CN106529584A - Flue-cured tobacco aroma type and quality judgment intelligent evaluation method - Google Patents
Flue-cured tobacco aroma type and quality judgment intelligent evaluation method Download PDFInfo
- Publication number
- CN106529584A CN106529584A CN201610933870.3A CN201610933870A CN106529584A CN 106529584 A CN106529584 A CN 106529584A CN 201610933870 A CN201610933870 A CN 201610933870A CN 106529584 A CN106529584 A CN 106529584A
- Authority
- CN
- China
- Prior art keywords
- grrf
- node
- flue
- fragrance
- evaluation
- 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
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
Abstract
The invention relates to a flue-cured tobacco aroma type and quality judgment intelligent evaluation method. The method is characterized by screening out aroma classification and aroma score principal component factors from tobacco aroma substances through a GRRF method; establishing a tobacco aroma style judgment model through GRRF+SVM and GRRF+RF methods; analyzing evaluation performance of the two methods through a 10 fold cross-validation method, wherein for classification evaluation results, adopting accuracy ACC and receiver operating characteristic curve as evaluation criterion, and for regression evaluation results, adopting Pearson's correlation coefficient and mean square error as evaluation criterion, and determining an optimal plan of a tobacco aroma type judgment model; and on the basis of the above, constructing a tobacco aroma type judgment online tool and carrying out evaluation on the aroma type and quality of tobaccos to be detected. The method can carry out analysis and judgment on aroma type characteristics of the tobaccos in different ecological areas objectively and accurately.
Description
Technical field
The present invention relates to the intelligence of the field of flue-cured tobacco flavor types and quality judging, particularly a kind of flue-cured tobacco flavor types and quality judging
Evaluation methodology.
Background technology
What the main body that Nicotiana tabacum L. is constituted as cigarette composition, its quality good or not and style characteristic directly determined cigarette sucks spy
Property.In the course of processing of cigarette products, the raw tobacco material of different parts, different-style characteristic is needed to constitute metastable leaf
Group of formula, and acted on by the mediation of adjuvant (flavouring essence for tobacco, spice), form the distinctive smoking property of cigarette characteristic of itself brand.
During this, stable Medicated cigarette style (sucking characteristic), the physiology of smoker is met and cigarette enterprise stable development all
It is particularly important, therefore, the evaluation to tobacco style characteristic and quality uses the extensive concern by Cigarette Industrial Enterprise.
Fresh tobacco leaves form the cigarette required for cigarette products processing after just roasting, redrying processing and the process refined
Leaf raw material.The intrinsic chemical composition and flavor matter component content of Nicotiana tabacum L. directly determines its inherent quality and aromatic style characteristic,
And then affect status and use of the raw tobacco material in cigarette composition.Although people are to flavor matter content and Nicotiana tabacum L. in Nicotiana tabacum L.
The research of relation between quality has been achieved for very big progress, but the evaluation to quality of tobacco and aromatic style is still led at present
Cross Nicotiana tabacum L. smoke panel test expert manually smoked panel test score evaluate.Domestic cured tobacco leaf style can be divided in terms of big delicate fragrance type,
The three types such as flat-flavor and Luzhou-flavor;And note can be divided into fresh and sweet fragrant, just fragrant and sweet and Jiao fragrant and sweet etc. more than ten and plant;Inherent quality
Marking calculating is carried out by aroma quality, perfume quantity, zest, miscellaneous QI, the ash Sensory quality index that grades.
Due to the multiformity and complexity of compound contained by cured tobacco leaf and during manually smokeing panel test by environment, state and
Subjective will etc. affects, and different degrees of deviation occurs, or even occur in the evaluation to inherent quality of tobacco leaf and style characteristic
Same sample difference is smoked panel test the difference of Type of aroma between personnel evaluation result.At present, domestic Cigarette Industrial Enterprise is to quality of tobacco
With not yet form unified evaluation criterion in the evaluation of style characteristic, it is special by itself enterprise's Medicated cigarette style more than the evaluation criterion of Nicotiana tabacum L.
The impact of color.
In sum, the method for evaluating quality of tobacco and aromatic style characteristic using manually smokeing panel test at present, easily by environment
Affect with individual factor etc., the evaluation and effectively utilizes and industrial undertaking's cigarette quality and style characteristic to raw tobacco material is stablized
Very big impact is brought, while increasing the harm brought to personnel's own health of smokeing panel test to a certain extent.
The content of the invention
It is an object of the invention to provide a kind of intelligent Evaluation method of flue-cured tobacco flavor types and quality judging, to overcome existing skill
Defect present in art.
For achieving the above object, the technical scheme is that:A kind of intelligent Evaluation side of flue-cured tobacco flavor types and quality judging
Method, is realized in accordance with the following steps:
Step S1:The main constituent factor of tabacco fragrance classification and fragrance score is screened with GRRF methods, from flavor matter
In filtered out the main constituent factor of odor classification and fragrance score;
Step S2:RF is used on the basis of GRRF using the method and employing with SVM classifier on the basis of GRRF
The method of grader sets up tabacco fragrance style discrimination model;
Step S3:The evaluation performance of above two method is adopted with the analysis of 10 folding cross-validation methods, wherein, for classification
Evaluation result, using accuracy ACC and Receiver operating curve as evaluation criterion;For the evaluation knot for returning
Really, using Pearson's correlation coefficient and mean square error as evaluation criterion, determine tabacco fragrance style discrimination model;
Step S4:Other model is judged by the tabacco fragrance style for determining, tabacco fragrance style is set up and is differentiated instrument online,
The odor type and quality of Nicotiana tabacum L. to be detected are evaluated.
In an embodiment of the present invention, in step S1, the utilization GRRF methods screening is realized in accordance with the following steps
The main constituent factor that tabacco fragrance is classified with fragrance score:
Step S11:One Random Forest model is built to sample set;
Step S12:After the completion of random forest builds, GRRF is carried out to all nodes of original random forest again time
Go through, and calculate variable importance;For each variable XiModel all provides corresponding importance score value, importance score value according to
Lower formula is calculated:
Wherein, k is decision tree quantity in forest, xiFor i-th variable, v is x used in treeiThe node of division, Gain
(Xi, V) be the node Gini impurity levels reduction amount;
Step S13:For Gain (Xi, V), process as follows:
Wherein, F represents the set of the feature selected in traverse node, and λ is to be selected into feature to receive punishment
Penalty coefficient;λ is bigger, and punishment dynamics are less, and λ is calculated as follows:
λ=1- γ (1-imp ');
Wherein, γ is that GRRF needs unique parameter for determining, imp ' is the importance scores that original random forest is given.
Step S14:The importance scores of each variable that GRRF is calculated according to Jing above-mentioned steps, and variable is important
Property the variable main constituent factor as GRRF chosen of the fraction more than zero.
In an embodiment of the present invention, in step S12, for the node v, from node v to its left and right two
Child node, Gini impurity level decrements are:
Gain(xi, v)=Gini (xi,v)-(wRGini(xi,vR)+wLGini(xi,vL));
Wherein, R, L represent child node W of left and right two of the node v respectivelyRAnd WLFor two child nodes Gini in left and right
The weighted value of impurity level decrement;
For the node v, Gini impurity levels are calculated as follows:
Wherein, pcFor the probability of classification c in the node v, it is the sample size of classification c divided by present node v
The sum of sample.
In an embodiment of the present invention, in step S1, screen from 92 kinds of flavor matters 22 kinds of odor classifications and
The main constituent factor of 19 kinds of fragrance scores.
In an embodiment of the present invention, tabacco fragrance is being set up using the method with SVM classifier on the basis of GRRF
During style discrimination model, an optimal separating hyper plane is found, make edge between two classes between adjacent nearest sample point most
Bigization;Sample point on margo is maximized is referred to as supporting vector, and the middle tangent plane at edge is optimal separating hyper plane;
The point for being divided by edge by mistake, its weight will be lowered, to reduce its impact;When data linearly inseparable, by kernel function by number
Strong point is mapped to higher dimensional space so as to linear separability.
In an embodiment of the present invention, the method with RF graders on the basis of using GRRF builds tabacco fragrance style
During discrimination model, three parameters are set:In forest, quantity ntree of decision tree, internal node randomly choose the individual of attribute
The smallest sample number nodesize of number mtry and terminal note;
Tabacco fragrance style discrimination model is set up as follows:
(1) concentrated from initial data using bootstrap sampling techniquess and extract ntree training set;
(2) post-class processing is set up respectively for each bootstrap training set, common property gives birth to ntree decision tree, and structure
Into a piece of forest, these decision trees do not carry out beta pruning, as unpruned yet;In each tree growth course, it is not choosing
The optimum attributes selected in whole M attributes carry out branch split as internal node, but from randomly selected mtry≤M
Optimum attributes are selected to carry out branch in attribute;
(3) set ntree decision tree predicts the outcome, and the classification of new samples is determined by the way of ballot.
In an embodiment of the present invention, online tool is differentiated using tabacco fragrance style described in PHP and R platform constructions.
Compared to prior art, the invention has the advantages that:
(1) present invention to tabacco fragrance type and showing in the evaluation of degree, broken away from because environment and it is artificial it is subjective because
The impact of element, compares artificial smoking result more objective, accurate.
(2) application of the invention can be to set up the unified mark that kinds of cigarettes industrial undertaking quality of tobacco and style characteristic are evaluated
It is accurate that theoretical foundation and technical support are provided.
(3) screening of the main constituent factor of the present invention on affecting tabacco fragrance style characteristic in flavor matter, to a certain degree
On reduce testing cost.
(4) the further apparent contribution degree of flavor matter factor pair aromatic style characteristic of the present invention, is to instruct flue-cured tobacco kind
Plant and Optimizing Tobacco production measure provide effectively reference, the evaluation system set up can be set up unified quality of tobacco and
Style characteristic evaluation methodology provides scientific basis and technical support.
(5) invention compared for manually smokeing panel test on to tabacco fragrance Style Evaluation, it is to avoid the personnel that smoke panel test excessively take out
The injury that smoking is caused to body.
Description of the drawings
Fig. 1 is Liu's flow chart of a kind of intelligent Evaluation method of flue-cured tobacco flavor types and quality judging in the present invention.
Fig. 2 is tabacco fragrance Style Evaluation system homepage in one embodiment of the invention.
Fig. 3 is DXY010 Sample intraocular's smoking result schematic diagrams in one embodiment of the invention.
Fig. 4 is that the input page of tabacco fragrance stylistic category differentiation main constituent content of material in one embodiment of the invention is illustrated
Figure.
Fig. 5 is the differentiation results page schematic diagram of tobacco sample odor type style in one embodiment of the invention.
Fig. 6 is the input page schematic diagram of tabacco fragrance score differentiation main constituent content of material in one embodiment of the invention.
Fig. 7 is the differentiation results page schematic diagram of tobacco sample fragrance score in one embodiment of the invention.
Specific embodiment
Below in conjunction with the accompanying drawings, technical scheme is specifically described.
The present invention provides a kind of intelligent Evaluation method of flue-cured tobacco flavor types and quality judging, realizes in accordance with the following steps:
Step S1:With GRRF, (Guided Regularized Random Forest, that is, have the regularization of guiding random
Forest) method screening tabacco fragrance classification and the main constituent factor of fragrance score, 22 kinds have been filtered out from 92 kinds of flavor matters
The main constituent factor of odor classification and 19 kinds of fragrance scores;
Step S2:Tabacco fragrance wind is set up with the method for GRRF+SVM (support vector machine) and GRRF+RF (random forest)
Lattice discrimination model;
Step S3:The evaluation of 2 kinds of methods is analyzed with 10 folding cross validations (10 fold Cross-validation) method
Performance, wherein, for the evaluation result of classification, using accuracy ACC (Accuracy) and Receiver operating curve
(ROC) as evaluation criterion;For the evaluation result for returning, made using Pearson's correlation coefficient (PCC) and mean square error (MSE)
For evaluation criterion, tabacco fragrance style discrimination model is determined;In the present embodiment, it is determined that tobacco incense is set up using GRRF+SVM
Gas style discrimination model;
Step S4:Tabacco fragrance style differentiates online tool, the perfume (or spice) to Nicotiana tabacum L. to be detected using PHP and R platform constructions
Type and quality are evaluated.
Further, in the present embodiment, the main constituent of tabacco fragrance classification and fragrance score is screened with GRRF methods
During the factor, also comprise the steps:
Step S11:A Random Forest model is built to sample set first before using GRRF.
Step S12:After the completion of random forest builds, GRRF can be carried out to all nodes of original random forest again time
Go through and calculate variable importance.For each variable XiModel all gives corresponding importance score value, importance score value according to
Below equation is calculated:
Wherein, k is decision tree quantity in forest, xiFor i-th variable, v is x used in treeiThe node of division, Gain
(Xi, V) be the node Gini impurity levels reduction amount.
Further, for node v, from node v to its two child node in left and right, Gini impurity level decrements are:
Gain(xi, v)=Gini (xi,v)-(wRGini(xi,vR)+wLGini(xi,vL));
Wherein, R, L represent child node W of left and right two of v respectivelyRAnd WLSubtract for the two child node Gini impurity levels in left and right
A small amount of weighted value.
For node v, Gini impurity levels are calculated as follows:
Wherein, pcFor the probability of classification c in node v, be classification c sample size divided by present node v sample it is total
Number.
Step S13:For Gain (Xi, V), to also process as follows:
Wherein, F represents the set of the feature selected in traverse node.Need to connect when certain feature is selected into
By a certain degree of punishment, the size of punishment is determined that by coefficient lambda λ is bigger, and punishment dynamics are less, and λ is calculated as follows:
λ=1- γ (1-imp ');
Wherein, γ is that GRRF needs unique parameter for determining, according to experimental result, during γ=0.1, the feature of selection makes mould
Type maintains preferable accuracy.Wherein imp ' is the importance scores that original random forest is given.
Step S14:Final GRRF also can provide the importance scores of each variable, those variable weights according to above-mentioned principle
Variable of the property the wanted fraction more than zero is exactly the main constituent that GRRF chooses.
Further, in the present embodiment, aromatic style is built with RF (random forest) grader on the basis of GRRF
Discrimination model, needs to arrange three main parameters in the process:In forest quantity ntree of decision tree, internal node with
Machine selects the smallest sample number nodesize of number mtry and terminal note of attribute.Its basic thought is:
One is to concentrate to extract ntree training set using bootstrap sampling techniquess from initial data;
Two is to set up post-class processing respectively for each bootstrap training set, and common property is given birth to ntree decision tree and constituted
A piece of " forest ", these decision trees do not carry out beta pruning (unpruned).In each tree growth course, it is not to select all
Optimum attributes in M attribute carry out branch (split) as internal node, but from randomly selected mtry≤M attribute
Middle selection optimum attributes carry out branch;
Three is predicting the outcome for ntree decision tree of set, and the class of new samples is determined by the way of ballot (voting)
Not.
Further, in the present embodiment, fragrance wind is built with MVB (support vector machine) grader on the basis of GRRF
Lattice discrimination model, basic thought are to find an optimal separating hyper plane (separating hyperplane), make phase between two classes
Edge (margin) between adjacent nearest sample point is maximized.Maximize margo on sample point be referred to as support to
Amount (support vectors), the middle tangent plane at edge is optimal separating hyper plane.By edge point by mistake point, its weight will be by
Reduce, to reduce its impact.When data linearly inseparable, by kernel function by Mapping of data points to higher dimensional space so as to line
Property can be divided.
Further, in the present embodiment, tabacco fragrance style discrimination model is set up based on GRRF+SVM, using PHP and R
Platform construction tabacco fragrance style differentiates online tool.As shown in Fig. 2 at the top of tabacco fragrance Style Evaluation system homepage
Selection operation item is provided, including:Integrated quality evaluation, odor type evaluation, using guide, function introduction and data download.
As shown in figure 3, introducing Nicotiana tabacum L. by taking the sample in DXY010 (the overseas village in Yuxi Xin Ping counties Xin Hua small towns) as an example below
The use of quality online rating system and the accuracy of constructed model evaluation.
First, in selecting tabacco fragrance Style Evaluation system homepage, odor type evaluates menu, as shown in Figure 4.Then, according to
Popup menu input frame is pointed out, after being input into 22 kinds of flavor matter content detection results of decision Type of aroma of DXY010 samples,
Click on and submit data to.As a result it is as shown in Figure 5:The Type of aroma for obtaining DXY010 samples is " delicate fragrance type ".Obtain DXY010 Nicotiana tabacum Lves
After the aromatic style types results of sample, the fragrance scoring event of the tobacco sample such as need to be further obtained, please click on " after
It is continuous ", tobacco aroma score main constituent content of material input page is will appear from, as shown in Figure 6.Then, it is input into according to popup menu
Frame is pointed out, and after being input into 19 kinds of flavor matter content detection results of decision fragrance score of DXY010 samples, is clicked on and " is submitted number to
According to ".As a result it is as shown in Figure 7:The fragrance for obtaining DXY010 samples must be divided into " 4.4 ".
In the present embodiment, the main constituent factor of 22 kinds of odor classifications that table 1 is obtained for screening, table 2 are obtained for screening
The main constituent factor of 19 kinds of fragrance scores.
Table 1
Sequence number | Material title | Sequence number | Material title |
X1 | 1- penten-3-ones | X40 | Dorinone |
X3 | 3- methyl-1-butanols | X43 | Geranyl acetone |
X4 | Pyridine | X49 | Megastigmatrienone B |
X5 | 3- methyl-2-butene aldehydes | X55 | Myristic acid |
X9 | Furfuryl alcohol | X63 | Palmic acid |
X10 | 2- cyclopentenes-Isosorbide-5-Nitrae-diketone | X78 | Octanoic acid |
X14 | Furancarboxylic acid | X80 | Oxalic acid |
X29 | 1- (3- pyridine radicals)-ethyl ketone | X83 | Malic acid |
X31 | Oxidation isophorone | X84 | Citric acid |
X36 | 2,3- Dihydrobenzofuranes | X89 | Stearic acid |
X38 | 2- methoxyl-4-vinyl phenols | X91 | Scopoletin |
Table 2
Sequence number | Material title | Sequence number | Material title |
X1 | 1- penten-3-ones | X54 | Methyl myristate |
X8 | Furfural | X60 | Dibutyl phthalate |
X9 | Furfuryl alcohol | X65 | It is very little to visit alcohol |
X15 | Benzaldehyde | X66 | Methyl linolenate |
X20 | 4- pyridine carboxaldehydes | X74 | Valeric acid |
X30 | Phenethanol | X80 | Oxalic acid |
X40 | Dorinone | X89 | Stearic acid |
X43 | Geranyl acetone | X90 | Chlorogenic acid |
X44 | Alpha, beta-lonone | X91 | Scopoletin |
X47 | Dihydroactinidiolide |
Can be seen that according to the above results, the intelligent evaluation system of flue-cured tobacco flavor types and quality judging is to tabacco fragrance stylistic category
Judgement with fragrance score is basically identical with artificial smoking result in Fig. 3, and artificial smoking result is 4.5, and both differences are 0.1,
Within error allowed band.Show that the system objective, accurate can judge the aromatic style type and fragrance score of Nicotiana tabacum L.,
Can be used as the judgement for carrying out aromatic style type and fragrance score in raw material of cigarette formula to raw tobacco material.
It is more than presently preferred embodiments of the present invention, all changes made according to technical solution of the present invention, produced function are made
During with scope without departing from technical solution of the present invention, protection scope of the present invention is belonged to.
Claims (7)
1. a kind of intelligent Evaluation method of flue-cured tobacco flavor types and quality judging, it is characterised in that realize in accordance with the following steps:
Step S1:The main constituent factor of tabacco fragrance classification and fragrance score is screened with GRRF methods, is sieved from flavor matter
Have selected the main constituent factor of odor classification and fragrance score;
Step S2:Classify with RF using the method that SVM classifier is used on the basis of GRRF and using on the basis of GRRF
The method of device sets up tabacco fragrance style discrimination model;
Step S3:The evaluation performance of above two method is adopted with the analysis of 10 folding cross-validation methods, wherein, for commenting for classification
Valency result, using accuracy ACC and Receiver operating curve as evaluation criterion;For the evaluation result for returning, adopt
With Pearson's correlation coefficient and mean square error as evaluation criterion, tabacco fragrance style discrimination model is determined;
Step S4:Other model is judged by the tabacco fragrance style for determining, tabacco fragrance style is set up and is differentiated instrument online, treat
The odor type and quality of detection Nicotiana tabacum L. is evaluated.
2. the intelligent Evaluation method of a kind of flue-cured tobacco flavor types according to claim 1 and quality judging, it is characterised in that in institute
State step S1, realize in accordance with the following steps the main constituent of utilization GRRF methods screening tabacco fragrance classification and fragrance score because
Son:
Step S11:One Random Forest model is built to sample set;
Step S12:After the completion of random forest builds, GRRF is traveled through again to all nodes of original random forest, and
Calculate variable importance;For each variable XiModel all provides corresponding importance score value, and importance score value is according to following public affairs
Formula is calculated:
Wherein, k is decision tree quantity in forest, xiFor i-th variable, v is x used in treeiThe node of division, Gain (Xi,V)
It is the reduction amount of the Gini impurity levels of the node;
Step S13:For Gain (Xi, V), process as follows:
Wherein, F represents the set of the feature selected in traverse node, and λ is to be selected into the punishment that feature receives punishment
Coefficient;λ is bigger, and punishment dynamics are less, and λ is calculated as follows:
λ=1- γ (1-imp ');
Wherein, γ is that GRRF needs unique parameter for determining, imp ' is the importance scores that original random forest is given.
Step S14:The importance scores of each variable that GRRF is calculated according to Jing above-mentioned steps, and variable importance is divided
The main constituent factor that variable of the number more than zero is chosen as GRRF.
3. the intelligent Evaluation method of a kind of flue-cured tobacco flavor types according to claim 2 and quality judging, it is characterised in that in institute
State in step S12, for the node v, from node v to its two child node in left and right, Gini impurity level decrements are:
Gain(xi, v)=Gini (xi,v)-(wRGini(xi,vR)+wLGini(xi,vL));
Wherein, R, L represent child node W of left and right two of the node v respectivelyRAnd WLIt is impure for two child nodes Gini in left and right
The weighted value of degree decrement;
For the node v, Gini impurity levels are calculated as follows:
Wherein, pcFor the probability of classification c in the node v, it is the sample size of classification c divided by sample in present node v
Sum.
4. the intelligent Evaluation method of a kind of flue-cured tobacco flavor types according to claim 2 and quality judging, it is characterised in that in institute
State in step S1, the main constituent factor of 22 kinds of odor classifications and 19 kinds of fragrance scores is screened from 92 kinds of flavor matters.
5. the intelligent Evaluation method of a kind of flue-cured tobacco flavor types according to claim 1 and quality judging, it is characterised in that adopting
During tabacco fragrance style discrimination model is set up with the method for SVM classifier on the basis of GRRF, a most optimal sorting is found
Class hyperplane, maximizes the edge between two classes between adjacent nearest sample point;Sample point on margo is maximized
It is referred to as supporting vector, the middle tangent plane at edge is optimal separating hyper plane;The point for being divided by edge by mistake, its weight will be lowered,
To reduce its impact;When data linearly inseparable, by kernel function by Mapping of data points to higher dimensional space so as to linearly may be used
Point.
6. the intelligent Evaluation method of a kind of flue-cured tobacco flavor types according to claim 1 and quality judging, it is characterised in that adopting
During tabacco fragrance style discrimination model is built with the method on the basis of GRRF with RF graders, three parameters are set:It is gloomy
The smallest sample number of quantity ntree of decision tree, number mtry of internal node random selection attribute and terminal note in woods
nodesize;
Tabacco fragrance style discrimination model is set up as follows:
(1) concentrated from initial data using bootstrap sampling techniquess and extract ntree training set;
(2) post-class processing is set up respectively for each bootstrap training set, common property gives birth to ntree decision tree, and constitutes one
Piece forest, these decision trees do not carry out beta pruning, as unpruned yet;In each tree growth course, it is not to select complete
Optimum attributes in M, portion attribute carry out branch split as internal node, but from randomly selected mtry≤M attribute
Middle selection optimum attributes carry out branch;
(3) set ntree decision tree predicts the outcome, and the classification of new samples is determined by the way of ballot.
7. the intelligent Evaluation method of a kind of flue-cured tobacco flavor types according to claim 1 and quality judging, it is characterised in that adopt
Tabacco fragrance style described in PHP and R platform constructions differentiates online tool.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610933870.3A CN106529584A (en) | 2016-10-25 | 2016-10-25 | Flue-cured tobacco aroma type and quality judgment intelligent evaluation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610933870.3A CN106529584A (en) | 2016-10-25 | 2016-10-25 | Flue-cured tobacco aroma type and quality judgment intelligent evaluation method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106529584A true CN106529584A (en) | 2017-03-22 |
Family
ID=58292205
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610933870.3A Pending CN106529584A (en) | 2016-10-25 | 2016-10-25 | Flue-cured tobacco aroma type and quality judgment intelligent evaluation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106529584A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108169420A (en) * | 2017-12-19 | 2018-06-15 | 云南瑞升烟草技术(集团)有限公司 | A kind of method using index characterization cigarette style characteristic difference degree |
CN109614714A (en) * | 2018-12-13 | 2019-04-12 | 上海应用技术大学 | A kind of method for establishing model of acidified milk fragrance aesthetic quality control |
CN110009191A (en) * | 2019-03-04 | 2019-07-12 | 中国地质大学(武汉) | A kind of flue-cured tobacco cultivation decision-making technique and system based on genetic algorithm |
CN110389200A (en) * | 2019-07-15 | 2019-10-29 | 浙江大学 | A kind of tea aroma of the same race differentiation detection method of different brands grade |
CN111311088A (en) * | 2020-02-11 | 2020-06-19 | 中国农业科学院烟草研究所 | Flue-cured tobacco aroma style characteristic detection method, system, storage medium and program |
CN113408945A (en) * | 2021-07-15 | 2021-09-17 | 广西中烟工业有限责任公司 | Method and device for detecting purity of flue-cured tobacco, electronic equipment and storage medium |
CN113488113A (en) * | 2021-07-12 | 2021-10-08 | 浙江中烟工业有限责任公司 | Industrial use value identification method of redried strip tobacco |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102435709A (en) * | 2011-09-12 | 2012-05-02 | 红云红河烟草(集团)有限责任公司 | Method for accurately identifying flue-cured tobacco flavor types |
CN105092789A (en) * | 2015-08-10 | 2015-11-25 | 江苏中烟工业有限责任公司 | Analysis method for tobacco leaf style characteristic variation trend in lamina aging process |
CN105447525A (en) * | 2015-12-15 | 2016-03-30 | 中国科学院软件研究所 | Data prediction classification method and device |
-
2016
- 2016-10-25 CN CN201610933870.3A patent/CN106529584A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102435709A (en) * | 2011-09-12 | 2012-05-02 | 红云红河烟草(集团)有限责任公司 | Method for accurately identifying flue-cured tobacco flavor types |
CN105092789A (en) * | 2015-08-10 | 2015-11-25 | 江苏中烟工业有限责任公司 | Analysis method for tobacco leaf style characteristic variation trend in lamina aging process |
CN105447525A (en) * | 2015-12-15 | 2016-03-30 | 中国科学院软件研究所 | Data prediction classification method and device |
Non-Patent Citations (3)
Title |
---|
DENG, HOUTAO等: "Gene selection with guided regularized random forest", 《PATTERN RECOGNITION》 * |
GU, LI等: "Classification of the fragrant styles and evaluation of the aromatic quality of flue-cured tobacco leaves by machine-learning methods", 《JOURNAL OF BIOINFORMATICS AND COMPUTATIONAL BIOLOGY》 * |
黄衍 等: "随机森林与支持向量机分类性能比较", 《软件》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108169420A (en) * | 2017-12-19 | 2018-06-15 | 云南瑞升烟草技术(集团)有限公司 | A kind of method using index characterization cigarette style characteristic difference degree |
CN109614714A (en) * | 2018-12-13 | 2019-04-12 | 上海应用技术大学 | A kind of method for establishing model of acidified milk fragrance aesthetic quality control |
CN110009191A (en) * | 2019-03-04 | 2019-07-12 | 中国地质大学(武汉) | A kind of flue-cured tobacco cultivation decision-making technique and system based on genetic algorithm |
CN110389200A (en) * | 2019-07-15 | 2019-10-29 | 浙江大学 | A kind of tea aroma of the same race differentiation detection method of different brands grade |
CN111311088A (en) * | 2020-02-11 | 2020-06-19 | 中国农业科学院烟草研究所 | Flue-cured tobacco aroma style characteristic detection method, system, storage medium and program |
LU102506A1 (en) * | 2020-02-11 | 2021-08-02 | Tobacco Res Inst Caas | A method, system, storage medium and program for detecting the flavor style characteristics of smoke hardened tobacco |
CN113488113A (en) * | 2021-07-12 | 2021-10-08 | 浙江中烟工业有限责任公司 | Industrial use value identification method of redried strip tobacco |
CN113488113B (en) * | 2021-07-12 | 2024-02-23 | 浙江中烟工业有限责任公司 | Industrial use value identification method for redried strip tobacco |
CN113408945A (en) * | 2021-07-15 | 2021-09-17 | 广西中烟工业有限责任公司 | Method and device for detecting purity of flue-cured tobacco, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106529584A (en) | Flue-cured tobacco aroma type and quality judgment intelligent evaluation method | |
CN104914720B (en) | Electronic nose intelligence baking control system and control method with autolearn feature | |
Voss et al. | Peach growth cycle monitoring using an electronic nose | |
CN102338780B (en) | Method for discriminating cigarette brands | |
CN108181263B (en) | Tobacco leaf position feature extraction and discrimination method based on near infrared spectrum | |
CN107392399A (en) | A kind of SVM Sensory Quality of Cigarette Forecasting Methodologies based on improved adaptive GA-IAGA | |
CN103344713B (en) | Method for analyzing importance of flavoring in flue-cured tobacco fragrance type classification and forecasting fragrance type based on random forest method | |
CN107463766A (en) | Generation method, device and the computer-readable recording medium of blood glucose prediction model | |
CN106295802A (en) | A kind of Folium Camelliae sinensis based on particle cluster algorithm Optimized BP Neural Network storage method chronological classification | |
CN105092789B (en) | Tobacco style changing features trend analysis in a kind of sheet nicotinyl alcohol process | |
CN105445421A (en) | Method for predicting sensory quality in lamina alcoholization process via appearance indexes | |
CN108414471B (en) | Method for distinguishing sensory characterization information based on near infrared spectrum and sensory evaluation mutual information | |
Voss et al. | E-nose Prototype to Monitoring the Growth and Maturation of Peaches in the Orchard | |
CN110132880B (en) | Tobacco leaf overall sensory quality evaluation method based on near infrared spectrum | |
CN105138834A (en) | Tobacco chemical value quantifying method based on near-infrared spectrum wave number K-means clustering | |
CN109324016A (en) | A kind of determination method of redried odor type style | |
CN110132879A (en) | A kind of flue-cured tobacco note discrimination method based near infrared spectrum | |
CN110990782B (en) | Cigarette sensory quality evaluation method based on weighted hesitation fuzzy power aggregation operator | |
CN108520276B (en) | Characterization method for internal sensory quality of tobacco leaf raw material | |
CN109781870B (en) | Method for identifying flavor type of flue-cured tobacco | |
CN102680427A (en) | Method for identifying surface fragrance quality of cigarettes by near infrared spectral analysis technology | |
CN109744580B (en) | Intelligent smoking control method and device | |
CN115983692A (en) | Tobacco leaf moistening and sweet flavor evaluation method and device based on random forest algorithm | |
CN110659691A (en) | Method and device for monitoring quality and style of tobacco leaf raw material and computer readable storage medium | |
CN114689746B (en) | Method, device, electronic equipment and medium for screening tobacco extract characteristics |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170322 |
|
RJ01 | Rejection of invention patent application after publication |