CN107341613B - Method for assisting balance replacement of leaf group formula - Google Patents
Method for assisting balance replacement of leaf group formula Download PDFInfo
- Publication number
- CN107341613B CN107341613B CN201710549990.8A CN201710549990A CN107341613B CN 107341613 B CN107341613 B CN 107341613B CN 201710549990 A CN201710549990 A CN 201710549990A CN 107341613 B CN107341613 B CN 107341613B
- Authority
- CN
- China
- Prior art keywords
- replacement
- raw materials
- raw material
- replaceable
- formula
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 239000002994 raw material Substances 0.000 claims abstract description 161
- 239000000463 material Substances 0.000 claims abstract description 33
- 238000011156 evaluation Methods 0.000 claims abstract description 18
- 238000010801 machine learning Methods 0.000 claims abstract description 11
- 235000019504 cigarettes Nutrition 0.000 claims description 14
- 238000004519 manufacturing process Methods 0.000 claims description 12
- 238000003066 decision tree Methods 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 7
- 239000000203 mixture Substances 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000001788 irregular Effects 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 230000002194 synthesizing effect Effects 0.000 claims description 2
- 238000009472 formulation Methods 0.000 claims 4
- 238000013461 design Methods 0.000 abstract description 6
- 238000012216 screening Methods 0.000 abstract description 6
- 238000012423 maintenance Methods 0.000 abstract description 2
- 238000012827 research and development Methods 0.000 abstract description 2
- 238000001914 filtration Methods 0.000 abstract 1
- 241000208125 Nicotiana Species 0.000 description 16
- 235000002637 Nicotiana tabacum Nutrition 0.000 description 16
- 238000004458 analytical method Methods 0.000 description 8
- 238000000926 separation method Methods 0.000 description 4
- 238000006467 substitution reaction Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000013441 quality evaluation Methods 0.000 description 2
- 201000004569 Blindness Diseases 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000007794 irritation Effects 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 235000013599 spices Nutrition 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- A—HUMAN NECESSITIES
- A24—TOBACCO; CIGARS; CIGARETTES; SIMULATED SMOKING DEVICES; SMOKERS' REQUISITES
- A24C—MACHINES FOR MAKING CIGARS OR CIGARETTES
- A24C5/00—Making cigarettes; Making tipping materials for, or attaching filters or mouthpieces to, cigars or cigarettes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Educational Administration (AREA)
- Animal Husbandry (AREA)
- General Health & Medical Sciences (AREA)
- Mining & Mineral Resources (AREA)
- Marine Sciences & Fisheries (AREA)
- Agronomy & Crop Science (AREA)
- Development Economics (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Manufacture Of Tobacco Products (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for assisting leaf group formula balance replacement, which is a raw material classification method for forming rules by converting artificial replacement experience informatization into screening conditions; selecting replaceable materials from existing inventory materials, screening and comparing the materials by using a clustering algorithm, submitting information of the replaceable materials with the most common characteristics as selectable replacement raw materials, and correcting and filtering deviations possibly occurring in characteristic selection by using historical replacement formula records; and integrating two screening results, substituting the screening results into an environment with complete formula requirements, associating the raw material data with complete formula design data and professional evaluation data by using an evaluation and prediction method of machine learning, carrying out prediction scoring on the screened replaceable raw materials, and finally, optimally selecting a reliable replaceable material selection list. The invention provides intuitive replacement selection for workers and scientific and informationized data prediction support for formula replacement, formula maintenance and formula research and development.
Description
Technical Field
The invention relates to a method for assisting balance replacement of a leaf group formula, and belongs to the technical field of cigarette leaf group formulas.
Background
As an agricultural product, tobacco leaves of different varieties, different regions and even different parts of the same plant have great differences in quality and style under the influence of multiple aspects such as self genetic genes, cultivation measures, soil conditions, climate factors, modulation methods and the like. Therefore, the various good quality factors of tobacco leaves cannot be present in the same or a few tobacco leaves. The quality factors of tobacco leaves of one variety are often contradictory. For example, some tobacco leaves have sufficient aroma, but heavy miscellaneous gas and large irritation; some tobacco leaves have moderate strength but insufficient aroma. Cigarettes made with a single variety or grade of tobacco have quality defects that cannot be overcome. Even high-grade tobacco leaves have difficulty in obtaining a completely satisfactory effect. Therefore, different characteristics of various tobacco leaves are required to be utilized, and the optimal leaf group formula combination is selected, so that various tobacco leaves participating in the formula can make good use of advantages and disadvantages, complement each other, and coordinately and consistently play respective roles.
China has made a lot of research and application in the aspect of leaf group formula design, but has not been practically applied, and still relies on the traditional experience of designers to design cigarette products, and the traditional experience is considered as follows: the tobacco leaves in the same area, the same variety, the same part and different grades can be mixed; and the tobacco leaf blending agent can also blend tobacco leaves with different areas, the same variety, the same part and stronger adaptability, and the like. However, practice proves that the traditional empirical formula is not beneficial to the development of new cigarette products, so that the research on algorithm design of cigarette products is a necessary trend in cigarette development on the basis of empirical design.
Disclosure of Invention
The invention aims to provide a method for assisting in balancing and replacing a leaf group formula aiming at the defects of the prior art, and aims to solve the problems that the traditional method for balancing and replacing a raw material leaf group formula is complex in process, comprehensive influence factors need to be considered in a combined manner, the accuracy of a test result fluctuates greatly, and time and labor are wasted.
The invention adopts the following technical scheme: a method for assisting balance replacement of leaf group formulas is characterized in that through large-batch historical data and current production data including raw materials, formulas, internal quality evaluation and other relevant data, data are cleaned and an incidence relation is extracted for research, the influence angle of raw material change on formula adjustment is analyzed, raw material formula change is analyzed at the same time, the condition that the quality is kept unchanged after the analysis of formula change is recorded, positive and negative influence algorithms of essence, spice, tobacco materials and production process parameter adjustment on the quality evaluation are applied, and finally a method for balance replacement of raw material leaf group by raw material quality information is found. The method specifically comprises the following steps:
(1) rule-based raw material classification: summarizing all data of raw material historical replacement, comparing and analyzing the frequency of each attribute change in the historical replacement data, and obtaining the ranking of the importance degree of the stock information and the raw material attribute identity information to the raw material replacement according to the frequency; establishing a decision tree according to the ranking of the importance degrees, taking the first important attribute as the maximum independent variable for improving the prediction effect, splitting the node firstly, then splitting the second important attribute for the second time, and so on, and finally finding out the replaceable raw material;
(2) classifying raw materials based on a clustering algorithm: sorting and summarizing the existing available stock raw materials, establishing a multi-dimensional space system, taking characteristic identity information of the raw materials as dimensions, generating a plurality of spatial coordinate points according to the dimensions of different characteristic attributes of each raw material, wherein the closer the dimensions of the coordinate points are, the closer the characteristic attributes of the raw materials are, and the higher the accuracy of replacement is; selecting a raw material with approximately convergent multidimensional characteristics from characteristic attributes of existing raw materials as a replaceable alternative through a clustering algorithm; then, correcting clustering deviation by using a historical replacement record to obtain a part of replaceable raw materials;
(3) evaluating and predicting based on a machine learning algorithm: and (3) combining the results of the replaceable raw materials in the step (1) and the step (2) to obtain an orthogonal set, wherein the obtained set is a final replaceable raw material list.
The higher the frequency of the step (1), the lower the influence degree of the attribute characteristics on the raw material replacement, and the lower the importance.
The inventory information in the step (1) is whether the user owns the inventory or not. Stock is a decisive factor for influencing replacement, which is in no way consistent with the actual situation, and stock replacement is not possible at all without stock.
The material attribute identity information in the step (1) comprises whether the trade mark, the variety information, the secondary producing area of the material, the year, the tertiary producing area of the material and the price are related.
Wherein, whether the associated mark is data generated according to the mark associated with the raw material, and if the raw material can be used for producing a plurality of marks, the associated mark is obtained; the raw materials are only used for producing cigarettes of one brand and are not related. Secondly, the influence of material grade association, if a certain replaced raw material is associated with other grades, the replacement does not occur as much as possible under the condition that other grades are considered to be preferentially used during replacement; thirdly, the variety replacement factor, certain similarity exists among the raw material attributes of certain varieties, and proper selection should be given during replacement; of course, the influence of the producing area, the year and the price cannot be ignored, the quality of the raw materials of adjacent or similar producing areas is similar, and the replacement can also be considered preferentially.
The characteristic identity information of the raw materials in the step (2) comprises year, producing area, grade, position and the like.
The clustering algorithm in the step (2) is a multi-dimensional space system algorithm which is a fuzzy and rapid algorithm; the specific algorithm is as follows: substituting characteristic identity information (year, part, place of production and the like) of required raw materials by a formula to serve as discrete points with different latitudes, carrying out set division on the given grouping number of the discrete data points under the same latitude to form an irregular stereogram, randomly extracting a numerical point k from the data set under the same latitude, calculating the central point of the distance from k to the next random numerical value, and setting the central point as k; randomly extracting the next point (third point) in a local weighting mode, and calculating the central point ko from k to the new point; sequentially calculating the distances from k to all non-midpoint points and the distances from ko to all non-midpoint points and the distances from Lo; and if L is greater than Lo, assigning k (k = ko) to ko, when all sets are calculated, finally regressing each set to obtain a midpoint and storing, correspondingly feeding back the characteristic identity information defined by each dimension to the center point of each dimension, traversing and searching the stock raw materials according to the calculated center value to obtain the same or most similar value, synthesizing the characteristic identity information of each dimension to give stock raw materials possibly existing or stock raw materials most similar to the characteristic factor in the calculation result, and sequencing according to the similar value.
Substituting the set obtained in the step (3) into an environment with complete formula replacement requirements, comparing professional evaluation scores of the cigarette formulas with the replaced raw materials correspondingly, and verifying; or, combining historical data, adopting the clustering algorithm in the step (2), continuously and circularly verifying the obtained set inclusion algorithm to determine the accuracy and stability of the formula balance replacement method, continuously summarizing replacement rules in a machine learning mode in the replacement process, optimizing a replacement method model, and improving the accuracy and reliability of the auxiliary leaf group formula balance replacement method.
The invention has the beneficial effects that: the invention selects, analyzes, evaluates and predicts in the existing inventory, and optimally screens out the optimal selection range. The method provides intuitive replacement selection for workers, provides scientific and informatization data prediction support for formula replacement, formula maintenance and formula research and development, solves the problems of blindness, low efficiency and the like, improves the use efficiency of formula raw materials, further optimizes the design of the leaf group formula, and achieves the aim of optimizing and improving the internal quality of the cigarettes.
Drawings
FIG. 1 is a graph of prediction of the importance of the feed replacement variables of example 1;
FIG. 2 is a graph showing the prediction of the frequency of the substitution variables of the raw material in example 1;
FIG. 3 is a decision tree of example 1;
FIG. 4 shows an example of the replacement of artificial experience materials for the model set No. 1 and Specification A of example 1;
FIG. 5 is a schematic diagram of the orthogonal set acquisition in step (3) of example 1;
FIG. 6 is a graph showing the prediction of the importance of the substitution variables of the feedstock in example 2;
FIG. 7 is a graph showing the prediction of the frequency of the substitution variables of the raw material in example 2;
FIG. 8 is a decision tree of example 2;
FIG. 9 shows an example of manual empirical raw material replacement of the model set No. 1 and Specification B in example 2.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples, which are not intended to limit the technical scope of the present invention.
Example 1
The substitution of the raw materials of '2013 flue-cured tobacco/Kunming 2/Hongda/WBBSF/FL/P/' in the xx cigarette factory specification A, No. 1 module is explained as follows:
(1) rule-based raw material classification: summarizing all data of raw material historical replacement, comparing and analyzing the frequency of each attribute change in the historical replacement data, and obtaining the ranking of the importance degree of the stock information and the raw material attribute identity information to the raw material replacement according to the frequency; establishing a decision tree according to the ranking of the importance degrees, taking the first important attribute as the maximum independent variable for improving the prediction effect, splitting the node firstly, then splitting the second important attribute for the second time, and so on, and finally finding out the replaceable raw material;
transverse pair analysis: summarizing all data of the historical replacement of the raw materials, comparing and analyzing the frequency of each attribute change in the historical replacement data, wherein the higher the replacement frequency is, the lower the influence degree of the attribute characteristics on the raw material replacement is, and the lower the importance is.
And (3) analysis results: the replaceable importance degree of the inventory information and the raw material attribute identity information to the raw materials sequentially comprises the existence of inventory, whether the inventory is associated with the brand, variety information, a secondary producing area of the raw materials, the year, a tertiary producing area of the raw materials and the price, wherein whether the associated brand is data generated by the system according to the brand associated with the raw materials, and if the raw materials can be used for producing a plurality of brands, the raw materials are associated brands; the raw materials are only used for producing cigarettes of one brand and are not related. The stock is a decisive factor influencing the replacement, is consistent with the actual situation without any doubt, and the raw material replacement cannot be carried out under the condition of no stock; secondly, the influence of material grade association, if a certain replaced raw material is associated with other grades, the replacement does not occur as much as possible under the condition that other grades are considered to be preferentially used during replacement; thirdly, the variety replacement factor, certain similarity exists among the raw material attributes of certain varieties, and proper selection should be given during replacement; of course, the influence of the producing area, the year and the price cannot be ignored, the quality of the raw materials of adjacent or similar producing areas is similar, and the replacement can also be considered preferentially. As shown in fig. 1 and 2.
And (5) drawing a conclusion that: year-to-year replacement is a relatively priority factor when replacement occurs, followed by replacement of varieties and tertiary producing areas, and finally secondary producing areas for replacement of raw materials are considered when other conditions cannot be met.
Establishing a decision tree prediction model: through analysis of historical artificial replacement experience, sorting according to the influence degree of the raw material attributes on replacement, and establishing a decision tree (as shown in figure 3): having inventory as the largest argument for improvement of the prediction effect is used first for splitting nodes; then, the second separation is carried out according to whether other brands are related or not, and then the third separation is carried out on the second-level producing area, and finally, the replaceable raw materials are found. The factors which are preferably considered during the raw material replacement are respectively inventory, brand association information and secondary production places, the screening is carried out layer by layer according to the mode, the manual replacement experience is converted into a replaceable raw material change rule, and part of replaceable materials are screened out.
As shown in fig. 4, is an artificial experience alternative. The replaced material was compared to the replacement material as follows:
therefore, the raw materials of the same variety, the same grade, the same production place and different years are selected for manual replacement. And a method for assisting leaf group formula balance replacement refers to historical replacement records by taking similar manual replacement experience as a rule to obtain partially selectable replacement materials.
(2) Classifying raw materials based on a clustering algorithm: sorting and summarizing the existing available stock raw materials, establishing a multi-dimensional space system, taking characteristic identity information (year, place of production, grade, position and the like) of the raw materials as dimensions, generating a plurality of spatial coordinate points according to the dimensions of different characteristic attributes of each raw material, wherein the closer the dimensions of the coordinate points are, the closer the characteristic attributes of the raw materials are, the higher the accuracy of replacement is; if the year is a dimension, the level is a dimension, and the position is a dimension, etc. The specific information of the characteristic attributes of the raw materials is the scale on the dimension, namely, the scale on the year dimension is 2009, 2010 and 2012 …, and the scale on the part dimension is provided with an upper part, a middle part and a lower part ….
Specifically, the dimensions in the multidimensional space system are not horizontal and vertical coordinates, and if the dimensions are not to be expressed as horizontal and vertical coordinates, any dimension in the multidimensional space system may be horizontal or vertical coordinates. That is, in the multidimensional space system, the horizontal and vertical coordinates are opposite, and as long as two dimensions are vertical, the two dimensions can be used as the x and y axes on the surface.
Points in space: selecting a raw material, wherein the specific description of the identity information of the raw material can generate a point in space according to different scales, and the point represents the raw material.
Selecting a raw material with approximately convergent multidimensional characteristics from characteristic attributes of existing raw materials as a replaceable alternative through a clustering algorithm; then, correcting clustering deviation by using a historical replacement record to obtain a part of replaceable raw materials; that is, by establishing a multidimensional space system, the position of the raw material in the space, which meets the requirement of the production inventory, is determined, and then the similarity of the raw material is judged by calculating the shortest distance between two different points in the space.
Taking coordinate points at the same latitude (same year, same part and the like) as a set, randomly extracting a numerical point K, calculating the distance from K to the next random data from a central point, and setting the central point as K; randomly extracting the next point (third point) in a local weighting mode, and calculating the central point ko from k to the new point; the sum of the distances k to all non-midpoints and L, and the sum of the distances ko to all non-midpoints and Lo are calculated in turn. If L > Lo, K = Ko is assigned; if L < Lo, K = K. That is, the center point where the distance sum is the smallest is taken as the relative center point K of the set.
Finally, the conclusion is drawn: when all the sets are calculated to obtain the central points, the distances between the central points are calculated to obtain new central points. And repeating continuously until a more concentrated set is obtained by final regression, wherein the raw material information represented by the set is part of replaceable materials screened by the raw material classification method based on the clustering algorithm.
(3) Evaluating and predicting based on a machine learning algorithm: and (3) combining the results of the alternative raw materials in the step (1) and the step (2) to obtain an orthogonal set, as shown in fig. 5, wherein the obtained set is a final alternative raw material list.
The replaceable raw materials are obtained by combining empirical rules and a machine learning algorithm for measurement and calculation, and the value comparison is established with professional evaluation of the replaceable raw materials, wherein the professional evaluation is from the internal professional evaluation worker for daily evaluation, and the evaluation result is obtained by recording and uploading result data through professional evaluation software. And comparing the score of the evaluation data of the replaceable raw materials with the score of the evaluation result data after actual replacement to verify the accuracy and stability of the formula balance replacement method. As in the following table:
as can be seen, the materials which are artificially and empirically replaced appear in the alternative materials screened by the auxiliary leaf group formula equilibrium replacement method shown in FIG. 4, and the preferred sequence is listed in the second, which shows that the auxiliary leaf group formula equilibrium replacement method has certain accuracy.
Example 2
The explanation is given by taking the raw material replacement of '2014 flue-cured tobacco/curved Jing 2/K326/WDC 3F/FW/P/' in the xx cigarette factory specification B, No. 1 module:
(1) rule-based raw material classification: summarizing all data of raw material historical replacement, comparing and analyzing the frequency of each attribute change in the historical replacement data, and obtaining the ranking of the importance degree of the stock information and the raw material attribute identity information to the raw material replacement according to the frequency; establishing a decision tree according to the ranking of the importance degrees, taking the first important attribute as the maximum independent variable for improving the prediction effect, splitting the node firstly, then splitting the second important attribute for the second time, and so on, and finally finding out the replaceable raw material;
transverse pair analysis: summarizing all replaced raw material data in the raw material historical replacement records of the formula, and comparing and analyzing the frequency of each attribute change in the historical replacement data, namely the frequency of the raw material attribute change after the raw material is replaced, wherein the higher the replacement frequency is, the lower the influence degree of the attribute characteristic on the raw material replacement is, and the importance is low.
And (3) analysis results: the replaceable importance degree of the inventory information and the raw material attribute identity information to the raw materials sequentially comprises the existence of inventory, whether the inventory is associated with the brand, variety information, a secondary producing area of the raw materials, the year, a tertiary producing area of the raw materials and the price, wherein whether the associated brand is data generated by the system according to the brand associated with the raw materials, and if the raw materials can be used for producing a plurality of brands, the raw materials are associated brands; the raw materials are only used for producing cigarettes of one brand and are not related. The stock is a decisive factor influencing the replacement, is consistent with the actual situation without any doubt, and the raw material replacement cannot be carried out under the condition of no stock; secondly, the influence of material grade association, if a certain replaced raw material is associated with other grades, the replacement does not occur as much as possible under the condition that other grades are considered to be preferentially used during replacement; thirdly, the variety replacement factor, certain similarity exists among the raw material attributes of certain varieties, and proper selection should be given during replacement; of course, the influence of the producing area, the year and the price cannot be ignored, the quality of the raw materials of adjacent or similar producing areas is similar, and the replacement can also be considered preferentially. As shown in fig. 6 and 7.
And (5) drawing a conclusion that: year-to-year replacement is a relatively priority factor when replacement occurs, followed by replacement of varieties and tertiary producing areas, and finally secondary producing areas for replacement of raw materials are considered when other conditions cannot be met.
Establishing a decision tree prediction model: through analysis of historical artificial replacement experience, sorting according to the influence degree of the raw material attributes on replacement, and establishing a decision tree (as shown in fig. 8): having inventory as the largest argument for improvement of the prediction effect is used first for splitting nodes; then, the second separation is carried out according to whether other brands are related or not, and then the third separation is carried out on the second-level producing area, and finally, the replaceable raw materials are found. The factors which are preferably considered during the raw material replacement are respectively inventory, brand association information and secondary production places, the screening is carried out layer by layer according to the mode, the manual replacement experience is converted into a replaceable raw material change rule, and part of replaceable materials are screened out.
As shown in fig. 9, is a human experience alternative. The replaced material was compared to the replacement material as follows:
therefore, the raw materials of the same variety, the same grade, the same year and different producing areas are selected for manual replacement. And a method for assisting leaf group formula balance replacement refers to historical replacement records by taking similar manual replacement experience as a rule to obtain partially selectable replacement materials.
(2) Classifying raw materials based on a clustering algorithm: sorting and summarizing the existing available stock raw materials, establishing a multi-dimensional space system, taking characteristic identity information (year, place of production, grade, position and the like) of the raw materials as dimensions, generating a plurality of spatial coordinate points according to the dimensions of different characteristic attributes of each raw material, wherein the closer the dimensions of the coordinate points are, the closer the characteristic attributes of the raw materials are, the higher the accuracy of replacement is; if the year is a dimension, the level is a dimension, and the position is a dimension, etc. The specific information of the characteristic attributes of the raw materials is the scale on the dimension, namely, the scale on the year dimension is 2009, 2010 and 2012 …, and the scale on the part dimension is provided with an upper part, a middle part and a lower part ….
Specifically, the dimensions in the multidimensional space system are not horizontal and vertical coordinates, and if the dimensions are not to be expressed as horizontal and vertical coordinates, any dimension in the multidimensional space system may be horizontal or vertical coordinates. That is, in the multidimensional space system, the horizontal and vertical coordinates are opposite, and as long as two dimensions are vertical, the two dimensions can be used as the x and y axes on the surface.
Points in space: selecting a raw material, wherein the specific description of the identity information of the raw material can generate a point in space according to different scales, and the point represents the raw material.
Selecting a raw material with approximately convergent multidimensional characteristics from characteristic attributes of existing raw materials as a replaceable alternative through a clustering algorithm; then, correcting clustering deviation by using a historical replacement record to obtain a part of replaceable raw materials; that is, by establishing a multidimensional space system, the position of the raw material in the space, which meets the requirement of the production inventory, is determined, and then the similarity of the raw material is judged by calculating the shortest distance between two different points in the space.
Taking coordinate points at the same latitude (same year, same part and the like) as a set, randomly extracting a numerical point K, calculating the distance from K to the next random data from a central point, and setting the central point as K; randomly extracting the next point (third point) in a local weighting mode, and calculating the central point ko from k to the new point; the sum of the distances k to all non-midpoints and L, and the sum of the distances ko to all non-midpoints and Lo are calculated in turn. If L > Lo, K = Ko is assigned; if L < Lo, K = K. That is, the center point where the distance sum is the smallest is taken as the relative center point K of the set.
Finally, the conclusion is drawn: when all the sets are calculated to obtain the central points, the distances between the central points are calculated to obtain new central points. And repeating continuously until a more concentrated set is obtained by final regression, wherein the raw material information represented by the set is part of replaceable materials screened by the raw material classification method based on the clustering algorithm.
(3) Evaluating and predicting based on a machine learning algorithm: and (3) combining the results of the replaceable raw materials in the step (1) and the step (2) to obtain an orthogonal set, wherein the obtained set is a final replaceable raw material list.
The replaceable raw materials are obtained by combining empirical rules and a machine learning algorithm for measurement and calculation, and the value comparison is established with professional evaluation of the replaceable raw materials, wherein the professional evaluation is from the internal professional evaluation worker for daily evaluation, and the evaluation result is obtained by recording and uploading result data through professional evaluation software. And comparing the score of the evaluation data of the replaceable raw materials with the score of the evaluation result data after actual replacement to verify the accuracy and stability of the formula balance replacement method. As in the following table:
as can be seen, the materials which are artificially and empirically replaced appear in the alternative materials screened by the auxiliary leaf group formula equilibrium replacement method shown in FIG. 9, and the preferred sequence is shown in the first, which shows that the auxiliary leaf group formula equilibrium replacement method has certain accuracy.
In general, the manual experience replacement is mainly performed from the producing area, the year and the like. The invention combines the manual replacement experience rule and the clustering algorithm and provides a plurality of alternative raw material optimal ordering for enterprises by a machine learning prediction analysis method. Meanwhile, the screened replaceable raw materials are verified by comparing with manual replacement experience, the top three of the sorted replaceable raw materials of the manually replaced raw materials are found, the score predicted by the auxiliary leaf group formula balance replacement method through machine learning is about 80% -90% of the actual score, and the method for assisting leaf group formula balance replacement is high in accuracy.
Claims (5)
1. A method for assisting leaf group formula balance replacement based on machine learning is characterized in that: the method comprises the following steps:
(1) rule-based raw material classification: summarizing all data of historical replacement of raw materials, and comparing and analyzing historical replacement numbers
According to the frequency of each attribute change, the sequence of the importance degree of the stock information and the material attribute identity information to the replaceability of the materials is obtained according to the frequency; establishing a decision tree according to the ranking of the importance degrees, taking the first important attribute as the maximum independent variable for improving the prediction effect, splitting the node firstly, then splitting the second important attribute for the second time, and so on, and finally finding out the replaceable raw material;
(2) classifying raw materials based on a clustering algorithm: sorting and summarizing the existing available stock raw materials, and establishing a multi-dimensional space system
Taking characteristic identity information of the materials as dimensions, and generating a plurality of spatial coordinate points according to the dimensions of different characteristic attributes of each raw material;
selecting a raw material with approximately convergent multidimensional characteristics from characteristic attributes of existing raw materials as a replaceable alternative through a clustering algorithm; then, correcting clustering deviation by using a historical replacement record to obtain a part of replaceable raw materials;
the clustering algorithm refers to a multi-dimensional space system algorithm, and the specific algorithm is as follows: substituting characteristic identity information of required raw materials by a formula to serve as discrete points with different latitudes, performing set division on the given grouping number of discrete data points under the same latitude to form an irregular stereogram, randomly extracting a numerical point k from the data set under the same latitude, calculating the central point of the distance from k to the next random numerical value, and setting the central point as k; randomly extracting the next point in a local weighting mode, and calculating k to the central point ko of the new point; sequentially calculating the distances from k to all non-central points, L and the distances from ko to all non-central points, Lo; if L is greater than Lo, assigning ko to k, after all sets are calculated, finally regressing each set to obtain a central point and storing, finally, correspondingly feeding back the characteristic identity information defined by each dimension by the central point of each dimension, traversing and searching the stock raw materials according to the calculated central value to obtain the same or most similar value, synthesizing the characteristic identity information of each dimension to give stock raw materials possibly existing or stock raw materials most similar to the characteristic factor in the calculation result, and sequencing according to the similar value;
(3) evaluating and predicting based on a machine learning algorithm: combining the results of the replaceable raw materials in the step (1) and the step (2) to obtain an orthogonal set, wherein the obtained set is a final replaceable raw material list;
substituting the obtained set into an environment with complete formula replacement requirements, comparing professional evaluation scores of the cigarette formulas corresponding to the replaced raw materials, and verifying; or, combining historical data, adopting a clustering algorithm in the step (2), and continuously and circularly verifying the obtained set inclusion algorithm.
2. The method of assisted leaf group formulation equilibrium replacement of claim 1, wherein: the step (1) is
The higher the frequency, the lower the degree of influence of this property characteristic on the raw material replacement.
3. The method of assisted leaf group formulation equilibrium replacement of claim 1, wherein: the step (1) is
The stock information indicates whether or not the stock is owned.
4. The method of assisted leaf group formulation equilibrium replacement of claim 1, wherein: the step (1) is
The material attribute identity information comprises whether the brand is related or not, variety information, a secondary production area of the material, a year, a tertiary production area of the material and a price.
5. The method of assisted leaf group formulation equilibrium replacement of claim 1, wherein: the step (2) is
The characteristic identity information of the raw materials comprises year, producing area, grade and position.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710549990.8A CN107341613B (en) | 2017-07-07 | 2017-07-07 | Method for assisting balance replacement of leaf group formula |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710549990.8A CN107341613B (en) | 2017-07-07 | 2017-07-07 | Method for assisting balance replacement of leaf group formula |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107341613A CN107341613A (en) | 2017-11-10 |
CN107341613B true CN107341613B (en) | 2021-05-25 |
Family
ID=60219610
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710549990.8A Active CN107341613B (en) | 2017-07-07 | 2017-07-07 | Method for assisting balance replacement of leaf group formula |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107341613B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110244670A (en) * | 2019-05-31 | 2019-09-17 | 山东中烟工业有限责任公司 | A kind of volume packet procedures technical parameter management-control method and system |
CN110161011A (en) * | 2019-06-12 | 2019-08-23 | 浙江中烟工业有限责任公司 | Rapid detection method of the Semblance to the essence spice for cigarette of the amounts of mixing different in production |
CN110250553A (en) * | 2019-06-25 | 2019-09-20 | 红云红河烟草(集团)有限责任公司 | Formula replacement method for maintaining stable quality of cigarette cut tobacco |
CN112397156B (en) * | 2019-07-31 | 2022-08-16 | 湖南中烟工业有限责任公司 | Digital flavoring method based on K-means clustering |
CN112395553A (en) * | 2019-08-16 | 2021-02-23 | 湖南中烟工业有限责任公司 | Digital flavoring method based on empirical formula network migration model |
CN112508470A (en) * | 2019-09-16 | 2021-03-16 | 比亚迪股份有限公司 | Method and device for changing bill of material, storage medium and electronic equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1524460A (en) * | 2003-02-25 | 2004-09-01 | 中国海洋大学 | Method for establishing mixed expert system of maintaining cigarette leaf group formulation |
CN101419454A (en) * | 2008-12-04 | 2009-04-29 | 哈尔滨工程大学 | Cigarette recipe maintenance method based on artificial immunity method |
CN103020765A (en) * | 2012-12-10 | 2013-04-03 | 红塔烟草(集团)有限责任公司 | Tobacco leaf raw material dynamic balancing method |
CN106096748A (en) * | 2016-04-28 | 2016-11-09 | 武汉宝钢华中贸易有限公司 | Entrucking forecast model in man-hour based on cluster analysis and decision Tree algorithms |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8185749B2 (en) * | 2008-09-02 | 2012-05-22 | Apple Inc. | System and method for revising boolean and arithmetic operations |
-
2017
- 2017-07-07 CN CN201710549990.8A patent/CN107341613B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1524460A (en) * | 2003-02-25 | 2004-09-01 | 中国海洋大学 | Method for establishing mixed expert system of maintaining cigarette leaf group formulation |
CN101419454A (en) * | 2008-12-04 | 2009-04-29 | 哈尔滨工程大学 | Cigarette recipe maintenance method based on artificial immunity method |
CN103020765A (en) * | 2012-12-10 | 2013-04-03 | 红塔烟草(集团)有限责任公司 | Tobacco leaf raw material dynamic balancing method |
CN106096748A (en) * | 2016-04-28 | 2016-11-09 | 武汉宝钢华中贸易有限公司 | Entrucking forecast model in man-hour based on cluster analysis and decision Tree algorithms |
Also Published As
Publication number | Publication date |
---|---|
CN107341613A (en) | 2017-11-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107341613B (en) | Method for assisting balance replacement of leaf group formula | |
CN101414183B (en) | Cigarette working procedure quality overall evaluation system and method based on gray correlation analysis | |
JP6979527B2 (en) | Chinese medicine production process knowledge system | |
CN108898479B (en) | Credit evaluation model construction method and device | |
Mpampa et al. | A new methodology for the development of sizing systems for the mass customization of garments | |
CN109242037B (en) | Tobacco leaf quality similarity measurement method | |
CN103844344A (en) | Method for regulating and controlling tobacco shred quality uniformity of different batches of cigarettes and application of method | |
CN103578037A (en) | Method and system for evaluating quality of tobacco materials | |
CN110097529B (en) | Farmland grading unit dividing method and system based on semantic rules | |
CN109447167A (en) | A kind of intelligent cigarette composition maintenance method based on Non-negative Matrix Factorization | |
CN111126865A (en) | Technology maturity judging method and system based on scientific and technological big data | |
CN107807206A (en) | A kind of piece cigarette based on industrial requirement suitably refines environment and its suitably refines the decision method of time | |
CN107897995A (en) | A kind of segmentation split based on former cigarette formula module is homogenized regulation and control method | |
CN101866368B (en) | Method for carrying out computer assisted design of tobacco group formula by near infrared spectrum technology | |
CN104572900B (en) | The properties and characteristicses system of selection that a kind of crop breeding is evaluated | |
CN111652516A (en) | Tobacco base applicability evaluation method based on formula efficacy | |
CN101419454A (en) | Cigarette recipe maintenance method based on artificial immunity method | |
CN111667176A (en) | Tobacco leaf style characteristic highlighting degree evaluation method based on cigarette brand requirements | |
CN109034486B (en) | Cotton production process optimization method based on big data hierarchical clustering | |
CN111543668B (en) | Design method of threshing and redrying formula module | |
CN110309424A (en) | A kind of socialization recommended method based on Rough clustering | |
CN113076309B (en) | System and method for predicting water addition amount of raw tobacco | |
CN114246356A (en) | Design method, system, medium and device of cigarette leaf group formula | |
CN115345487A (en) | Cigarette classification method and system | |
CN110432539B (en) | Cigarette raw material vacancy substitution method |
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 |