Characteristic selection method for tobacco threshing and redrying yield prediction model of flue-cured tobacco

By conducting correlation analysis of the physical and chemical indicators of tobacco leaves, combined with the weight fusion selection characteristics, a sheet rate prediction model was established, which solved the problem of difficult to predict the leaf-beating and re-baking sheet rate in the existing technology, and achieved more efficient and more Accurate film yield prediction and corporate decision-making support.

CN118917479AActive Publication Date: 2024-11-08CHINA TOBACCO HEBEI INDUSTRIAL CO LTD

Patent Information

Application Number
CN202410999551.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-08
Estimated Expiration
2044-07-24

AI Technical Summary

Technical Problem

The characteristics of the existing technology are difficult to select the prediction model for the leaf-beating and re-baked tobacco leaf-beating and re-baked tobacco leaf rate prediction, which can not be accurately predicted, affecting the accurate prediction and subsequent preparation of tobacco leaf raw materials by cigarette factories and re-baked enterprises. .

Method used

By detecting the physical and chemical indicators of tobacco leaves, Pearson, Kendall and Spearman correlation analysis was used, combined with weight fusion, indicators with significant characteristics were selected, and a prediction model for leaf-beating and re-baking sheet rate was established.

Benefits of technology

It reduces detection indicators and costs, improves the efficiency and accuracy of cell yield prediction, provides enterprises with faster and more accurate decision-making support, accurately adjusts procurement and warehousing planning, and improves cell yield rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917479A_ABST
    Figure CN118917479A_ABST
Patent Text Reader

Abstract

The invention provides a characteristic selection method for a tobacco threshing and redrying yield prediction model. The characteristic selection method comprises the following steps: processing multiple batches of tobacco leaves to collect characteristics of the tobacco leaves and corresponding threshing and redrying yield data; feature selection is carried out through Pearson correlation analysis, Kendall correlation analysis and Spearman correlation analysis, and corresponding prediction models are established through regression analysis; and calculating the comprehensive correlation and comprehensive significance of the characteristics of the tobacco leaves, and finally selecting the characteristics of the tobacco leaf threshing and redrying yield model of the flue-cured tobacco. According to the feature selection method, the number of selected features and the preparation work time for predicting the threshing and redrying slice yield are reduced, the workload of detection personnel is relieved, the detection cost is effectively reduced, the threshing and redrying slice yield result can be predicted more accurately, faster and more accurate decision support is provided for cigarette factories and redrying enterprises, and the tobacco leaf threshing and redrying efficiency is improved. The purchase plan can be adjusted more accurately, subsequent packaging materials, logistics transport capacity and storage sites can be prepared, and cost reduction and efficiency improvement of enterprises are facilitated.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to feature selection of flue-cured tobacco leaf threshing and redrying yield rate, and in particular to a feature selection method for improving the accuracy of a flue-cured tobacco leaf threshing and redrying yield rate prediction model, including physical and chemical indicators, correlation analysis and feature selection. Background Art

[0002] The flake yield refers to the percentage of the weight of the flakes obtained after the tobacco leaves are processed during the threshing and redrying process, which accounts for the standard weight of the raw tobacco leaves. Threshing and redrying is an indispensable and important link in the production of flue-cured tobacco. It is not only a simple separation of flakes from stems, but also a reasonable leaf structure after threshing and redrying is an important guarantee for product quality. The flake yield of threshing and redrying is also the most concerned economic and technical indicator for processing companies. The main factors affecting the flake yield of threshing and redrying are the physical properties of tobacco leaves, the performance of threshing equipment and the process parameters during the threshing process.

[0003] The physical properties of tobacco leaves, such as leaf structure, moisture content, thickness, density, and mechanical properties such as shearing force and penetration force, will have a significant impact on the yield of leaves after threshing and redrying. The chemical properties of tobacco leaves also have a certain degree of influence on the physical properties. For example, tobacco leaves with high total nitrogen and ash content have poor crushing resistance, while tobacco leaves with high reducing sugar content have better crushing resistance, which affects the yield of leaves.

[0004] The performance of key equipment such as vacuum rehumidification machines, leaf moistening machines, leaf threshing machines and redrying machines directly affects the sheet yield. For example, equipment from different manufacturers, equipment usage time and maintenance quality all have a great impact on the sheet yield. The process parameters during leaf threshing include roller speed, beating force and tobacco leaf moisture content. Moderately increasing the roller speed is conducive to improving the sheet yield, but too high a speed will cause excessive crushing of tobacco leaves and reduce the sheet yield. Tobacco leaf moisture content refers to the moisture content of tobacco leaves before threshing. Reasonable tobacco leaf moisture content is conducive to improving the sheet yield of threshing and redrying. Too high or too low a tobacco leaf moisture content is not conducive to improving the sheet yield of threshing and redrying.

[0005] For the redrying plant, the performance of the leaf threshing machine and the technological parameters during the leaf threshing process are relatively constant. Therefore, it is necessary to find the characteristics that affect the leaf threshing and redrying yield from the physical and chemical indicators of the tobacco raw materials.

[0006] The invention with application number 201911228338.1 discloses a method for detecting the accuracy of leaf threshing and redrying data, including: dividing tobacco leaves into light fragrance type, strong fragrance type and intermediate fragrance type according to the origin and flavor of tobacco leaves, and dividing tobacco leaves of each flavor type into upper leaves, middle leaves and lower leaves according to the parts; establishing a mathematical model for the flake yield of tobacco leaves of different flavors and different parts and the stem content of original tobacco to characterize the corresponding relationship between the flake yield of tobacco leaves and the stem content of original tobacco; calculating the upper and lower limits of the flake yield corresponding to tobacco leaves of different flavors and different parts according to the mathematical model, and establishing the flake yield range corresponding to tobacco leaves of different flavors and different parts; obtaining the flake yield data of the threshing and redrying tobacco leaves to be detected, judging whether the flake yield data is within the corresponding flake yield range, and if not, judging that the flake yield data is abnormal. The above invention can improve the intelligence of enterprise production management and improve the accuracy of product data analysis. However, the above invention does not select the features for the prediction model of the slice yield rate of flue-cured tobacco leaf threshing and redrying, and is unable to more accurately predict the slice yield rate of leaf threshing and redrying, which is not conducive to cigarette factories and redrying enterprises to accurately predict the slice yield rate of tobacco raw materials, and cannot accurately prepare subsequent packaging materials, logistics capacity and storage space. Summary of the invention

[0007] In order to solve the technical problem that it is difficult to select features for a prediction model of the yield of slices from leaf threshing and redrying of flue-cured tobacco, the present invention proposes a feature selection method for a prediction model of the yield of slices from leaf threshing and redrying of flue-cured tobacco. By testing the physical and chemical indicators of tobacco leaf raw materials, an innovative correlation analysis between the physical and chemical indicators of tobacco leaves and the yield of slices from leaf threshing and redrying is carried out, indicators with significant features are selected from numerous physical and chemical indicators, and a prediction model for the yield of slices from leaf threshing and redrying is established, thereby achieving accurate prediction of the yield of slices from leaf threshing and redrying of tobacco.

[0008] In order to achieve the above object, the technical solution of the present invention is achieved as follows:

[0009] A feature selection method for a flue-cured tobacco leaf threshing and redrying yield prediction model, comprising:

[0010] Step 1: Collect the characteristics of Q tobacco leaves and the corresponding threshing and redrying yield data by processing multiple batches of tobacco leaves;

[0011] Step 2: Select Q1 features that are correlated with the threshing and redrying rate from the Q tobacco leaf features through Pearson correlation analysis, and establish a Pearson prediction model through regression analysis;

[0012] Step 3: Select Q2 features that are correlated with the threshing and redrying yield rate from the Q tobacco leaf features through Kendall correlation analysis, and establish a Kendall prediction model through regression analysis;

[0013] Step 4: Select Q3 features that are correlated with the threshing and redrying rate from the Q tobacco leaf features through Spearman correlation analysis, and establish a Spearman prediction model through regression analysis;

[0014] Step 5: Calculate the error proportions of the Pearson prediction model, the Kendall prediction model, and the Spearman prediction model, respectively, and obtain the credibility of the Pearson correlation analysis, the Kendall correlation analysis, and the Spearman correlation analysis;

[0015] Step 6: Based on the credibility obtained in step 5, calculate the weights corresponding to the Pearson correlation analysis, Kendall correlation analysis, and Pierceman correlation analysis, respectively, so as to calculate the comprehensive correlation and comprehensive significance of the characteristics of Q tobacco leaves, set the significance critical value of the significance test to 0.05, and finally select Q4 characteristics for the flue-cured tobacco leaf threshing and redrying slice rate model.

[0016] The comprehensive correlation R = W1r + W2τ + W3ρ, the comprehensive significance R p =W1r p +W2τ p +W3ρ p , r p is the significance of Pearson correlation analysis, τ p is the significance of Kendall correlation analysis, ρ p is the significance of Spearman correlation analysis, W1 is the weight of Pearson correlation analysis, W2 is the weight of Kendall correlation analysis, and W3 is the weight of Spearman correlation analysis.

[0017] The implementation method of step 2 is:

[0018] Step 2.1: Calculate the Pearson correlation coefficients between the characteristics of Q tobacco leaves and the threshing and redrying yield;

[0019] Step 2.2: Perform a significance test on the Pearson correlation coefficient to obtain significance. Select Q1 features that are significantly correlated with the leaf threshing and redrying rate from the Q tobacco leaf features according to the Pearson coefficient. Establish a Pearson prediction model through regression analysis and calculate the mean square error of the Pearson prediction model.

[0020] The significance level threshold of the Pearson chi-square test was set to 0.05.

[0021] The implementation method of step 3 is:

[0022] Step 3.1: Calculate the Kendall correlation coefficients between the characteristics of Q tobacco leaves and the threshing and redrying yield;

[0023] Step 3.2: Perform a significance test on the Kendall correlation coefficient to obtain significance. Select Q2 features that are significantly correlated with the threshing and redrying rate from the Q tobacco leaf features based on the Kendall correlation coefficient. Establish a Kendall prediction model through regression analysis and calculate the mean square error of the Kendall prediction model.

[0024] The significance level threshold of the Kendall correlation coefficient chi-square test was set at 0.05.

[0025] The implementation method of step 4 is:

[0026] Step 4.1: Calculate the Spearman correlation coefficient between the characteristics of Q tobacco leaves and the threshing and redrying yield rate respectively;

[0027] Step 4.2: Perform a significance test on the Spearman correlation coefficient to obtain significance. Select Q3 tobacco leaf characteristics that are significantly correlated with the leaf threshing and redrying rate from the Q tobacco leaf characteristics based on the Spearman correlation coefficient. Establish a Spearman prediction model through regression analysis and calculate the mean square error of the Spearman prediction model.

[0028] The significance level of the Spearman correlation coefficient chi-square test was set at 0.05.

[0029] The Pearson correlation coefficient calculation method is: r k is the Pearson correlation coefficient of the kth feature, represents the kth eigenvalue of the i-th tobacco leaf sample, y i represents the threshing and re-baking rate of the i-th tobacco leaf sample, N is the number of samples, represents the average value of all tobacco leaf samples on the kth feature, It represents the mean value of the leaf threshing and re-roasting slice yield rate of all samples.

[0030] The Kendall correlation coefficient calculation method is: In the formula, τ g is the Kendall correlation coefficient of the g-th feature, e is the number of observations, is the rank of the lth observation of the gth feature, rank(y j ) is the rank of the label j-th observation, It compares all possible combinations of features and labels and calculates the sum of the absolute values ​​of the level differences.

[0031] The Spearman correlation coefficient calculation method is: is the rank difference between the rank of the sth observation of the fth feature and the label rank, e is the number of observations, and ρ fis the Kendall correlation coefficient between the fth feature and the threshing and redrying slice yield.

[0032] The error ratio = mean square error of the prediction model / total error, the total error = mean square error of the Pearson prediction model + mean square error of the Kendall prediction model + mean square error of the Spearman prediction model, and the credibility = 100% - error ratio.

[0033] The weight is calculated as follows: W u is the weight corresponding to the correlation analysis, h u is the reliability of the corresponding correlation analysis, u=1, 2, 3 correspond to Pearson correlation analysis, Kendall correlation analysis, and Spearman correlation analysis respectively.

[0034] Compared with the prior art, the present invention has the following beneficial effects: the present invention reduces the detection indicators of physical and chemical indicators, significantly reduces the workload of detection personnel, and effectively reduces the detection cost. It also greatly reduces the preparation time for predicting the leaf threshing and redrying yield rate, thereby greatly improving the efficiency of leaf threshing and redrying yield rate prediction, providing faster and more accurate decision support for cigarette factories and redrying enterprises, and can more accurately adjust procurement plans, prepare subsequent packaging materials, logistics capacity and storage sites, which is beneficial to enterprises to reduce costs and increase efficiency. The yield rate prediction model can also be used to guide tobacco cultivation and modulation, and improve the yield rate by adjusting the characteristic indicators of tobacco raw materials. The features selected by the three analysis methods are fused to form a new set of feature selection schemes, and a regression equation is established to predict the leaf threshing and redrying yield rate; compared with the single use of Pearson correlation analysis, Kendall correlation analysis or Spearman correlation analysis, the feature selection method of the present invention not only reduces the number of selected features, but also can more accurately predict the leaf threshing and redrying yield rate results. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 The present invention is a schematic diagram of the overall process of the feature selection method for the prediction model of the flue-cured tobacco leaf threshing and redrying yield rate. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] like Figure 1 As shown, a feature selection method for a flue-cured tobacco leaf threshing and redrying slice rate prediction model includes the following steps:

[0039] Step 1: Collect the characteristics of Q tobacco leaves and the corresponding threshing and redrying yield data by processing multiple batches of tobacco leaves.

[0040] More than 70 physical and chemical indicators of a batch of flue-cured tobacco produced by Hebei China Tobacco Industry Co., Ltd. were determined, including shear force, penetration force, tension, adhesion, thickness, total sugar, reducing sugar, total nitrogen, total potassium, total chlorine, nicotine, starch, sulfide ion, anti-drying and breaking property, dichloromethane extract, oxalic acid, succinic acid, malonic acid, neochlorogenic acid, scopoletin, rutin, malic acid, asparagine, glycine, phenylalanine, histidine, Glu-An (glutamine), Fru-Asn (fruit aspartic acid), Fru-Ile (fruit isoleucine), Fru-Tyr (fruit tyrosine), and neophytadiene.

[0041] Table 1 Physical and chemical characteristics

[0042]

[0043] Specifically, the physical indicators of tobacco leaves such as shear force, penetration force, tension, and adhesion are tested in accordance with the requirements of "YQ-GY / T 7.2-2023 Determination of Physical Properties of Tobacco Leaves". The national standard "YC / T 159-2019 Continuous Flow Determination" is adopted to determine the total sugar, reducing sugar, total nitrogen, total potassium, total chlorine, nicotine, and starch content in tobacco leaves. Chemical indicators such as dichloromethane extract, oxalic acid, succinic acid, malonic acid, neochlorogenic acid, scopoletin, rutin, malic acid, asparagine, glycine, phenylalanine, histidine, Glu-An, Fru-Asn, Fru-Ile, Fru-Tyr, and neophytadiene are detected by near-infrared spectroscopy. The leaf threshing and re-baking rate is expressed by the percentage of the weight of the tobacco leaves obtained after the tobacco leaves are processed to the standard weight of the raw tobacco leaves.

[0044] Step 2: Select Q1 features that are correlated with the leaf threshing and redrying rate from the Q tobacco leaf features through Pearson correlation analysis, and establish a Pearson prediction model through regression analysis.

[0045] Step 2.1: Calculate the Pearson correlation coefficients between the characteristics of the Q tobacco leaves and the leaf threshing and redrying yield.

[0046] The Pearson correlation coefficient, also known as the Pearson product-moment correlation coefficient, is a statistic that measures the degree of linear correlation between two variables and is used to evaluate the linear correlation between two continuous variables. Its value range is between -1 and 1. The closer the value is to 1 or -1, the stronger the linear relationship between the two variables; the closer the value is to 0, the weaker the linear relationship between the two variables.

[0047] Features: Measures the strength and direction of the linear relationship between two variables.

[0048] Advantages: When there is a linear relationship between variables, the Pearson correlation coefficient can effectively capture the strength of this relationship. It is the most commonly used correlation measurement tool and the results are easy to understand.

[0049] Applicable scenarios: Applicable to continuous data and variables that conform to the normal distribution.

[0050] The calculation formula of Pearson correlation coefficient is as follows:

[0051]

[0052] r k is the Pearson correlation coefficient of the kth feature, represents the kth eigenvalue of the i-th tobacco leaf sample, y i It represents the rate of threshing and redrying of tobacco leaves of the i-th tobacco leaf sample, and N is the number of samples. represents the average value of all tobacco leaf samples on the kth feature, It represents the mean value of the leaf threshing and re-roasting slice yield rate of all samples.

[0053] Step 2.2: Perform a significance test on the Pearson correlation coefficient to obtain significance. According to the Pearson coefficient, select Q1 features that are significantly correlated with the leaf threshing and redrying rate from the Q tobacco leaf features. Establish a Pearson prediction model through regression analysis and calculate the mean square error of the Pearson prediction model.

[0054] In order to determine the statistical significance of the Pearson correlation coefficient, the purpose of the significance test is to determine whether there is a significant difference between tobacco leaf characteristics and labels (threshing and redrying slice rate), so as to determine the correlation between tobacco leaf characteristics and threshing and redrying slice rate. The significance level threshold of the Pearson chi-square test is set to 0.05.

[0055] The significance P value of the Pearson correlation coefficient is usually calculated through a hypothesis test based on the t distribution of the Pearson coefficient. The formula for calculating the t statistic is:

[0056]

[0057] t k is the t statistic of the kth feature, r k is the Pearson correlation coefficient of the kth feature, and n is the number of trials. Determine the degrees of freedom (df) of the t statistic. For the Pearson correlation coefficient, the degrees of freedom are n-2. Use a t distribution table or computer software (such as R, Python, SPSS, etc.) to determine the P value corresponding to the t statistic, that is, significance.

[0058] The selected features are F1…F Q1 , conduct production experiments on multiple batches of tobacco raw materials, perform regression analysis based on the selected Q1 features (such as linear regression model, neural network, etc., determined by the number of samples), establish a Pearson prediction model, and calculate the mean square error (MSE) of the Pearson prediction model r To evaluate the prediction accuracy of the Pearson prediction model.

[0059] Step 3: Select Q2 features that are correlated with the leaf threshing and redrying rate from the Q tobacco leaf features through Kendall correlation analysis, and establish a Kendall prediction model through regression analysis.

[0060] Step 3.1: Calculate the Kendall correlation coefficients between the characteristics of the Q tobacco leaves and the threshing and redrying yield.

[0061] Kendall correlation analysis is a statistical method used to quantify the correlation of the rank or order relationship between two variables. It is a type of rank correlation coefficient, usually represented by the Greek letter τ. The Kendall correlation coefficient is mainly applied to ordered categorical variables. It can measure the similarity between the rankings of two variables. The value of the Kendall correlation coefficient τ is between -1 and 1. When τ = 1, it means that the two variables are completely consistent; when τ = -1, it means that the two variables are completely opposite; when τ = 0, it means that there is no correlation between the two variables.

[0062] Features: A non-parametric method based on rank to measure the correlation between two variables.

[0063] Advantages: It is highly robust to outliers, does not require data to conform to a specific distribution, and is suitable for processing small sample data.

[0064] Applicable scenarios: Applicable to situations where the data does not conform to the normal distribution, the sample size is small, or there are outliers.

[0065] The calculation formula of Kendall correlation coefficient is as follows:

[0066]

[0067] In the formula, τ g is the Kendall correlation coefficient of the g-th feature, e is the number of observations (the number of observations is the measured features of Q tobacco leaves), is the rank of the lth observation of the gth feature; rank(y j ) is the rank of the jth observation value of the label (threshing and re-roasting slice yield). (The rank of the observation value is 1 for the first observation value, 2 for the second observation value, and so on). It is to compare all possible combinations of features and labels (leaf threshing and re-roasting slice yield) and calculate the sum of the absolute values ​​of the grade differences.

[0068] Step 3.2: Perform a significance test on the Kendall correlation coefficient to obtain significance. According to the Kendall correlation coefficient, select Q2 features that are significantly correlated with the leaf threshing and redrying rate from the Q tobacco leaf characteristics. Establish the Kendall prediction model through regression analysis and calculate the mean square error of the Kendall prediction model.

[0069] In order to determine the statistical significance of the Kendall correlation coefficient, the purpose of the significance test is to determine whether there is a significant difference between tobacco leaf characteristics and labels (threshing and redrying slice rate), so as to determine the correlation between tobacco leaf characteristics and threshing and redrying slice rate. The significance level threshold of the Kendall correlation coefficient chi-square test is set to 0.05.

[0070] The significance test of Kendall's correlation coefficient is usually performed by calculating its corresponding t statistic, and the formula is as follows:

[0071]

[0072] t g is the t statistic of the g-th feature, τ g is the Kendall correlation coefficient for the gth feature, and n is the number of trials. After calculating the t value, determine the degrees of freedom (df) of the t statistic. For the Kendall correlation coefficient, the degrees of freedom are n-2. Use a t distribution table or computer software (such as R, Python, SPSS, etc.) to determine the P value corresponding to the t statistic, that is, significance.

[0073] The selected features are F1…F Q2 Through the production experiments of multiple batches of tobacco raw materials, the Kendall prediction model was constructed through linear regression analysis based on the obtained data, and the mean square error (MSE) of the Kendall prediction model was calculated. τ To evaluate the prediction accuracy of the Kendall prediction model.

[0074] Step 4: Select Q3 features that are correlated with the leaf threshing and redrying rate from the Q tobacco leaf features through Spearman correlation analysis, and establish a Spearman prediction model through regression analysis.

[0075] Step 4.1: Calculate the Spearman correlation coefficient between the characteristics of Q tobacco leaves and the leaf threshing and redrying yield.

[0076] Spearman's rank correlation coefficient, also known as Spearman's rank correlation coefficient, is a nonparametric statistical indicator that measures the dependence of two variables. It is suitable for evaluating the correlation between the rank or ordinal data of two variables. The Spearman correlation coefficient can also be used to evaluate the correlation between two continuous variables, but they do not need to be normally distributed. When ρ is close to 1, it indicates that there is a strong positive correlation between the two variables. When ρ is close to -1, it indicates that there is a strong negative correlation between the two variables. When ρ is close to 0, it means that there is no or only a very weak correlation between the two variables.

[0077] Features: It is also a rank-based nonparametric method, but compared with the Kendall correlation coefficient, it pays more attention to the order and hierarchical structure of the data.

[0078] Advantages: It also has good resistance to outliers, does not require data distribution, and is suitable for processing ordered categorical data and data with nonlinear relationships.

[0079] Applicable scenarios: Applicable when the relationship between variables may be nonlinear or the variables are ordered categorical data.

[0080] The calculation formula of Kendall correlation coefficient is as follows:

[0081]

[0082] is the difference between the rank of the sth observation of the fth feature and the rank of the label (threshing and re-roasting slice yield), e is the number of observations, and ρ f is the Kendall correlation coefficient between the fth feature and the threshing and redrying slice yield.

[0083] Step 4.2: Perform a significance test on the Spearman correlation coefficient to obtain significance. Select Q3 tobacco leaf characteristics that are significantly correlated with the leaf threshing and re-roasting rate from the Q tobacco leaf characteristics based on the Spearman correlation coefficient. Establish a Spearman prediction model through regression analysis and calculate the mean square error of the Spearman prediction model.

[0084] The significance test of the Spearman correlation coefficient is to determine whether there is a significant difference between tobacco leaf characteristics and labels (threshing and redrying slice rate), so as to determine the correlation between tobacco leaf characteristics and threshing and redrying slice rate. The significance level of the Spearman correlation coefficient chi-square test is set to 0.05.

[0085] Convert ρ to t statistic. The calculation formula of t statistic is:

[0086]

[0087] t f is the t statistic of the fth feature, ρ f is the Spearman correlation coefficient of the f-th feature, and n is the number of trials.

[0088] Determine the degrees of freedom (df) of the t statistic. For the Spearman correlation coefficient, the degrees of freedom are n - 2. Use a t distribution table or computer software (such as R, Python, SPSS, etc.) to determine the P value corresponding to the t statistic, that is, the significance.

[0089] The selected features are F1…F Q3 By conducting production experiments on multiple batches of tobacco raw materials, and building a Spearman prediction model based on the obtained data through linear regression analysis, the mean square error (MSE) of the Spearman prediction model was calculated. ρ To evaluate the prediction accuracy of the Spearman prediction model.

[0090] Step 5: Calculate the error proportions of the Pearson prediction model, the Kendall prediction model, and the Spearman prediction model, respectively, and obtain the credibility of the Pearson correlation analysis, the Kendall correlation analysis, and the Spearman correlation analysis, respectively.

[0091] According to the mean square error MSE of the Pearson prediction model r , mean square error MSE of Kendall prediction model τ , mean square error MSE of the Spearman prediction model ρ , the total error is:

[0092] Total error = MSE r +MSE τ +MSE ρ (7)

[0093] The error percentage of the Pearson prediction model is:

[0094]

[0095] The error ratio of the Kendall prediction model is:

[0096]

[0097] The error percentage of the Spearman prediction model is:

[0098]

[0099] According to the error percentage of the Pearson prediction model, Kendall prediction model and Spearman prediction model, credibility = 100% - error percentage, it can be concluded that the credibility of Pearson correlation analysis is (1-a), the credibility of Kendall correlation analysis is (1-b), and the credibility of Spearman correlation analysis is (1-c).

[0100] Step 6: Based on the credibility obtained in step 5, calculate the weights corresponding to the Pearson correlation analysis, Kendall correlation analysis, and Pierceman correlation analysis, respectively, so as to calculate the comprehensive correlation and comprehensive significance of the characteristics of Q tobacco leaves, set the significance critical value of the significance test to 0.05, and finally select Q4 characteristics for the flue-cured tobacco leaf threshing and redrying slice rate model.

[0101] From step 2, step 3 and step 4, it is found that the features selected by the three correlation analysis methods are inconsistent and the number of features is large. Pearson correlation analysis selects Q1 features, Kendall correlation analysis selects Q2 features, and Spearman correlation analysis selects Q3 features.

[0102] Compared with the single use of Pearson correlation analysis, Kendall correlation analysis or Spearman correlation analysis, all three correlation analyses have certain limitations. In statistical analysis, Pearson correlation analysis requires the data to be normally distributed, while Spearman correlation analysis is applicable to non-normally distributed data. Pearson correlation analysis may not reveal the significant relationship between them. As a non-parametric test method, Spearman correlation analysis does not require the distribution characteristics of the data, so in this case it can show the significant correlation between the two variables and the leaf threshing and re-roasting rate. According to practical experience, although the Pearson correlation analysis did not show a significant relationship between a certain feature and the leaf threshing and re-roasting rate, in fact, the role of these chemical components in the production process may be very critical. Therefore, in practical applications, even if the statistical data is not fully supported, it is still necessary to pay attention to the potential impact of these factors and give them appropriate consideration in process optimization and production regulation.

[0103] In order to solve this problem, the present invention comprehensively considers Pearson correlation analysis, Kendall correlation analysis and Spearman correlation analysis, and deeply integrates the correlation between each feature and the label (leaf threshing and re-roasting slice rate), and adopts a weight-based fusion strategy and feature selection strategy: the weight of the corresponding correlation method is calculated according to the credibility ratio, and the corresponding correlation is weighted. The three different correlation coefficients and significance are recalculated by weight to obtain the comprehensive correlation and comprehensive significance. Although they may not be significant in Pearson correlation analysis, Kendall correlation analysis or Spearman correlation analysis, they show significance in other types of correlation analysis. This may be because different correlation analysis methods have different requirements for the distribution, type and relationship assumptions of data, and therefore reveal different association patterns.

[0104] Weight-based fusion strategy and feature selection strategy: In order to fuse the features, the weight of the corresponding correlation method is calculated according to the credibility ratio, and the corresponding correlation method is weighted. The calculation formula of the weight W is shown in formula (10):

[0105]

[0106] W u is the weight corresponding to the correlation analysis, h u is the reliability of the corresponding correlation analysis, u=1, 2, 3 correspond to Pearson correlation analysis, Kendall correlation analysis, and Spearman correlation analysis respectively.

[0107] The weights of Pearson correlation analysis, Kendall correlation analysis and Spearman correlation analysis are calculated by formula (10) as W1, W2 and W3 respectively. The comprehensive correlation R and comprehensive significance R of the characteristics of Q tobacco leaves are calculated by formula (12) and formula (13) respectively. p .

[0108] R=W1r+ W2τ+W3ρ (12)

[0109] R p =W1r p + W2τ p +W3ρ p (13)

[0110] In the formula, r p is the significance of Pearson correlation analysis, τ p is the significance of Kendall correlation analysis, ρ p is the significance of Spearman correlation analysis.

[0111] The significance level threshold of the significance test of the present invention is set to 0.05. p≤0.05 is the threshold value to select Q4 features for the flue-cured tobacco leaf threshing and redrying rate model. The selected features are F1…F Q4 , a comprehensive prediction model was established through linear regression analysis.

[0112] The following is a specific embodiment of the feature selection method for the flue-cured tobacco leaf threshing and redrying yield prediction model of the present invention.

[0113] Hebei China Tobacco Industry Co., Ltd. purchased a batch of tobacco raw materials in Liancheng, Fujian, and then sent these raw materials to the Fujian Longyan Golden Leaf Redrying Factory for deep processing. During the processing, the tobacco leaves were comprehensively tested for physical and chemical indicators, and relevant data such as the rate of leaf threshing and redrying were recorded. The Pearson correlation coefficient and significance, Kendall correlation coefficient and significance, and Spearman correlation coefficient and significance of each characteristic index and the rate of leaf threshing and redrying were calculated; 21 features related to the rate of leaf threshing and redrying were selected through Pearson correlation analysis, 25 features related to the rate of leaf threshing and redrying were selected through Kendall correlation analysis, and 26 features related to the rate of leaf threshing and redrying were selected through Spearman correlation analysis. Prediction models of three correlation analysis methods were established, and the mean square error of each prediction model was calculated. According to the weight fusion strategy and feature selection strategy of the present invention, the weight of the corresponding correlation method was calculated according to the credibility ratio, and the corresponding correlation method was weighted to calculate the correlation coefficient and significance after fusion.

[0114] Specifically, 21 features were selected that were correlated with the rate of leaf threshing and redrying, as shown in Table 2.

[0115] Table 2 Pearson correlation coefficients between physical and chemical indices and leaf threshing and redrying yield

[0116]

[0117] Through the Pearson correlation coefficient and its significance, it can be seen that the shear force, thickness, brittleness, nicotine, total nitrogen, dichloromethane extract, solanesol, myristic acid, linoleic acid, asparagine, glutamic acid, methionine, phenylalanine, arginine, Glu-An, and Fru-Leu, sixteen indicators are positively correlated with the leaf threshing and redrying rate, that is, the larger the values ​​of these indicators, the higher the leaf threshing and redrying rate, among which the shear force and thickness have the largest positive correlation with the leaf threshing and redrying rate and the highest significance; the five indicators, potassium, magnesium, succinic acid, Fru-Ala, and neophytadiene, are negatively correlated with the leaf threshing and redrying rate, that is, the larger the values ​​of these indicators, the lower the leaf threshing and redrying rate, among which succinic acid has the largest negative correlation with the leaf threshing and redrying rate and the highest significance.

[0118] The Pearson prediction model of the features selected by Pearson correlation coefficient and significance and the leaf threshing and redrying yield was established:

[0119] Y p =10.30+1.66×10 -2 X1+4.27X2+…+1.34×10 -2 X 21 .

[0120] Specifically, 25 features correlated with the leaf threshing and redrying yield were selected through Kendall correlation analysis, as shown in Table 3.

[0121] Table 3 Kendall correlation coefficients between physical and chemical parameters and threshing and redrying yield

[0122]

[0123] The Kendall correlation coefficient and its significance show that twenty indicators, including shear force, tension, thickness, brittleness, myristic acid, oleic acid + linolenic acid, asparagine, glutamic acid, glutamine, glycine, alanine, lysine, tryptophan, arginine, Glu-An, Fru-Val, Fru-Gly, Fru-Asp eicosanoic acid, and scopoletin, are positively correlated with the threshing and redrying rate of slices, that is, the larger the values ​​of these indicators, the higher the threshing and redrying rate of slices, among which tension has the largest positive correlation with the threshing and redrying rate and the highest significance; five indicators, including potassium, succinic acid, cystine, Fru-Ala, and neophytadiene, are negatively correlated with the threshing and redrying rate of slices, that is, the larger the values ​​of these indicators, the lower the threshing and redrying rate of slices, among which potassium has the largest negative correlation with the threshing and redrying rate and the highest significance.

[0124] The Kendall prediction model of the leaf threshing and redrying yield was established based on the Kendall correlation coefficient and the significant selected features:

[0125] Y k =3.78+4.12×10 -1 X1+9.28×10 -2 X2+…+2.16×10 -2 X 24 .

[0126] Specifically, 26 features were selected that were correlated with the leaf threshing and redrying yield rate through Spearman correlation analysis, as shown in Table 4.

[0127] Table 4 Spearman correlation coefficients between physical and chemical indices and threshing and redrying yield

[0128]

[0129] The Spearman correlation coefficient and its significance show that twenty indicators, including shear force, tension, thickness, penetration, adhesion, brittleness, nicotine, solanesol, myristic acid, oleic acid + linolenic acid, asparagine, glutamic acid, glycine, alanine, methionine, phenylalanine, 4-aminobutyric acid, arginine, Glu-An, and Fru-His, are positively correlated with the rate of leaf threshing and redrying. That is, the larger the values ​​of these indicators, the higher the rate of leaf threshing and redrying. Among them, tension, asparagine, and arginine have the largest positive correlation with the rate of leaf threshing and redrying, and the highest significance; five indicators, including potassium, succinic acid, octadecanoic acid, tryptophan, Fru-Ala, and neophytadiene, are negatively correlated with the rate of leaf threshing and redrying. That is, the larger the values ​​of these indicators, the lower the rate of leaf threshing and redrying. Among them, potassium has the largest negative correlation with the rate of leaf threshing and redrying, and the highest significance.

[0130] The Spearman prediction model of the features selected by Spearman correlation coefficient and significance and the threshing and redrying yield was established:

[0131] Y (斯皮尔曼) =25.62+4.12×10 -1 X1+2.62×10 -2 X2+…+1.38×10 -1 X 26 .

[0132] The three prediction models of Pearson prediction model, Kendall prediction model and Spearman prediction model were verified in production, and data of multiple batches of raw cigarettes were collected. The three prediction models were deployed in the production environment to ensure that the input and output of the model matched the production data interface. For the same batch of raw materials, the three prediction models were used to predict respectively, and the prediction results were recorded. The differences between the prediction results of the three prediction models and the actual production results were compared to evaluate the accuracy and reliability of each prediction model. The weight calculation was performed according to the method of step 6, and the results are shown in Table 5.

[0133] Table 5 Errors of each model

[0134] Pearson Kendall Spearman Select quantity 21 24 26 Mean square error 1.09 2.48 0.97 Weight 0.24 0.55 0.21

[0135] According to step 6, the weights corresponding to each feature index were calculated, and the correlation and significance of each feature were recalculated according to the weights. Feature fusion was performed, and the critical value of the significance level of the significance test was set to 0.05 to select new features. Finally, 13 features for the flue-cured tobacco leaf threshing and redrying slice rate model were selected, as shown in Table 6. The feature correlation coefficient was iteratively calculated based on the weights, and a comprehensive prediction model was established for production verification.

[0136] Table 6 Correlation coefficients between physical and chemical indices and leaf threshing and redrying yield based on weight calculation

[0137] feature thickness pull Shear force brittleness Potassium Succinic acid Calculate new correlation coefficient based on weights 0.786 0.851 0.763 0.730 -0.756 -0.725 Significance 0.017 0.005 0.025 0.029 0.021 0.036 feature Myristoyl Asparagine Glutamate Arginine Glu-An Fru-His Calculate new correlation coefficient based on weights 0.706 0.751 0.785 0.805 0.754 0.701 Significance 0.043 0.029 0.015 0.012 0.022 0.040 feature Fru-Ala Calculate new correlation coefficient based on weights -0.688 Significance 0.047

[0138] Through the weighted iterative calculation of the characteristic correlation coefficient and its significance, it can be seen that the ten indicators of shear force, tension, thickness, brittleness, myristic acid, asparagine, glutamic acid, arginine, Glu-An, and Fru-His are positively correlated with the leaf threshing and redrying rate, that is, the larger the values ​​of these indicators, the higher the leaf threshing and redrying rate, among which the tension has the largest positive correlation with the leaf threshing and redrying rate and the highest significance; potassium, succinic acid, and Fru-Ala are negatively correlated with the leaf threshing and redrying rate, that is, the larger the values ​​of these indicators, the lower the leaf threshing and redrying rate, among which potassium has the largest negative correlation with the leaf threshing and redrying rate and the highest significance.

[0139] Establishment of a comprehensive prediction model based on weighted fusion of selected features and threshing and redrying yield

[0140] Y (发明) =-165.54+4.12×10 -1 X1+5.34×10 -3 X2…5.14×10 -2 X 13 .

[0141] Comparison of selected features: Pearson correlation analysis selected 21 features in total, Kendall correlation analysis selected 24 features in total, Spearman correlation analysis selected 26 features in total, and the feature selection method of the present invention selected 13 features in total. The feature selection used in the present invention reduced the number of features selected by 8 compared with Pearson correlation analysis, reduced the number of features selected by Kendall correlation analysis by 11, and reduced the number of features selected by Spearman correlation analysis by 13.

[0142] Error comparison of model establishment by selecting features: Production verification is performed on the four prediction models of Pearson prediction model, Kendall prediction model, Spearman prediction model and the comprehensive prediction model obtained by the present invention, and data of multiple batches of raw cigarettes are collected. The four prediction models are deployed in the production environment to ensure that the input and output of the model match the production data interface. For the same batch of raw materials, four prediction models are used to make predictions respectively, and the prediction results are recorded. The differences between the prediction results of the four prediction models and the actual production results are compared to evaluate the accuracy and reliability of each prediction model. The mean square error of the Pearson prediction model established by selecting features through Pearson correlation analysis is 1.13; the mean square error of the Kendall prediction model established by selecting features through Kendall correlation analysis is 2.31; the mean square error of the Spearman prediction model established by selecting features through Spearman correlation analysis is 1.01; the average error of the comprehensive prediction model established by selecting features using the feature selection method of the present invention is 0.31. The comprehensive prediction model established by the feature selection used in the present invention has an error reduction of 0.88 compared with the Pearson prediction model established by selecting features through Pearson correlation analysis, an error reduction of 2.00 compared with the Kendall prediction model established by selecting features through Kendall correlation analysis, and an error reduction of 0.70 compared with the Spearman prediction model established by selecting features through Spearman correlation analysis.

[0143] Table 7 Comparison of feature selection between Pearson, Kendall, Spearman and the present invention

[0144] Pearson Kendall Spearman The present invention Select quantity 21 24 26 13 Model Error 1.29 2.31 1.01 0.31

[0145] In summary, it can be seen that the feature selection method proposed in the present invention has the following advantages:

[0146] Comprehensiveness: By combining the Pearson correlation coefficient, Kendall correlation coefficient, and Spearman correlation coefficient, the linear relationship, rank correlation, and sensitivity to outliers of the data can be fully considered. This allows reliable feature correlation assessment to be obtained under different types of data relationships and distribution conditions.

[0147] Robustness: The Kendall and Spearman correlation coefficients provide additional stability in the face of outliers or non-normally distributed data, which can help improve the quality of feature selection and the ultimate performance of the model.

[0148] Flexibility: Applicable to a variety of data types and structures, whether linear or nonlinear, continuous or categorical.

[0149] The present invention reduces the detection indexes of physical and chemical indicators, significantly reduces the workload of detection personnel, and effectively reduces the detection cost. It also greatly reduces the preparation time for predicting the yield rate of leaf threshing and redrying, thereby greatly improving the efficiency of leaf threshing and redrying yield rate prediction, and provides faster and more accurate decision support for cigarette factories and redrying enterprises.

[0150] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A feature selection method for a prediction model of tobacco leaf threshing and redrying yield, characterized in that: include: Step 1: Collect the characteristics of Q tobacco leaves and the corresponding threshing and redrying yield data by processing multiple batches of tobacco leaves; Step 2: Select Q1 features that are correlated with the threshing and redrying rate from the Q tobacco leaf features through Pearson correlation analysis, and establish a Pearson prediction model through regression analysis; Step 3: Select Q2 features that are correlated with the threshing and redrying yield rate from the Q tobacco leaf features through Kendall correlation analysis, and establish a Kendall prediction model through regression analysis; Step 4: Select Q3 features that are correlated with the threshing and redrying rate from the Q tobacco leaf features through Spearman correlation analysis, and establish a Spearman prediction model through regression analysis; Step 5: Calculate the error proportions of the Pearson prediction model, the Kendall prediction model, and the Spearman prediction model, respectively, and obtain the credibility of the Pearson correlation analysis, the Kendall correlation analysis, and the Spearman correlation analysis; Step 6: Based on the credibility obtained in step 5, calculate the weights corresponding to the Pearson correlation analysis, Kendall correlation analysis, and Pierceman correlation analysis, respectively, so as to calculate the comprehensive correlation and comprehensive significance of the characteristics of Q tobacco leaves, set the significance critical value of the significance test to 0.05, and finally select Q4 characteristics for the flue-cured tobacco leaf threshing and redrying slice rate model.

2. The feature selection method for the prediction model of flue-cured tobacco leaf threshing and redrying yield according to claim 1, characterized in that: The comprehensive correlation R = W1r + W2τ + W3ρ, the comprehensive significance R p =W1r p +W2τ p +W3ρ p , r p is the significance of Pearson correlation analysis, τ p is the significance of Kendall correlation analysis, ρ p is the significance of Spearman correlation analysis, W1 is the weight of Pearson correlation analysis, W2 is the weight of Kendall correlation analysis, and W3 is the weight of Spearman correlation analysis.

3. The feature selection method for the prediction model of flue-cured tobacco leaf threshing and redrying yield according to claim 1, characterized in that: The implementation method of step 2 is: Step 2.1: Calculate the Pearson correlation coefficients between the characteristics of Q tobacco leaves and the threshing and redrying yield; Step 2.2: Perform a significance test on the Pearson correlation coefficient to obtain significance. Select Q1 features that are significantly correlated with the leaf threshing and redrying rate from the Q tobacco leaf features according to the Pearson coefficient. Establish a Pearson prediction model through regression analysis and calculate the mean square error of the Pearson prediction model. The significance level threshold of the Pearson chi-square test was set to 0.

05.

4. The feature selection method for the prediction model of flue-cured tobacco leaf threshing and redrying yield according to claim 3, characterized in that: The implementation method of step 3 is: Step 3.1: Calculate the Kendall correlation coefficients between the characteristics of Q tobacco leaves and the threshing and redrying yield; Step 3.2: Perform a significance test on the Kendall correlation coefficient to obtain significance. Select Q2 features that are significantly correlated with the threshing and redrying rate from the Q tobacco leaf features based on the Kendall correlation coefficient. Establish a Kendall prediction model through regression analysis and calculate the mean square error of the Kendall prediction model. The significance level threshold of the Kendall correlation coefficient chi-square test was set at 0.

05.

5. The feature selection method for the prediction model of flue-cured tobacco leaf threshing and redrying yield according to claim 4, characterized in that: The implementation method of step 4 is: Step 4.1: Calculate the Spearman correlation coefficient between the characteristics of Q tobacco leaves and the threshing and redrying yield rate respectively; Step 4.2: Perform a significance test on the Spearman correlation coefficient to obtain significance. Select Q3 tobacco leaf characteristics that are significantly correlated with the leaf threshing and redrying rate from the Q tobacco leaf characteristics based on the Spearman correlation coefficient. Establish a Spearman prediction model through regression analysis and calculate the mean square error of the Spearman prediction model. The significance level of the Spearman correlation coefficient chi-square test was set at 0.

05.

6. The feature selection method for the prediction model of flue-cured tobacco leaf threshing and redrying yield according to claim 5, characterized in that: The Pearson correlation coefficient calculation method is: r k is the Pearson correlation coefficient of the kth feature, represents the kth eigenvalue of the i-th tobacco leaf sample, y i represents the threshing and re-baking rate of the i-th tobacco leaf sample, N is the number of samples, represents the average value of all tobacco leaf samples on the kth feature, It represents the mean value of the leaf threshing and re-roasting slice yield rate of all samples.

7. The feature selection method for the prediction model of flue-cured tobacco leaf threshing and redrying yield according to claim 6, characterized in that: The Kendall correlation coefficient calculation method is: In the formula, τ g is the Kendall correlation coefficient of the g-th feature, e is the number of observations, is the rank of the lth observation of the gth feature, rank(y j ) is the rank of the label j-th observation, It compares all possible combinations of features and labels and calculates the sum of the absolute values ​​of the level differences.

8. The feature selection method for the prediction model of flue-cured tobacco leaf threshing and redrying yield according to claim 7, characterized in that: The Spearman correlation coefficient calculation method is: is the rank difference between the rank of the sth observation of the fth feature and the label rank, e is the number of observations, and ρ f is the Kendall correlation coefficient between the fth feature and the threshing and redrying slice yield.

9. The feature selection method for the prediction model of flue-cured tobacco leaf threshing and redrying yield according to claim 8, characterized in that: The error ratio = mean square error of the prediction model / total error, the total error = mean square error of the Pearson prediction model + mean square error of the Kendall prediction model + mean square error of the Spearman prediction model, and the credibility = 100% - error ratio.

10. The feature selection method for the prediction model of flue-cured tobacco leaf threshing and redrying yield according to claim 2 or 9, characterized in that: The weight is calculated as follows: W u is the weight corresponding to the correlation analysis, h u is the reliability of the corresponding correlation analysis, u=1, 2, 3 correspond to Pearson correlation analysis, Kendall correlation analysis, and Spearman correlation analysis respectively.

Citation Information

Patent Citations

  • Threshing and redrying processing quality comprehensive evaluation method based on multivariate statistics

    CN110796350A

  • Method for detecting threshing and redrying data accuracy

    CN110866221A

  • Establishment method of threshing process shrinkage rate control model

    CN111165858A

  • Threshing and redrying process parameter correlation analysis method

    CN116882629A

  • Method and system for predicting moisture removal amount of dried tobacco shreds

    CN117494880A

Cited By

  • Drainage basin water environment quality dynamic evaluation method and system based on space-time diagram neural network

    CN121563264A