154410000133
s70, judging whether iteration of the BP neural network algorithm is finished, if the number of times of iteration is specified or the difference value between two adjacent errors is smaller than a specified value, and outputting a predicted value after the iteration is finished
Figure BDA0004031154410000134
Otherwise, the process returns to S20.
Specifically, as shown in fig. 9, in S100, the learning parameter optimization is performed by using a GA algorithm, which specifically includes:
s100, initializing a population;
s200, taking the training error of the BP neural network as an fitness function;
s300, repeatedly executing the operations of selection, crossing and mutation;
s400, stopping iteration until the individual fitness meets the condition or the iteration times reach the designated times, and outputting the optimal weight and the threshold value.
In step S5, the test set is used as an input of the trained GA-BP network, and a spare part demand prediction result is obtained.
And carrying out regression prediction by using the trained GA-BP network which obtains the optimal weight. And using the test set for performance evaluation to obtain a spare part demand prediction result.
In addition, the embodiment of the invention also adopts four evaluation indexes to verify the prediction accuracy of the combined model based on the CNN-GA-BP, and specifically comprises the following steps:
(1) Mean absolute error MAE:
Figure BDA0004031154410000141
(2) Mean square error MSE:
Figure BDA0004031154410000142
(3) Root mean square error RMSE:
Figure BDA0004031154410000143
(4) R2 determines the coefficient:
Figure BDA0004031154410000151
wherein n is the number of samples;
Figure BDA0004031154410000152
is a predicted value; y is t Is an actual value; />
Figure BDA0004031154410000153
Is the average of the actual values.
And the top-layer prediction model adopts a BP neural network based on GA optimization to predict the spare part demand. The embodiment of the invention selects a three-layer neural network, and can be theoretically fit with any linear function. The activate function selects the common Sigmoid function. Only considering the influence of the number of hidden layer neurons on the performance of the BP neural network, along with the increase of the number of neurons, the prediction precision is improved, but the training time is also increased. Prediction accuracy and training time are weighed. Considering comprehensively, the embodiment of the invention selects 10 neuron numbers in one hidden layer. Genetic parameters are selected, population scale is set to 100, crossover probability and mutation probability are set to 0.3 and 0.1 respectively, and the maximum iteration number is set to 50. And selecting the root mean square error of the neural network as the fitness function of the genetic algorithm, and obtaining the optimal fitness function through 50 iterations.
In order to verify the effectiveness of the spare part demand combination prediction method of the convolutional neural network and the BP neural network based on genetic algorithm optimization provided by the embodiment of the invention, the CNN-GA-BP combination model provided by the embodiment of the invention is compared with a single BP model or the CNN-BP model aiming at the spare part demand prediction result, and the specific prediction result is shown in table 1.
Table 1 effect table of different models for spare part demand prediction
Figure BDA0004031154410000154
As can be seen from table 1, four evaluation index values of the CNN-GA-BP method are lower than those of the other two methods, wherein compared with the BP method, the CNN-GA-BP method reduces the Mean Absolute Error (MAE) by 13.8%, the Mean Square Error (MSE) by 24.87%, the Root Mean Square Error (RMSE) by 13.33%, the R2 coefficient by 1.99%, and the prediction accuracy is improved greatly; compared with the CNN-BP method, the CNN-GA-BP reduces the Mean Absolute Error (MAE) by 6.6%, reduces the Mean Square Error (MSE) by 20.13%, reduces the Root Mean Square Error (RMSE) by 10.64%, improves the R2 coefficient by 1.34%, and improves the prediction precision well.
As shown in fig. 10, the CNN-GA-BP method has a faster convergence speed and a smaller fitness function value than the BP neural network and the CNN-BP method.
As shown in fig. 11, the predicted value and the true value are obtained by using the BP, CNN-BP, and CNN-GA-BP prediction methods, respectively.
In conclusion, the electronic equipment spare part demand is predicted based on the CNN-GA-BP combined model, and compared with a single BP model or CNN-BP model, the method has the advantages of higher precision, higher convergence speed, more accurate prediction result and superiority.
The embodiment of the invention provides a CNN-GA-BP-based combined model spare part demand prediction system which is used for predicting the demand quantity of middle-long-term equipment spare parts, and in the training process of the CNN-GA-BP-based combined model, the demand quantity of the middle-long-term equipment spare parts is calculated by the CNN-GA-BP-based combined model demand prediction system:
the vectorization module is used for acquiring and vectorizing historical demand data of the electronic equipment spare parts;
the direct prediction module is used for acquiring a direct prediction result by adopting a bottom layer prediction model according to the historical demand data vector;
the convolution module is used for splicing the historical demand data vector and the direct prediction result, and a middle layer prediction model CNN network is adopted to obtain a convolution result;
the training module is used for splicing the historical demand data vector, the direct prediction result and the convolution result, and dividing the spliced result into a training set and a testing set; the training set is used as input of a top layer prediction model GA-BP network, and input data are trained;
and the test module is used for taking the test set as the input of the trained GA-BP network and obtaining the spare part demand prediction result.
An embodiment of the present invention provides a storage medium storing a computer program for CNN-GA-BP-based combined model spare part demand prediction, wherein the computer program causes a computer to execute the CNN-GA-BP-based combined model spare part demand prediction method as described above.
The embodiment of the invention provides electronic equipment, which comprises:
one or more processors;
a memory; and
one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs comprising means for performing a CNN-GA-BP based combined model spare part demand prediction method as described above.
In summary, compared with the prior art, the method has the following beneficial effects:
1. in the embodiment of the invention, a combined prediction model of a CNN (CNN) network and a BP (back propagation) neural network is combined, wherein the BP neural network is optimized by adopting a GA (genetic algorithm), and a regression model for predicting future spare part demands of equipment by using historical demand data of the spare parts is established. Comprising the following steps: firstly, three types of individual direct prediction models are predicted, then a CNN (computer numerical network) is used for convolving a direct prediction result, and then a back propagation BP neural network is optimized based on a GA (genetic algorithm) to form a three-layer combined prediction model which is superior to the individual combined model and the individual direct prediction model, and finally, the prediction performance is estimated by utilizing historical demand data of electronic equipment spare parts. The prediction result proves the superiority of the combined prediction method; the method can effectively provide a new way for predicting the needs of the medium-and-long-term spare parts, so that the management of the internal inventory of the enterprise is optimized.
2. In the embodiment of the invention, the electronic equipment spare part demand is predicted based on the CNN-GA-BP combined model, and compared with a single BP model or CNN-BP model, the method has the advantages of higher precision, faster convergence speed, more accurate prediction result and superiority.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. The method for predicting the demand quantity of the spare parts of the combined model based on the CNN-GA-BP is characterized by being used for predicting the demand quantity of the spare parts of the medium-and-long-term electronic equipment, and in the training process of the combined model based on the CNN-GA-BP:
s1, acquiring and vectorizing historical demand data of electronic equipment spare parts;
s2, acquiring a direct prediction result by adopting a bottom layer prediction model according to the historical demand data vector;
s3, splicing the historical demand data vector and the direct prediction result, and acquiring a convolution result by adopting a middle layer prediction model CNN network;
s4, splicing the historical demand data vector, the direct prediction result and the convolution result, and dividing the spliced result into a training set and a testing set; the training set is used as input of a top layer prediction model GA-BP network, and input data are trained;
s5, taking the test set as input of the trained GA-BP network, and obtaining a spare part demand prediction result.
2. The CNN-GA-BP based combined model spare part demand prediction method of claim 1, wherein the S2 mid-floor prediction model comprises an exponential smoothing method and a differential autoregressive moving average model.
3. The CNN-GA-BP based combined model spare part demand prediction method of claim 2, wherein the exponential smoothing method and differential autoregressive moving average model comprise:
(1) Exponential smoothing method
Figure FDA0004031154400000011
wherein ,yt For the actual value of the t-th period, the time series consists of historical demand data of spare parts, 1<t<T, T is the total period number; alpha is a smoothing coefficient, 0<α<1；
Figure FDA0004031154400000021
The predicted value of the t-th phase i method is shown as a predicted value, wherein i=1 and 2 respectively represent a primary exponential smoothing method and a secondary exponential smoothing method;
is provided with
Figure FDA0004031154400000022
For the actual value of the i method at phase t+1, then the two exponential smoothing modes become:
Figure FDA0004031154400000023
Figure FDA0004031154400000024
wherein ,
Figure FDA0004031154400000025
the actual values of the T+1st stage of the primary and secondary exponential smoothing methods are respectively;
(2) Assume { X t And a group of weak stationary time sequences, which are subjected to d-level difference:
Figure FDA0004031154400000026
the differential autoregressive moving average model is expressed as:
W t ＝α 1 W t-1 +α 2 W t-2 …+α p W t-p -θ 1 β t-1 -θ 2 β t-2 …-θ p β t-p
wherein ,Wt The sequence is subjected to d-order differential processing; b is a backward shift operator and satisfies the condition B k X t ＝X t-k ；
Figure FDA0004031154400000027
Representing differential calculation, and having ∈>
Figure FDA0004031154400000028
X t Is the time sequence value at the t time; beta t Mean 0 and variance sigma a White noise of (a); alpha t 、θ t Is a model parameter.
4. The method for predicting the spare part demand of a combined model based on CNN-GA-BP as recited in claim 1, wherein said step S3 is performed with zero padding after splicing said historical demand data vector and direct prediction result, and then a CNN network is used for convolution.
5. The CNN-GA-BP based combined model spare part demand prediction method of claim 1, wherein the training process of the GA-BP network in S4 comprises:
s10, initializing a network structure of a BP neural network and learning parameters; the network structure comprises an input layer, an implicit layer and an output layer, wherein the GA algorithm is adopted to execute learning parameter optimization, and the optimized weight and threshold value are obtained;
s20, inputting samples in a training set, and calculating the output of an hidden layer;
Figure FDA0004031154400000031
wherein ,Hj For the j-th node of the hidden layer, f is the excitation function, l is the node number of the input layer, omega ij For the weight value between the i node of the input layer and the j node of the hidden layer, x i An input value of an i-th node of the input layer, a j A threshold value for the j-th node of the hidden layer;
s30, forward propagation, calculating output O of an output layer k ；
Figure FDA0004031154400000032
wherein ,Ok For the output of the kth node of the output layer, q is the number of nodes of the hidden layer, omega jk B is the weight between the jth node of the hidden layer and the kth node of the output layer k A threshold value for the kth node of the output layer;
s40, calculating the actual output of the current network through the training set, comparing the actual output with the expected output of the network,
e k ＝Y k -O k
wherein ,ek For the output error of the kth node of the output layer, Y k A desired output for a kth node of the output layer;
s50, back propagation, correcting the weight of the neural network layer by layer according to the error,
Figure FDA0004031154400000033
wherein eta is the learning rate and m is the node number of the output layer;
s60, updating the threshold value, and correcting the threshold value layer by layer according to the error;
Figure FDA0004031154400000041
s70, judging whether iteration of the BP neural network algorithm is finished, and reaching the designated iteration times or two adjacent iterationsThe difference between the errors is smaller than the appointed value, and the iteration ends to output the predicted value
Figure FDA0004031154400000042
Otherwise, the process returns to S20.
6. The method for predicting the spare parts of a combined model based on CNN-GA-BP according to claim 5, wherein the step S100 of performing learning parameter optimization by using a GA algorithm comprises:
s100, initializing a population;
s200, taking the training error of the BP neural network as an fitness function;
s300, repeatedly executing the operations of selection, crossing and mutation;
s400, stopping iteration until the individual fitness meets the condition or the iteration times reach the designated times, and outputting the optimal weight and the threshold value.
7. The method for predicting the demand for a CNN-GA-BP based composite model according to any one of claims 1 to 6, wherein the method for verifying the accuracy of the prediction of the CNN-GA-BP based composite model using four evaluation indexes comprises:
(1) Mean absolute error MAE:
Figure FDA0004031154400000043
(2) Mean square error MSE:
Figure FDA0004031154400000044
(3) Root mean square error RMSE:
Figure FDA0004031154400000051
(4) R2 determines the coefficient:
Figure FDA0004031154400000052
wherein n is the number of samples;
Figure FDA0004031154400000053
is a predicted value; y is t Is an actual value; />
Figure FDA0004031154400000054
Is the average of the actual values.
8. The system for predicting the demand quantity of the spare parts of the combined model based on the CNN-GA-BP is characterized by being used for predicting the demand quantity of the spare parts of the middle-long-term equipment, and in the training process of the combined model based on the CNN-GA-BP:
the vectorization module is used for acquiring and vectorizing historical demand data of the electronic equipment spare parts;
the direct prediction module is used for acquiring a direct prediction result by adopting a bottom layer prediction model according to the historical demand data vector;
the convolution module is used for splicing the historical demand data vector and the direct prediction result, and a middle layer prediction model CNN network is adopted to obtain a convolution result;
the training module is used for splicing the historical demand data vector, the direct prediction result and the convolution result, and dividing the spliced result into a training set and a testing set; the training set is used as input of a top layer prediction model GA-BP network, and input data are trained;
and the test module is used for taking the test set as the input of the trained GA-BP network and obtaining the spare part demand prediction result.
9. A storage medium, characterized in that it stores a computer program for CNN-GA-BP-based combined model spare part demand prediction, wherein the computer program causes a computer to execute the CNN-GA-BP-based combined model spare part demand prediction method according to any one of claims 1 to 7.
10. An electronic device, comprising:
one or more processors;
a memory; and
one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs comprising instructions for performing the CNN-GA-BP based combined model spare part demand prediction method of any one of claims 1-7.
CN202211729757.5A 2022-12-28 2022-12-30 CNN-GA-BP-based combined model spare part demand prediction method and system Active CN116050600B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2023/094438 WO2024139014A1 (en) 2022-12-28 2023-05-16 Cnn-ga-bp combined model-based spare part demand prediction method and system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211696332 2022-12-28
CN2022116963329 2022-12-28

Publications (2)

Publication Number Publication Date
CN116050600A true CN116050600A (en) 2023-05-02
CN116050600B CN116050600B (en) 2024-08-23

Family

ID=86123159

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211729757.5A Active CN116050600B (en) 2022-12-28 2022-12-30 CNN-GA-BP-based combined model spare part demand prediction method and system

Country Status (2)

Country Link
CN (1) CN116050600B (en)
WO (1) WO2024139014A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024139014A1 (en) * 2022-12-28 2024-07-04 中国电子科技集团公司第三十八研究所 Cnn-ga-bp combined model-based spare part demand prediction method and system

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118468733A (en) * 2024-07-12 2024-08-09 国网山东省电力公司邹城市供电公司 Cable joint temperature prediction method and system

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108053054A (en) * 2017-11-22 2018-05-18 中国农业大学 A kind of water quality prediction of river method
CN111324989A (en) * 2020-03-19 2020-06-23 重庆大学 GA-BP neural network-based gear contact fatigue life prediction method
CN111539515A (en) * 2020-04-21 2020-08-14 中国电子科技集团公司第三十八研究所 Complex equipment maintenance decision method based on fault prediction
CN111967679A (en) * 2020-08-20 2020-11-20 江苏师范大学 Ionized layer total electron content forecasting method based on TCN model
CN112100940A (en) * 2020-09-17 2020-12-18 浙江大学 Method and device for predicting primary stretching technological parameters of optical fiber preform
WO2021082809A1 (en) * 2019-10-29 2021-05-06 山东科技大学 Training optimization method for foreign exchange time series prediction
CN113239569A (en) * 2021-05-31 2021-08-10 北京科技大学 Data-driven plate shape prediction method and device
CN114595890A (en) * 2022-03-09 2022-06-07 合肥工业大学 Ship spare part demand prediction method and system based on BP-SVR combined model
CN114707712A (en) * 2022-03-23 2022-07-05 辽宁红沿河核电有限公司 Method for predicting requirement of generator set spare parts
CN115237896A (en) * 2022-07-12 2022-10-25 四川大学 Data preprocessing method and system for forecasting air quality based on deep learning

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107466816B (en) * 2017-07-24 2020-04-07 东华大学 Irrigation method based on dynamic multilayer extreme learning machine
KR102271379B1 (en) * 2020-12-16 2021-06-30 주식회사 두잉랩 Method and Server for Recognizing and Determining the Object in the Image
CN112906957A (en) * 2021-02-05 2021-06-04 东北大学秦皇岛分校 BP neural network coal conveying fault prediction method based on genetic algorithm optimization
CN116050600B (en) * 2022-12-28 2024-08-23 中国电子科技集团公司第三十八研究所 CNN-GA-BP-based combined model spare part demand prediction method and system

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108053054A (en) * 2017-11-22 2018-05-18 中国农业大学 A kind of water quality prediction of river method
WO2021082809A1 (en) * 2019-10-29 2021-05-06 山东科技大学 Training optimization method for foreign exchange time series prediction
CN111324989A (en) * 2020-03-19 2020-06-23 重庆大学 GA-BP neural network-based gear contact fatigue life prediction method
CN111539515A (en) * 2020-04-21 2020-08-14 中国电子科技集团公司第三十八研究所 Complex equipment maintenance decision method based on fault prediction
CN111967679A (en) * 2020-08-20 2020-11-20 江苏师范大学 Ionized layer total electron content forecasting method based on TCN model
CN112100940A (en) * 2020-09-17 2020-12-18 浙江大学 Method and device for predicting primary stretching technological parameters of optical fiber preform
CN113239569A (en) * 2021-05-31 2021-08-10 北京科技大学 Data-driven plate shape prediction method and device
CN114595890A (en) * 2022-03-09 2022-06-07 合肥工业大学 Ship spare part demand prediction method and system based on BP-SVR combined model
CN114707712A (en) * 2022-03-23 2022-07-05 辽宁红沿河核电有限公司 Method for predicting requirement of generator set spare parts
CN115237896A (en) * 2022-07-12 2022-10-25 四川大学 Data preprocessing method and system for forecasting air quality based on deep learning

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
崔吉峰;乞建勋;杨尚东;: "基于粒子群改进BP神经网络的组合预测模型及其应用", 中南大学学报(自然科学版), no. 01, pages 190 - 194 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024139014A1 (en) * 2022-12-28 2024-07-04 中国电子科技集团公司第三十八研究所 Cnn-ga-bp combined model-based spare part demand prediction method and system

Also Published As

Publication number Publication date
CN116050600B (en) 2024-08-23
WO2024139014A1 (en) 2024-07-04

Similar Documents

Publication Publication Date Title
CN116050600B (en) CNN-GA-BP-based combined model spare part demand prediction method and system
CN111563706A (en) Multivariable logistics freight volume prediction method based on LSTM network
US11650968B2 (en) Systems and methods for predictive early stopping in neural network training
CN110245801A (en) A kind of Methods of electric load forecasting and system based on combination mining model
CN106815782A (en) A kind of real estate estimation method and system based on neutral net statistical models
CN107463993B (en) Medium-and-long-term runoff forecasting method based on mutual information-kernel principal component analysis-Elman network
CN105956690A (en) Water supply prediction method and water supply prediction system
CN113822499B (en) Train spare part loss prediction method based on model fusion
CN116432543B (en) Method for predicting remaining life of power semiconductor module, terminal device and storage medium
CN116303786B (en) Block chain financial big data management system based on multidimensional data fusion algorithm
CN112766548A (en) Order completion time prediction method based on GASA-BP neural network
CN116151581A (en) Flexible workshop scheduling method and system and electronic equipment
CN116108742A (en) Low-voltage transformer area ultra-short-term load prediction method and system based on improved GRU-NP model
CN115640888A (en) Yield prediction method of decreasing function embedded threshold sequence network
CN109540089B (en) Bridge deck elevation fitting method based on Bayes-Kriging model
CN114792126A (en) Convolutional neural network design method based on genetic algorithm
CN111695260B (en) Material performance prediction method and system
CN116703607A (en) Financial time sequence prediction method and system based on diffusion model
CN117076887A (en) Pump station unit running state prediction and health assessment method and system
CN114595890A (en) Ship spare part demand prediction method and system based on BP-SVR combined model
CN112651168B (en) Construction land area prediction method based on improved neural network algorithm
CN111414990B (en) Convolutional neural network processing method and device, electronic equipment and storage medium
CN111160419A (en) Electronic transformer data classification prediction method and device based on deep learning
CN111080160A (en) Interactive multi-criterion decision method and device containing preference assistant decision factor preference and computer equipment
CN113379455B (en) Order quantity prediction method and equipment

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
GR01 Patent grant
GR01 Patent grant