CN109034388A - A kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network - Google Patents
A kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network Download PDFInfo
- Publication number
- CN109034388A CN109034388A CN201810839016.XA CN201810839016A CN109034388A CN 109034388 A CN109034388 A CN 109034388A CN 201810839016 A CN201810839016 A CN 201810839016A CN 109034388 A CN109034388 A CN 109034388A
- Authority
- CN
- China
- Prior art keywords
- neural network
- model
- genetic algorithm
- training
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 63
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 40
- 230000002068 genetic effect Effects 0.000 title claims abstract description 35
- 235000019504 cigarettes Nutrition 0.000 title claims abstract description 24
- 239000000463 material Substances 0.000 title claims abstract description 22
- 239000000779 smoke Substances 0.000 title claims abstract description 17
- 239000000470 constituent Substances 0.000 title claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 48
- 238000003062 neural network model Methods 0.000 claims abstract description 24
- 238000005457 optimization Methods 0.000 claims abstract description 24
- 230000000694 effects Effects 0.000 claims abstract description 6
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000004519 manufacturing process Methods 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 14
- 238000012360 testing method Methods 0.000 claims description 14
- 210000002569 neuron Anatomy 0.000 claims description 8
- 230000035772 mutation Effects 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 5
- 230000014509 gene expression Effects 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 238000000034 method Methods 0.000 description 19
- 230000008569 process Effects 0.000 description 8
- 241000208125 Nicotiana Species 0.000 description 6
- 235000002637 Nicotiana tabacum Nutrition 0.000 description 6
- 238000013507 mapping Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 210000004218 nerve net Anatomy 0.000 description 3
- SNICXCGAKADSCV-JTQLQIEISA-N (-)-Nicotine Chemical compound CN1CCC[C@H]1C1=CC=CN=C1 SNICXCGAKADSCV-JTQLQIEISA-N 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000010219 correlation analysis Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 239000004615 ingredient Substances 0.000 description 2
- 229960002715 nicotine Drugs 0.000 description 2
- SNICXCGAKADSCV-UHFFFAOYSA-N nicotine Natural products CN1CCCC1C1=CC=CN=C1 SNICXCGAKADSCV-UHFFFAOYSA-N 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000000505 pernicious effect Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
Classifications
-
- 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
- G06N3/12—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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Genetics & Genomics (AREA)
- Physiology (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network, the following steps are included: sample preprocessing；The optimal weight and threshold parameter of neural network is obtained based on genetic algorithm；Based on best weight value and threshold value that genetic algorithm obtains, neural network, and training neural network model are constructed；The neural network model obtained to training is verified, and assessment models are applied to actual effect.Compared with the neural network for not using optimization algorithm, neural network model based on genetic algorithm optimization first uses genetic algorithm to select that model error is made to reach the smallest weight and threshold value as the initial parameter of training neural network, locally optimal solution can be fallen into avoid model, and is unable to get globally optimal solution.
Description
Technical field
The present invention relates to a kind of prediction models, belong to industrial design and production field, utilize genetic algorithm optimization nerve net
Network establishes the mapping relations between cigarette material and mainstream smoke constituents, reasonably adjusts cigarette material formula for Cigarette Manufacturing Industry,
It produces cigarette up to specification and provides model, specifically a kind of cigarette material neural network based and mainstream smoke constituents
Prediction model.
Background technique
In tobacco manufacturing industry, the pernicious gas ingredient and content of tobacco release are always tobacco manufacturing company and consumer
Concern.For optimizing product design, accelerate production efficiency, the product of production qualification and less harmful, cigarette making industry
It needs using the relevant knowledge of statistics, canonical correlation analysis, principal component analysis and partial Correlation Analysis is carried out to tobacco raw material
Deng, obtain with the highest cigarette material of main flume degree of correlation, then the obtained principal component of analysis and main flume are established
Model appropriate, to assist the materials and production in tobacco manufacturing process.With the development of information technology, the calculating of computer and
Processing capacity is continuously improved, and carries out statistical analysis to data using computer, handles and be modeled as trend.Utilize computer
It can be rapidly performed by analysis and modeling, but traditional statistical analysis method is that the priori based on low volume data and people is known
Know and carry out data analysis and modeling, this modeling method has very strong specific aim, and the applicability of model is very poor.
In recent years, deep learning is risen again, and neural network can be compared with by complicated network structure and self-learning capability
Good fitting is various to seem irregular data, and the only unconspicuous data modeling of rule provides new modeling method, and
And it is more preferable compared to the fitting effect of traditional statistics modeling method.BP neural network be it is a kind of by learning rules generalization, it is right
The adjustment that non-linear differentiable function carries out the multitiered network weight of Weight Training uses the learning algorithm of backpropagation.Study
Process is divided into two stages: (1) information forward-propagating；(2) error back propagation.In the training process of BP neural network, such as
Difference between the primary output result of fruit and desired output result is more than some standard, and the number of training does not reach
The maximum frequency of training of setting approaches output valve step by step then error returns with some form, and corrects the weight of each layer
Desired output valve, until error is reduced to acceptable degree or frequency of training reaches preset number.Currently, refreshing
It has begun through the relevant technology of network and is applied in cigarette sensory evaluation and the model of mainstream smoke constituents relationship, but do not have also
There is the precedent that cigarette material Yu main flume model are established using genetic algorithm optimization BP neural network.
Summary of the invention
Cigarette is established based on genetic algorithm optimization BP neural network in view of the deficiencies of the prior art, the present invention proposes a kind of
The method of material and mainstream smoke constituents prediction model.
The invention discloses a kind of prediction model for establishing cigarette material and mainstream smoke constituents and using model method,
First with the optimal initial weight of genetic algorithm structure choice neural network based and threshold value；Genetic algorithm is then based on to obtain
The optimal initial weight and Threshold-training neural network arrived falls into locally optimal solution to avoid model, rather than globally optimal solution；
Finally, being input with truthful data, it is practical whether assessment models can be applied to production based on trained model.
The technical solution adopted by the present invention is that: a kind of cigarette material and mainstream smoke based on Genetic Algorithm Optimized Neural Network
The prediction model of gas ingredient, which comprises the following steps:
Step 1: sample preprocessing；
Step 2: the optimal weight and threshold parameter of neural network is obtained based on genetic algorithm；
Step 3: the best weight value and threshold value obtained based on genetic algorithm constructs neural network, and training neural network mould
Type；
Step 4: the neural network model obtained to training is verified, and assessment models are applied to actual effect.
Further, the specific implementation of step 1 includes following sub-step:
Step 1.1: there are missing values and data obviously not to meet actual data record in removal sample；
Step 1.2: the data in initial data are normalized according to following expressions, and x represents initial data, and y is represented
Data after normalization, Minvalue and MaxValue respectively represent the minimum value maximum value of initial data, y=(x-
MinValue)/(MaxValue-Minvalue)；
Step 1.3: the data after normalization are divided into training set and test set.
Further, the specific implementation of step 2 includes following sub-step:
Step 2.1: defining the constant parameter of neural network, input layer number INPUT_NODE, hidden layer number of nodes
LAYER1_NODE, output layer number of nodes OUTPUT_NODE, the learning rate LEARNING_RATE_BASE on basis, learning rate decline
Coefficient of the regularization term in loss function of lapse rate LEARNING_RATE_DECAY, descriptive model complexity
REAULARIZATION_RATE, exercise wheel number TRAINING_SETPS, sliding average attenuation rate MOVING_AVERAGE_
DECAY；
Step 2.2: defining activation primitive tf.nn.relu (), loss function loss (), the optimization calculation for optimizing loss function
Method tf.train.GradientDescentOptimizer ()；
Step 2.3: the evolutionary generation of population, population scale, intersection and mutation probability are set, and to the initial power of neuron
Weight and threshold value carry out real coding；
Step 2.4: calculating the fitness of population, optimal some individuals are selected from current population；
Step 2.5: population selected, intersected, mutation operation；
Step 2.6: judging whether to have reached optimization aim, if having reached optimization aim, select optimal weight and threshold value
It performs the next step, it is no to then follow the steps 2.4；
Further, the specific implementation of step 3 includes following sub-step:
Step 3.1: BP nerve net is constructed based on best weight value obtained in the previous step and threshold value and other neural network parameters
Network；
Step 3.2: training neural network model；
Step 3.3: calculating the error between the true value in the value and data set of trained neural network model output, root
Successively update the weight and threshold value of each layer of neural network forward according to error；
Step 3.4: whether the model judged has reached expected precision or maximum number of iterations, if so, holding
It goes in next step, otherwise return step 3.2；
Further, the specific implementation of step 4 includes following sub-step:
Step 4.1: neural network model obtained in the data input step 3 in test set, obtaining corresponding output
Value；
Step 4.2: the error between computation model output valve and true value, if error has reached expected requirement, mould
Type can be applied in actual production, and otherwise, return step 2 adjusts parameters, re -training model.
Beneficial effects of the present invention and feature are: the present invention is first volume based on genetic algorithm optimization BP neural network
Prediction model of the cigarette material to mainstream smoke constituents.Traditional linear regression fit curve is non-linear in processing multiple-input and multiple-output
When mapping relations, usually poor fitting.BP network can learn and store a large amount of input-output mode map relationship, without thing
It is preceding to disclose the math equation for describing this mapping relations.Weight is corrected repeatedly during using BP neural network training pattern,
Model is continued to optimize, model can be made to be fitted training data with arbitrary accuracy.Exist between cigarette material and mainstream smoke constituents
Certain complicated Nonlinear Mapping relationship, therefore cigarette material and main flume are established using genetic algorithm optimization BP neural network
Between prediction model it is not only more suitable than traditional homing method, but also the neural network model due to being not optimised.
Detailed description of the invention
Fig. 1 is the principle framework structure of the embodiment of the present invention；
Fig. 2 is fitness change curve during selecting optimal neutral net initiation parameter based on genetic algorithm；
Fig. 3, Fig. 4 are prediction error comparison of the neural network model to tar and CO that genetic algorithm optimization is not used respectively
Figure；
Fig. 5, Fig. 6 are to be compared using the neural network model of genetic algorithm optimization to the prediction accuracy of tar and CO respectively
Figure.
Specific embodiment
Invention is further explained with reference to the accompanying drawing:
In this example, whole experimental data sets is divided into two parts, training set of a part as BP neural network,
Test set of the another part as BP neural network, the data in training set take 2/3-the 4/5 of total amount of data.Defining D is n × m
Data set, every a line represents data record, each column one attribute of expression.Set A and B is enabled to respectively indicate division data
Collect obtained training set and test set.Genetic algorithm is first based on based on training set A and obtains optimal neural network weight and threshold value,
Again using optimal neural network weight and threshold value as the initial parameter training neural network of training BP neural network model, work as net
Difference between the output result of network and desired output result reaches some standard or has reached maximum the number of iterations, just
Think that BP neural network is trained to complete, then examines prediction of the model of training in new data set using test set B
Precision.
The invention proposes one kind to establish cigarette material and mainstream smoke constituents based on genetic algorithm optimization BP neural network
Prediction model and using model method.Include the following steps and (can refer to Fig. 1):
Step 1: sample preprocessing；
Step 1.1: there are missing values and data obviously not to meet actual data record in removal sample；
Step 1.2: the data in initial data are normalized according to following expressions, and x represents initial data, and y is represented
Data after normalization, Minvalue and MaxValue respectively represent the minimum value maximum value of initial data, y=(x-
MinValue)/(MaxValue-Minvalue)；
Step 1.3: the data after normalization are divided into training set and test set.
The specific implementation process of embodiment is described as follows:
Concentrate the value x of each attribute in each data record according to y=(x-MinValue)/(MaxValue- data
Minvalue mode) normalized after value y, using each the y value being calculated as the value of corresponding attribute, thus
Form the data set of new n × m.
In data set after normalization 2/3-4/5 data record is divided into training set, remaining data are divided into
Test set will be by the way of random division, to avoid in data set partition process during dividing training set and test
Think influence of the factor to experimental result.
Step 2: the optimal weight and threshold parameter of neural network is obtained based on genetic algorithm；
Step 2.1: defining the constant parameter of neural network, input layer number INPUT_NODE, hidden layer number of nodes
LAYER1_NODE, output layer number of nodes OUTPUT_NODE, the learning rate LEARNING_RATE_BASE on basis, learning rate decline
Coefficient of the regularization term in loss function of lapse rate LEARNING_RATE_DECAY, descriptive model complexity
REAULARIZATION_RATE, exercise wheel number TRAINING_SETPS, sliding average attenuation rate MOVING_AVERAGE_
DECAY；
Step 2.2: defining activation primitive tf.nn.relu (), loss function loss (), the optimization calculation for optimizing loss function
Method tf.train.GradientDescentOptimizer ()；
Step 2.3: the evolutionary generation of population, population scale, intersection and mutation probability are set, and to the initial power of neuron
Weight and threshold value carry out real coding；
Step 2.4: calculating the fitness of population, optimal some individuals are selected from current population；
Step 2.5: population selected, intersected, mutation operation；
Step 2.6: judging whether to have reached optimization aim, if having reached optimization aim, select optimal weight and threshold value
It performs the next step, it is no to then follow the steps 2.4；
The specific implementation process of embodiment is described as follows:
Before obtaining optimal weight and threshold value using genetic algorithm, the input layer number of neural network is first determined
INPUT_NODE, the quantity of hidden layer and hidden layer number of nodes LAYER1_NODE, output layer number of nodes OUTPUT_NODE etc.
A series of parameter needed in the parameter and genetic algorithm that neural networks need, such as: evolutionary generation, population scale.Then just
The initial weight and threshold value of beginningization neuron carry out real coding, calculate the fitness of population one by one, and successively carry out neuron
Selection, intersection and mutation operation.As soon as every execution wheel selection intersects and makes a variation, population update is primary, needs to recalculate kind
The fitness of group.Population's fitness is an index for portraying population to Optimum Solution approximation ratio, here using calculating instruction
The mode for practicing the sum of predicted value error for concentrating total data and model solves the fitness of each individual.According to the adaptation of individual
Degree selects the lesser some individuals of fitness numerical value to form new population, in new population, continues selection, intersects and become
It is different, make the continuous iteration of population, it is improper to approach optimal solution.After calculating fitness every time, obtained most in all iteration before preservation
The neuron of good fitness, finally obtained best neuron will train the initial power of final neural network model as next step
Value and threshold value.
Step 3: the best weight value and threshold value obtained based on genetic algorithm constructs neural network, and training neural network mould
Type；
Step 3.1: BP nerve net is constructed based on best weight value obtained in the previous step and threshold value and other neural network parameters
Network；
Step 3.2: training neural network model；
Step 3.3: calculating the error between the true value in the value and data set of trained neural network model output, root
Successively update the weight and threshold value of each layer of neural network forward according to error；
Step 3.4: whether the model judged has reached expected precision or maximum number of iterations, if so, holding
It goes in next step, otherwise return step 3.2；
The specific implementation process of embodiment is described as follows:
The initial weight and threshold value of best neuron initialization neural network according to obtained in step 2.In training process
In, the weight of each layer will gradually be adjusted according to the difference between output valve and desired output valve, until obtaining according to current net
Optimum state under network structure.The training process of BP neural network not only needs take turns to train more, it is also possible to by repeatedly adjustment
The hiding layer number of network and the number of nodes of each layer continue to optimize output result.
In neural network hidden layer and output layer, needs to use nonlinear function as activation primitive, keep model foundation non-
Linear Mapping.It defines a loss function for the quality of assessment models training in output layer and model effectiveness is assessed.
Weight is adjusted using index decreased strategy for training for promotion speed in the incipient stage of network training just, reaches quick
The regularization term of descriptive model complexity is added in order to avoid the over-fitting of model in convergent effectiveness in loss function, passes through
Iteration several times uses the weight of each layer in optimization algorithm optimization network.
After network structure, activation primitive, loss function and optimization algorithm has been determined, so that it may start based on trained number
Neural network is trained according to collecting.During training BP neural network, the power of each node layer can be constantly adjusted inside neural network
Value, until reaching optimal training result under current network configuration.At this point, the training of BP neural network does not terminate, one
It completes after training, should also be determined whether according to experimental result to the structure of neural network and certain under the configuration of secondary neural network
A little parameters are adjusted rear re -training, to obtain optimal network model.
Step 4: the neural network model obtained to training is verified, and assessment models are applied to actual effect.
Step 4.1: neural network model obtained in the data input step 3 in test set, obtaining corresponding output
Value；
Step 4.2: the error between computation model output valve and true value, if error has reached expected requirement, mould
Type can be applied in actual production, and otherwise, return step 2 adjusts parameters, re -training model.
It is real training neural network and then whether can be applied to production with the model that test set detection training obtains
In border, if difference model as defined in some of the output valve in predicted value and test set in test set based on BP neural network
In enclosing, be considered as the BP neural network model that training obtains be can be applied to produce it is actual；Otherwise, BP neural network model
It needs to rebuild and train.
By the comparative illustration of Fig. 3, Fig. 4, Fig. 5, Fig. 6, this patent by using genetic algorithm optimization neural network mould
What the predicted value precision of tar, nicotine and CO that type obtains were obtained relative to the neural network model of unused genetic algorithm optimization
The predicted value precision of tar, nicotine and CO is enhanced, illustrate this patent use technical solution can obtain really compared with
Good actual effect can generate positive effect to the materials and production assisted in tobacco manufacturing process.
The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the above embodiments and description only describe this
The structural relation and principle of invention, without departing from the spirit and scope of the present invention, the present invention also have various change and
It improves, these changes and improvements all fall within the protetion scope of the claimed invention.The claimed scope of the invention is by appended power
Sharp claim and its equivalent thereof.
Claims (5)
1. a kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network, feature exist
In, comprising the following steps:
Step 1: sample preprocessing；
Step 2: the optimal weight and threshold parameter of neural network is obtained based on genetic algorithm；
Step 3: the best weight value and threshold value obtained based on genetic algorithm constructs neural network, and training neural network model；
Step 4: the neural network model obtained to training is verified, and assessment models are applied to actual effect.
2. the cigarette material and mainstream smoke constituents according to claim 1 based on Genetic Algorithm Optimized Neural Network is pre-
Survey model, which is characterized in that the specific implementation of step 1 includes following sub-step:
Step 1.1: there are missing values and data obviously not to meet actual data record in removal sample；
Step 1.2: the data in initial data are normalized according to following expressions, and x represents initial data, and y represents normalizing
Data after change, Minvalue and MaxValue respectively represent the minimum value maximum value of initial data, y=(x-
MinValue)/(MaxValue-Minvalue)；
Step 1.3: the data after normalization are divided into training set and test set.
3. the cigarette material and mainstream smoke constituents according to claim 1 based on Genetic Algorithm Optimized Neural Network is pre-
Survey model, which is characterized in that the specific implementation of step 2 includes following sub-step:
Step 2.1: defining the constant parameter of neural network, input layer number INPUT_NODE, hidden layer number of nodes LAYER1_
The attenuation rate of NODE, output layer number of nodes OUTPUT_NODE, the learning rate LEARNING_RATE_BASE on basis, learning rate
Coefficients R EAULARIZATION_ of the regularization term in loss function of LEARNING_RATE_DECAY, descriptive model complexity
RATE, exercise wheel number TRAINING_SETPS, sliding average attenuation rate MOVING_AVERAGE_DECAY；
Step 2.2: defining activation primitive tf.nn.relu (), loss function loss (), the optimization algorithm for optimizing loss function
tf.train.GradientDescentOptimizer()；
Step 2.3: the evolutionary generation of population is set, population scale intersects and mutation probability, and to the initial weight of neuron and
Threshold value carries out real coding；
Step 2.4: calculating the fitness of population, optimal some individuals are selected from current population；
Step 2.5: population selected, intersected, mutation operation；
Step 2.6: judging whether to have reached optimization aim, if having reached optimization aim, optimal weight and threshold value is selected to execute
In next step, no to then follow the steps 2.4.
4. the cigarette material and mainstream smoke constituents according to claim 1 based on Genetic Algorithm Optimized Neural Network is pre-
Survey model, which is characterized in that the specific implementation of step 3 includes following sub-step:
Step 3.1: BP neural network is constructed based on best weight value obtained in the previous step and threshold value and other neural network parameters；
Step 3.2: training neural network model；
Step 3.3: the error between the true value in the value and data set of trained neural network model output is calculated, according to accidentally
The poor weight and threshold value for successively updating each layer of neural network forward；
Step 3.4: whether the model judged has reached expected precision or maximum number of iterations, if so, under executing
One step, otherwise return step 3.2.
5. the cigarette material and mainstream smoke constituents according to claim 1 based on Genetic Algorithm Optimized Neural Network is pre-
Survey model, which is characterized in that the specific implementation of step 4 includes following sub-step:
Step 4.1: neural network model obtained in the data input step 3 in test set, obtaining corresponding output valve；
Step 4.2: the error between computation model output valve and true value, if error has reached expected requirement, model can
To be applied in actual production, otherwise, return step 2 adjusts parameters, re -training model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810839016.XA CN109034388A (en) | 2018-07-27 | 2018-07-27 | A kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810839016.XA CN109034388A (en) | 2018-07-27 | 2018-07-27 | A kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109034388A true CN109034388A (en) | 2018-12-18 |
Family
ID=64646907
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810839016.XA Pending CN109034388A (en) | 2018-07-27 | 2018-07-27 | A kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109034388A (en) |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110059821A (en) * | 2019-03-19 | 2019-07-26 | 埃睿迪信息技术（北京）有限公司 | Neural network optimization, device, computer equipment and storage medium |
| CN110457758A (en) * | 2019-07-16 | 2019-11-15 | 江西理工大学 | Prediction technique, device, system and the storage medium in Instability of Rock Body stage |
| CN110490244A (en) * | 2019-08-14 | 2019-11-22 | 吉林大学 | A kind of data processing method and device |
| CN110516954A (en) * | 2019-08-23 | 2019-11-29 | 昆明理工大学 | One kind referring to calibration method based on GA-BP neural network algorithm optimization mineral processing production |
| CN110656385A (en) * | 2019-10-28 | 2020-01-07 | 季华实验室 | Neural network-based electrostatic spinning fiber average diameter prediction method |
| CN110826229A (en) * | 2019-11-07 | 2020-02-21 | 长沙学院 | Cut tobacco drying head process modeling and optimal setting control method based on long-term prediction |
| CN111126560A (en) * | 2019-11-07 | 2020-05-08 | 云南民族大学 | Method for optimizing BP neural network based on cloud genetic algorithm |
| CN111210877A (en) * | 2020-01-03 | 2020-05-29 | 山东天岳先进材料科技有限公司 | Method and device for deducing physical property parameters |
| CN111242278A (en) * | 2019-12-29 | 2020-06-05 | 杭州拓深科技有限公司 | Composite smoke sense low-false-alarm method based on intelligent algorithm |
| CN111950099A (en) * | 2020-08-03 | 2020-11-17 | 中国石油大学(华东) | Method, system, medium and computer equipment for testing mechanical property of equipment material |
| CN112016249A (en) * | 2020-08-31 | 2020-12-01 | 华北电力大学 | SCR denitration system bad data identification method based on optimized BP neural network |
| CN112099355A (en) * | 2020-09-18 | 2020-12-18 | 湖北中烟工业有限责任公司 | BP-PID-based tobacco sheet coating rate control method and device |
| CN112231978A (en) * | 2020-10-15 | 2021-01-15 | 西安特种设备检验检测院 | Boiler flue gas acid dew point testing method based on artificial neural network |
| CN112308278A (en) * | 2019-08-02 | 2021-02-02 | 中移信息技术有限公司 | Method, device, equipment and medium for optimizing prediction model |
| CN112464566A (en) * | 2020-12-07 | 2021-03-09 | 广东电网有限责任公司佛山供电局 | Transformer oil temperature prediction method based on genetic algorithm and BP neural network |
| CN112464554A (en) * | 2020-11-03 | 2021-03-09 | 桂林理工大学 | Operating parameter optimization method of gasoline refining equipment |
| CN112456639A (en) * | 2020-11-10 | 2021-03-09 | 盐城工学院 | Intelligent prediction method for hydrogen sulfide production in anaerobic digestion |
| CN112485622A (en) * | 2020-12-09 | 2021-03-12 | 西安华谱电力设备制造有限公司 | Partial discharge pattern recognition method based on GA-BPNN |
| CN112765883A (en) * | 2021-01-18 | 2021-05-07 | 电子科技大学 | Method for determining valve closing process based on genetic algorithm and neural network |
| CN113077041A (en) * | 2020-01-06 | 2021-07-06 | 大唐环境产业集团股份有限公司 | Deep neural network modeling independent variable dimension reduction method based on genetic algorithm |
| CN113268927A (en) * | 2021-05-21 | 2021-08-17 | 哈尔滨工业大学 | High-power laser device output energy prediction method based on full-connection neural network |
| CN113449930A (en) * | 2021-07-27 | 2021-09-28 | 威海长和光导科技有限公司 | Optical fiber preform preparation quality prediction method based on BP neural network |
| CN113779858A (en) * | 2020-06-10 | 2021-12-10 | 国家能源投资集团有限责任公司 | Combustion optimization method and system, storage medium and electronic equipment |
| CN113887636A (en) * | 2021-10-09 | 2022-01-04 | 四川大学 | Selectable data enhancement method and system based on genetic algorithm |
| CN114354455A (en) * | 2022-01-17 | 2022-04-15 | 北京石油化工学院 | Method for online measurement of particle size distribution of battery slurry |
| CN115796011A (en) * | 2022-11-16 | 2023-03-14 | 四川大学 | Hydrogen storage bed heat transfer performance optimization method based on neural network and genetic algorithm |
| CN116611349A (en) * | 2023-07-18 | 2023-08-18 | 华东交通大学 | Neural network-based roller wire drying process parameter optimization method and system |
| CN117408166A (en) * | 2023-12-14 | 2024-01-16 | 沧州经济开发区武理工京津冀协同产业科技研究院 | Method for predicting concentration of dissolved gas in transformer oil |
| CN117852611A (en) * | 2024-03-07 | 2024-04-09 | 哈尔滨工业大学（威海） | Ionosphere F2 layer maximum height prediction method and system |
| CN118211755A (en) * | 2024-03-30 | 2024-06-18 | 常州通用自来水有限公司 | Genetic algorithm optimization-based BP neural network secondary water supply flow prediction method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1359646A (en) * | 2000-12-21 | 2002-07-24 | 青岛海洋大学 | Method for creating fuzzy-neural network expert system for evaluating sensing quality of cigarette |
| CN101385571A (en) * | 2008-10-27 | 2009-03-18 | 川渝中烟工业公司 | Cigarette tobacco leaf formulation design method |
| CN102854289A (en) * | 2012-10-11 | 2013-01-02 | 云南烟草科学研究院 | Method for predicting release amount of phenol in mainstream smoke of flue-cured tobaccos |
-
2018
- 2018-07-27 CN CN201810839016.XA patent/CN109034388A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1359646A (en) * | 2000-12-21 | 2002-07-24 | 青岛海洋大学 | Method for creating fuzzy-neural network expert system for evaluating sensing quality of cigarette |
| CN101385571A (en) * | 2008-10-27 | 2009-03-18 | 川渝中烟工业公司 | Cigarette tobacco leaf formulation design method |
| CN102854289A (en) * | 2012-10-11 | 2013-01-02 | 云南烟草科学研究院 | Method for predicting release amount of phenol in mainstream smoke of flue-cured tobaccos |
Non-Patent Citations (3)
| Title |
|---|
| L. B. FAZLIC ET AL.: "GA-ANFIS expert system prototype for detection of tar content in the manufacturing process", 《2015 38TH INTERNATIONAL CONVENTION ON INFORMATION AND COMMUNICATION TECHNOLOGY, ELECTRONICS AND MICROELECTRONICS (MIPRO)》 * |
| 张忠良 等: "基于DE_BPANN的单料烟烟气指标预测方法", 《烟草科技》 * |
| 徐翔禹: "基于数据挖掘的单料烟感官评吸指标预测硏究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》 * |
Cited By (40)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110059821A (en) * | 2019-03-19 | 2019-07-26 | 埃睿迪信息技术（北京）有限公司 | Neural network optimization, device, computer equipment and storage medium |
| CN110457758B (en) * | 2019-07-16 | 2023-03-24 | 江西理工大学 | Method, device and system for predicting unstable phase of rock mass and storage medium |
| CN110457758A (en) * | 2019-07-16 | 2019-11-15 | 江西理工大学 | Prediction technique, device, system and the storage medium in Instability of Rock Body stage |
| CN112308278A (en) * | 2019-08-02 | 2021-02-02 | 中移信息技术有限公司 | Method, device, equipment and medium for optimizing prediction model |
| CN110490244A (en) * | 2019-08-14 | 2019-11-22 | 吉林大学 | A kind of data processing method and device |
| CN110516954A (en) * | 2019-08-23 | 2019-11-29 | 昆明理工大学 | One kind referring to calibration method based on GA-BP neural network algorithm optimization mineral processing production |
| CN110656385A (en) * | 2019-10-28 | 2020-01-07 | 季华实验室 | Neural network-based electrostatic spinning fiber average diameter prediction method |
| CN110656385B (en) * | 2019-10-28 | 2021-01-12 | 季华实验室 | Neural network-based electrostatic spinning fiber average diameter prediction method |
| CN111126560A (en) * | 2019-11-07 | 2020-05-08 | 云南民族大学 | Method for optimizing BP neural network based on cloud genetic algorithm |
| CN110826229B (en) * | 2019-11-07 | 2023-06-09 | 长沙学院 | Modeling and optimizing setting control method for filament drying head drying process based on long-term prediction |
| CN110826229A (en) * | 2019-11-07 | 2020-02-21 | 长沙学院 | Cut tobacco drying head process modeling and optimal setting control method based on long-term prediction |
| CN111242278A (en) * | 2019-12-29 | 2020-06-05 | 杭州拓深科技有限公司 | Composite smoke sense low-false-alarm method based on intelligent algorithm |
| CN111210877B (en) * | 2020-01-03 | 2023-11-24 | 山东天岳先进科技股份有限公司 | Method and device for deducing physical parameters |
| CN111210877A (en) * | 2020-01-03 | 2020-05-29 | 山东天岳先进材料科技有限公司 | Method and device for deducing physical property parameters |
| CN113077041A (en) * | 2020-01-06 | 2021-07-06 | 大唐环境产业集团股份有限公司 | Deep neural network modeling independent variable dimension reduction method based on genetic algorithm |
| CN113779858A (en) * | 2020-06-10 | 2021-12-10 | 国家能源投资集团有限责任公司 | Combustion optimization method and system, storage medium and electronic equipment |
| CN111950099A (en) * | 2020-08-03 | 2020-11-17 | 中国石油大学(华东) | Method, system, medium and computer equipment for testing mechanical property of equipment material |
| CN112016249A (en) * | 2020-08-31 | 2020-12-01 | 华北电力大学 | SCR denitration system bad data identification method based on optimized BP neural network |
| CN112099355A (en) * | 2020-09-18 | 2020-12-18 | 湖北中烟工业有限责任公司 | BP-PID-based tobacco sheet coating rate control method and device |
| CN112231978A (en) * | 2020-10-15 | 2021-01-15 | 西安特种设备检验检测院 | Boiler flue gas acid dew point testing method based on artificial neural network |
| CN112231978B (en) * | 2020-10-15 | 2022-01-28 | 西安特种设备检验检测院 | Boiler flue gas acid dew point testing method based on artificial neural network |
| CN112464554A (en) * | 2020-11-03 | 2021-03-09 | 桂林理工大学 | Operating parameter optimization method of gasoline refining equipment |
| CN112456639A (en) * | 2020-11-10 | 2021-03-09 | 盐城工学院 | Intelligent prediction method for hydrogen sulfide production in anaerobic digestion |
| CN112456639B (en) * | 2020-11-10 | 2022-11-22 | 盐城工学院 | Intelligent prediction method for hydrogen sulfide generation amount in anaerobic digestion |
| CN112464566A (en) * | 2020-12-07 | 2021-03-09 | 广东电网有限责任公司佛山供电局 | Transformer oil temperature prediction method based on genetic algorithm and BP neural network |
| CN112464566B (en) * | 2020-12-07 | 2023-01-24 | 广东电网有限责任公司佛山供电局 | Transformer oil temperature prediction method based on genetic algorithm and BP neural network |
| CN112485622A (en) * | 2020-12-09 | 2021-03-12 | 西安华谱电力设备制造有限公司 | Partial discharge pattern recognition method based on GA-BPNN |
| CN112765883A (en) * | 2021-01-18 | 2021-05-07 | 电子科技大学 | Method for determining valve closing process based on genetic algorithm and neural network |
| CN113268927A (en) * | 2021-05-21 | 2021-08-17 | 哈尔滨工业大学 | High-power laser device output energy prediction method based on full-connection neural network |
| CN113268927B (en) * | 2021-05-21 | 2024-04-30 | 哈尔滨工业大学 | Output energy prediction method of high-power laser device based on fully-connected neural network |
| CN113449930A (en) * | 2021-07-27 | 2021-09-28 | 威海长和光导科技有限公司 | Optical fiber preform preparation quality prediction method based on BP neural network |
| CN113887636A (en) * | 2021-10-09 | 2022-01-04 | 四川大学 | Selectable data enhancement method and system based on genetic algorithm |
| CN114354455A (en) * | 2022-01-17 | 2022-04-15 | 北京石油化工学院 | Method for online measurement of particle size distribution of battery slurry |
| CN114354455B (en) * | 2022-01-17 | 2023-12-08 | 北京石油化工学院 | Method for online measurement of size distribution of battery slurry |
| CN115796011A (en) * | 2022-11-16 | 2023-03-14 | 四川大学 | Hydrogen storage bed heat transfer performance optimization method based on neural network and genetic algorithm |
| CN116611349A (en) * | 2023-07-18 | 2023-08-18 | 华东交通大学 | Neural network-based roller wire drying process parameter optimization method and system |
| CN116611349B (en) * | 2023-07-18 | 2023-10-10 | 华东交通大学 | Neural network-based roller wire drying process parameter optimization method and system |
| CN117408166A (en) * | 2023-12-14 | 2024-01-16 | 沧州经济开发区武理工京津冀协同产业科技研究院 | Method for predicting concentration of dissolved gas in transformer oil |
| CN117852611A (en) * | 2024-03-07 | 2024-04-09 | 哈尔滨工业大学（威海） | Ionosphere F2 layer maximum height prediction method and system |
| CN118211755A (en) * | 2024-03-30 | 2024-06-18 | 常州通用自来水有限公司 | Genetic algorithm optimization-based BP neural network secondary water supply flow prediction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109034388A (en) | A kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network | |
| CN107633301B (en) | A kind of the training test method and its application system of BP neural network regression model | |
| CN107689224A (en) | The deep neural network compression method of reasonable employment mask | |
| CN102622418B (en) | Prediction device and equipment based on BP (Back Propagation) nerve network | |
| CN109002917A (en) | Total output of grain multidimensional time-series prediction technique based on LSTM neural network | |
| CN104636801A (en) | Transmission line audible noise prediction method based on BP neural network optimization | |
| CN101599138A (en) | Land evaluation method based on artificial neural network | |
| CN103559537B (en) | Based on the template matching method of error back propagation in a kind of out of order data stream | |
| CN107346459A (en) | One kind is based on the improved multi-mode pollutant Integrated Forecast Method of genetic algorithm | |
| CN106447092A (en) | Marine reverse osmosis desalination system performance prediction method based on MEA-BP neural network | |
| CN104346389A (en) | Scoring method and system of semi-open-ended questions of oral test | |
| CN106897744A (en) | A kind of self adaptation sets the method and system of depth confidence network parameter | |
| CN112634019A (en) | Default probability prediction method for optimizing grey neural network based on bacterial foraging algorithm | |
| CN104865827A (en) | Oil pumping unit oil extraction optimization method based on multi-working-condition model | |
| CN116167415A (en) | Policy decision method in multi-agent cooperation and antagonism | |
| CN113705098A (en) | Air duct heater modeling method based on PCA and GA-BP network | |
| CN105259754A (en) | Board thickness intelligent control method based on active learning | |
| da Costa Martins et al. | Nonlinear system identification based on modified ANFIS | |
| CN102063640B (en) | Robot behavior learning model based on utility differential network | |
| CN115586801B (en) | Gas blending concentration control method based on improved fuzzy neural network PID | |
| CN117196368A (en) | Public staff assessment method and system based on GA-BP algorithm | |
| CN115879412A (en) | Layout level circuit diagram size parameter optimization method based on transfer learning | |
| CN112989812B (en) | Cloud data center-based distributed poetry generation method | |
| Qin et al. | Sales forecast based on BP neural network | |
| CN113426109B (en) | Method for cloning chess and card game behaviors based on factorization machine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181218 |