Method for evaluating comprehensive quality of fresh highland barley noodles based on SVM (Support Vector Machine) model

Abstract

The invention discloses a highland barley fresh noodle comprehensive quality evaluation method based on an SVM (Support Vector Machine) model, which comprises the following steps: taking different varieties of highland barley raw materials, and measuring the quality of the different varieties of highland barley raw materials to obtain raw material quality index data; different highland barley raw materials are prepared into highland barley fresh noodles, the cooking quality and the eating quality of the corresponding highland barley fresh noodles are evaluated, finished product quality index data are obtained, and a comprehensive evaluation model is established through principal component analysis; raw material quality index data and comprehensive evaluation are simplified by utilizing correlation analysis, and key index data are selected by utilizing a grey relational degree method; performing support vector machine model training by using the key index data to obtain a comprehensive quality evaluation model of the fresh highland barley noodles; and evaluating the quality of the fresh highland barley noodles by using the comprehensive quality evaluation model of the fresh highland barley noodles. According to the method, the comprehensive quality evaluation model of the fresh highland barley noodles based on the SVM model is established, a traditional analysis method is replaced, and the efficiency is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of food processing and relates to a method for comprehensively evaluating the quality of fresh highland barley noodles based on a SVM model. Background Art

[0002] The development of the highland barley industry faces multiple constraints: First, insufficient development of variety resources, a lack of specialized processing varieties, significant differences in physical and chemical properties among existing varieties, and low efficiency in traditional breeding, making it difficult to meet processing needs. Second, significant bottlenecks in processing technology exist. Low gluten protein content results in poor noodle extensibility, requiring reliance on wheat flour compounding or improvers. Furthermore, a lack of standardized processing techniques makes it difficult to retain functional ingredients. Third, weak industry chain coordination and a disconnect between cultivation and processing. Due to the lack of information on highland barley quality resources, the correlation between the quality of highland barley raw materials and the quality of processed products remains unclear, resulting in unstable quality of fresh highland barley noodles and a cumbersome and complex process for determining their overall quality. Summary of the Invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a comprehensive quality evaluation method for fresh highland barley noodles based on the SVM model to solve the problems of unstable quality of fresh highland barley noodles and complex measurement procedures.

[0004] Technical solution: A method for comprehensively evaluating the quality of fresh highland barley noodles based on a SVM model of the present invention comprises the following steps:

[0005] S1. Taking different varieties of highland barley raw materials, measuring the quality of the different varieties of highland barley raw materials to obtain raw material quality index data;

[0006] S2. Prepare fresh highland barley noodles from different highland barley raw materials, and evaluate the cooking quality and edible quality of the corresponding fresh highland barley noodles to obtain finished product quality index data, and establish a comprehensive evaluation model using principal component analysis;

[0007] S3, using correlation analysis to simplify the highland barley index of step S1 and the comprehensive evaluation obtained in S2, and then using the grey correlation method to select key indicator data;

[0008] S4, using the key indicator data of step S3 to perform support vector machine model training to obtain a comprehensive quality evaluation model for fresh highland barley noodles;

[0009] S5. Use the comprehensive quality evaluation model of fresh highland barley noodles to evaluate the quality of fresh highland barley noodles.

[0010] Furthermore, step S1 specifically includes the following steps: S11, preparing whole highland barley flour; S12, determining the nutritional components of highland barley, including moisture content determination, ash content determination, fat content determination, protein content determination, amino acid content determination, total starch, amylose content determination, and β-glucan content determination; S13, determining the protein component of highland barley; S14, determining the gelatinization properties of the whole highland barley flour.

[0011] Furthermore, the whole highland barley flour prepared in step S11 is passed through a 40-mesh sieve.

[0012] Furthermore, the cooking quality of the fresh highland barley noodles in step S2 includes noodle cooking characteristics, noodle texture characteristics, and noodle stretching characteristics.

[0013] Furthermore, the specific method for evaluating the edible quality is: using a sensory evaluation method to evaluate the color, appearance, palatability, elasticity, smoothness and taste of the fresh highland barley noodles.

[0014] Furthermore, in step S4, the support vector machine model adopts a radial basis kernel function, a cost parameter of 1, a kernel function parameter Gramma of 0.1, and a hyperplane formula as follows:

[0015] f(x)=ω T x+b,

[0016] Where: ω is the transpose of the weight vector, b is the bias;

[0017] The formula for finding the optimal hyperplane is as follows:

[0018]

[0019] In the formula, the value of the penalty factor c determines the weights of the two variables in the formula. ξi and ξi are slack variables used to relax the constraints on the optimal hyperplane. Sample data within the insensitive band defined by ε are not included in the loss. Only samples outside the insensitive band will affect the support vector regression model.

[0020] Beneficial Effects: Compared with existing technologies, the present invention has the following significant advantages: By deeply analyzing the quality of different highland barley raw materials and the corresponding edible quality of fresh highland barley noodles after processing, the present invention explores the relationship between the quality of highland barley raw materials and the quality of processed products. It then establishes a comprehensive quality evaluation model for highland barley noodles based on the Support Vector Machine (SVM) model, replacing traditional analysis methods and improving efficiency. This model provides scientific guidance for the industrial development of highland barley foods, helping companies optimize raw material selection standards, achieve efficient utilization of high-quality raw materials, improve product quality, and better meet market demand. In addition, the research results can also provide theoretical support for improving highland barley quality and selecting and breeding specialized varieties. DETAILED DESCRIPTION

[0021] The technical solution of the present invention is further described below.

[0022] In this embodiment, the following 19 highland barley varieties were collected as experimental raw materials, as shown in Table 1, and the physical and chemical quality, amino acid quality, and gelatinization characteristics of the highland barley raw materials were systematically analyzed. The cooking and edible quality of fresh highland barley noodles were studied, and a support vector machine model (SVM) was used for regression analysis to study the relationship between highland barley quality characteristics and edible quality. Finally, a highland barley fresh noodle processing suitability evaluation system was constructed.

[0023] Table 1:

[0024]

[0025]

[0026] 1. Quality determination of highland barley raw materials

[0027] 1. Preparation of whole barley flour

[0028] After removing impurities from the highland barley grains, they are crushed and passed through a 40-mesh sieve to obtain whole highland barley flour of different production areas and varieties for later use.

[0029] 2. Determination of nutritional components of highland barley

[0030] The moisture content was determined by referring to the first method of GB / T5009.3-2016 "National Food Safety Standard - Determination of Water in Foods" for the calculation of nutrients on a dry basis; the ash content was determined by referring to the first method of GB / T5009.4-2016 "National Food Safety Standard - Determination of Ash in Foods"; the fat content was determined by referring to the first method of GB / T5009.6-2016 "National Food Safety Standard - Determination of Fat in Foods"; the protein content was determined by referring to the first method of GB / T5009.5-2016 "National Food Safety Standard - Determination of Protein in Foods"; the amino acid content was determined by referring to GB / T5009.124-2016 "National Food Safety Standard - Determination of Amino Acids in Foods"; the total starch and amylose contents were determined using a microassay kit; and the β-glucan content was determined using a Megazyme kit.

[0031] 3. Determination of highland barley protein components

[0032] Protein concentration was determined using the Osborne fractionation method and the Coomassie Brilliant Blue method (G250 colorimetric assay). The extraction process was as follows: 2 grams of crushed highland barley sample was weighed and placed in a centrifuge tube. Deionized water was added proportionally, stirred evenly, and then extracted in a constant-temperature water bath with shaking. After centrifugation, supernatant A (albumin) and residue A were obtained. Residue A was added with NaCl solution, and the extraction was repeated to obtain supernatant B (globulin) and residue B. Ethanol solution was added to residue B to extract alcohol-soluble proteins, obtaining supernatant C and residue C. NaOH solution was added to residue C to extract gluten, obtaining supernatant D and residue D. With absorbance A as the ordinate (A) and the mass concentration of bovine serum albumin standard solution (mg / mL) as the abscissa (X), a protein standard curve was drawn, and a regression equation was established. The regression equation was Y=0.154X+0.01-0.00103 (R2=0.9994), and the protein extraction rate of each component was calculated according to the Coomassie Brilliant Blue method (G250 colorimetric method).

[0033] 4. Determination of gelatinization properties of whole barley flour

[0034] Different varieties of highland barley prepared using RVA standard procedure 1, using method 2.1.2.1, were tested. The samples were heated to 50°C, held for 1 minute, then raised to 95°C at a constant rate of 12°C / min⁻¹, held for 2.5 minutes, and then cooled to 50°C at the same rate. The stirring rate was set at 960 rpm for the first 10 seconds, followed by a stirring rate of 160 rpm⁻¹.

[0035] 2. Cooking and Edible Quality Determination of Fresh Barley Noodles

[0036] (1) Preparation of fresh highland barley noodles: refer to SB / T 10137-93.

[0037] Weigh 88g of highland barley flour, add 12g of gluten and 35mL of water, and mix into a dough. Let it rise at room temperature for 20 minutes. Press it using an MT-15 electric noodle machine, with a 4mm spacing between the extrusion rollers. Press it four times to form a sheet. Roll it six times at a 3.5mm gauge, and then roll it twice at gauges of 3.0, 2.5, 2.0, 1.5, and 1.0mm. Use a 2mm-wide cutting die and cut it into strips to produce noodles with a thickness of 1mm and a width of 2mm.

[0038] (2) Cooking and edible quality evaluation of fresh highland barley noodles

[0039] ①Determination of noodle cooking characteristics

[0040] Pour 300 mL of water into a container with a measured weight M1 and bring to a boil. Place 10 noodles in the container and cook. Starting at 2 minutes, remove one noodle every 10 seconds and press it against a glass plate to observe whether the white core disappears. The time when the white core disappears is recorded as the optimal cooking time. Accurately weigh 20 noodles with a mass of m1 and a length of approximately 15 cm. Cook until the optimal cooking time is reached. Quickly remove the noodles and soak them in cold water for 30 seconds. Remove the noodles and drain them through a strainer for 5 minutes. Weigh the mass of the noodles at this point, m2, and count the number of intact noodles, N. Continue boiling the noodle soup to evaporate most of the water. Dry the noodles in a 105°C oven to constant weight. Weigh the container and dry matter, M2. Calculate the noodle water absorption rate, breakage rate, and cooking loss rate using the following formulas.

[0041] Water absorption (%) = (m2-m1) / m1×100% (1)

[0042] Bar breakage rate (%) = (20-N) / 20×100% (2)

[0043] Cooking loss rate (%) = (M2-M1) / m1×100% (3)

[0044] ② Determination of noodle texture characteristics

[0045] Noodles were cooked for the optimal cooking time, soaked in cold water for 30 seconds, drained, and cooled to room temperature before being tested for texture. Compression tests were performed using a TA-XT2i texture analyzer with an N673035 probe. Parameters were set: a pre-test speed of 10 mm / s, a test speed of 15 mm / s, a post-test speed of 5 mm / s, a 50% compression degree, a test force of 5 g, and a 1-second interval between compressions. The stage was kept clean and flat. Three replicates were performed for each sample, with four noodles placed on each stage.

[0046] ③Determination of noodle tensile properties

[0047] Using a TA-XT2i texture analyzer with an A / KIE probe, tensile tests were performed according to the specified parameters to generate a dough tensile characteristic curve. The instrument's automated analysis software then generated values representing maximum tensile resistance, extensibility, and tensile area. Each sample was tested five times in parallel.

[0048] ④Edible quality evaluation:

[0049] A seven-person sensory evaluation panel evaluated the color, appearance, palatability, elasticity, smoothness, and taste of fresh highland barley noodles. The evaluation criteria for each item are shown in the table. The total sensory evaluation score was the sum of all individual scores converted to a percentage. The sensory evaluation criteria are shown in Table 2.

[0050] Table 2:

[0051]

[0052] 3. Use correlation analysis and principal component analysis to reduce the dimension and simplify the 14 indicators of fresh noodles of 19 different varieties of highland barley. Then use the fuzzy membership function method to calculate the membership function values of the nutritional indicators and gelatinization characteristics indicators of different varieties of highland barley. The membership values of the selected indicators are accumulated, and the average value is calculated, and a comprehensive ranking is performed based on the size of the mean.

[0053] The membership function calculation formula is as follows:

[0054] X(i)=(X-Xmin / Xmax-Xmin), i=1, 2, 3, 4, ..., n (4)

[0055] X(i)=1-(X-Xmin / Xmax-Xmin), i=1, 2, 3, 4, ..., n (5)

[0056] Where X(i) is the membership value of the i-th index of a certain barley variety; X is the measured value of the i-th index of a certain barley variety; Xmax and Xmin are the maximum and minimum measured values of each index, respectively. If the measured index is positively correlated with the barley quality, use formula (4) for calculation; if it is negatively correlated, use formula (5) for calculation.

[0057] Data standardization was performed on fresh highland barley noodles from 19 different highland barley varieties, and five principal components were extracted. Table 3 shows the eigenvalues and contribution rates of the principal components. The contribution rates of each principal component were 33.01%, 20.91%, 11.74%, 11.39%, and 7.76%, respectively, and the total contribution rate was 84.81%, indicating that these five principal components play a leading role in the evaluation of quality indicators of fresh highland barley noodles and can comprehensively reflect the main information of the comprehensive evaluation of the quality characteristics of fresh highland barley noodles.

[0058] Table 3:

[0059]

[0060]

[0061] Table 4 shows the loadings and coefficients of each principal component. The coefficient of each indicator in a principal component is the quotient of the square root of the eigenvalue and the corresponding loading. The linear relationships between the five principal components and each indicator were calculated using the component loading matrix and the eigenvalues of each principal component. In PC1, hardness, elasticity, chewiness, and adhesiveness have high loadings and have a positive impact on PC1; in PC2, cohesion and resilience have high loadings and have a positive impact; in PC3, adhesion has high loadings and has a positive impact; in PC4, optimal cooking time has high loadings and has a positive impact, and can be selected as an evaluation indicator; in PC5, maximum tensile force has high loadings and has a positive impact. Since the high loadings in PC1, PC2, PC3, and PC5 are textural properties, and the high loading in PC4 is cooking properties, it can be seen that textural properties have the greatest impact on the quality of fresh highland barley noodles, followed by cooking properties.

[0062] Table 4:

[0063] index PC1 PC2 PC3 PC4 PC5 hardness 0.98 0.02 0.14 0.03 0.07 Adhesion 0.20 -0.09 0.85 0.13 0.05 elasticity 0.68 0.39 -0.43 -0.23 -0.22 chewability 0.97 0.22 0.02 -0.06 0.03 Adhesion 0.97 0.16 0.14 0.03 0.09 Cohesion -0.02 0.94 -0.19 -0.07 0.05 Resilience 0.27 0.84 0.04 0.05 -0.02 Maximum tensile force 0.00 0.14 -0.01 0.64 0.57 Displacement at maximum force -0.15 -0.15 -0.86 0.04 0.22 Water absorption 0.41 -0.18 0.18 0.63 -0.32 Bar breakage rate -0.34 -0.42 0.58 -0.26 0.36 Cooking loss rate -0.15 -0.66 -0.39 0.42 0.10 Optimal steaming time -0.17 -0.09 -0.02 0.89 -0.09

[0064] Set the barley quality index values to X1~X 14 , the five principal components are set as Fn (n = 1, 2, 3, 4, 5), and the expression is as follows:

[0065] F1=0.61X1-0.493X2+0.819X3+0.905X4+0.776X5+0.686X6+0.502X7-0.115X8+0.021X9-0.25

[0066] 3X 10 -0.672X 11 -0.42X 12 -0.365X 13 -0.301X 14

[0067] F2=0.51X1+0.653X2-0.306X3+0.233X4+0.47X5-0.018X6+0.301X7+0.182X8-0.793X9+0.746

[0068] X 10 +0.206X 11 -0.296X 12 +0.463X 13 +0.145X 14

[0069] F3=0.457X1+0.052X2+0.284X3+0.076X4+0.316X5-0.543X6-0.452X7-0.329X8+0.292X9+0.3

[0070] 11X 10 -0.159X 11 +0.482X 12 -0.066X 13 +0.315X 14

[0071] F4=-0.23X1-0.309X2+0.059X3+0.084X4-0.102X5+0.321X6+0.071X7+0.303X8+0.105X9+0.

[0072] 325X 10 -0.472X 11 +0.456X 12 +0.737X 13 +0.186X 14

[0073] F5=-0.238X1+0.06X2-0.114X3+0.117X4+0.199X5-0.047X6-0.114X7+0.54X8+0.3X9-0.049X

[0074] 10 +0.3X 11 +0.335X 12 +0.081X 13 -0.736X 14

[0075] In order to further evaluate the comprehensive quality characteristics of different varieties of fresh highland barley noodles, a comprehensive evaluation mathematical model for highland barley varieties was established based on the above five principal components:

[0076] F=(33.01 / 84.81)×F1+(20.91 / 84.81)×F2+(11.74 / 84.81)×F3+(11.39 / 84.81)×F4+(7.76 / 84.81)

[0077] ×F5

[0078] Table 5 shows the comprehensive evaluation and ranking of different varieties of fresh highland barley noodles. The model was used to calculate the comprehensive scores of the main quality traits of each highland barley variety. The top three varieties were Zangqing 25, Kunlun 14, and Zangqing 3000.

[0079] Table 5:

[0080] variety <![CDATA[F1]]> <![CDATA[F2]]> <![CDATA[F3]]> <![CDATA[F4]]> <![CDATA[F5]]> Comprehensive evaluation (F) 0349-1 variety -2.85 1.00 1.80 0.56 -0.54 -0.59 1128 varieties -6.29 -1.22 1.35 1.85 -0.46 -2.36 Navy 320 1.05 2.55 0.71 -0.28 -0.23 1.08 Navy 18 4.94 -4.23 -0.94 0.03 0.19 0.77 Navy 2000 0.60 2.58 0.48 -0.01 -0.31 0.91 Navy 25 6.32 1.77 -2.06 1.87 -1.52 2.72 Navy 3000 3.42 2.48 -0.28 -0.72 -0.40 1.77 Holly No. 18 -6.08 -1.93 0.09 2.32 -0.43 -2.56 Ganqing No. 10 1.45 1.80 -0.81 -0.17 -0.77 0.80 Ganqing No. 11 -7.12 1.58 -1.20 -1.73 0.07 -2.77 Ganqing No. 4 2.26 -0.79 0.55 -1.56 0.80 0.62 Ganqing No. 8 2.73 -0.43 1.81 -1.03 0.82 1.14 Ganqing No. 9 0.91 -1.45 1.97 -1.15 0.34 0.15 Kangqing No. 9 -5.29 0.39 -1.44 -2.59 -0.18 -2.53 Kunlun 14 1.79 3.14 -0.27 2.13 1.88 1.89 Longzihei -0.61 0.71 0.26 0.32 -0.68 -0.05 Xila No. 22 1.96 -3.30 -1.56 0.20 -0.42 -0.28 Xila No. 23 3.03 -5.23 0.02 0.09 0.04 -0.09 Purple barley -2.23 0.60 -0.48 -0.13 1.78 -0.64

[0081] 4. The key indicator data were selected using the grey correlation method. In this embodiment, DPS software was used to perform grey correlation analysis on the dimensionless processed data. The resolution coefficient value range was (0, 1), and the specific resolution coefficient value was 0.5. The grey correlation and ranking of the main quality indicators of highland barley and the comprehensive quality score of fresh highland barley noodles are shown in Table 6. Ash content, β-glucan, protein, amylose, albumin, gluten, disintegration value, retrogradation value, trough viscosity, and peak viscosity are 10 indicators. As shown in Table 6, the order of correlation between the comprehensive quality score of fresh highland barley noodles and each quality indicator of highland barley is: ash content > β-glucan > protein > amylose > albumin > gluten > disintegration value > retrogradation value > trough viscosity > peak viscosity, indicating that ash content has the greatest impact on the comprehensive quality score of fresh highland barley noodles, followed by β-glucan. Protein can continue to screen indicators, and ash, β-glucan, protein, amylose, albumin, and gluten are selected to establish a comprehensive evaluation model.

[0082] Table 6:

[0083]

[0084] V. Establishment and comparison of three models: In this embodiment, the six selected highland barley quality indicators and the comprehensive score of highland barley noodles quality were used to establish PLS model, BP model and support vector machine model respectively. The accuracy of the constructed PLS model R 2 =0.40348, the accuracy of the BP model R 2 =0.66855, the accuracy of the SVM model R 2 =0.75452, and the RMSE (root mean square) values of the regression model are 0.4071, 0.61445, and 0.71086, respectively, indicating that the SVM model has a good ability to extract information in the present invention, the established model is relatively stable, has high prediction accuracy, and is highly reliable.

[0085] Sixth, a comprehensive quality evaluation model for fresh highland barley noodles was obtained by training a support vector machine model using the above key indicator data. Support vector machines (SVMs) can perform supervised classification and regression on samples. The present invention uses support vector machine regression. The support vector machine model fits a model based on a set of training data with known responses, and then uses this model to predict the category of new observations. The SVM model classifies data by optimizing the hyperplane used to separate the classes, which can also be viewed as finding the hyperplane that maximizes the margin between classes.

[0086] The present invention selects an optimization model with a radial basis kernel function (RBF). The optimal model parameters are configured as follows: the cost parameter is 1, and the kernel function parameter Gramma is 0.1. The goal is to find a hyperplane through training that minimizes the distance between all sample points. The hyperplane can be expressed by formula (6):

[0087] f(x)=ω T x+b (6)

[0088] Where: ω is the transpose of the weight vector, b is the bias.

[0089] The problem of finding the optimal hyperplane can be transformed into the corresponding convex quadratic programming problem:

[0090]

[0091] Where the penalty factor c determines the weights of the two variables in formula (7); ξi and ξi are slack variables used to relax the constraints on the optimal hyperplane. Sample data within the insensitive band defined by ε is not included in the loss; only samples outside the insensitive band will affect the support vector regression model.

[0092] 7. Use the comprehensive quality evaluation model of fresh highland barley noodles to evaluate the quality of fresh highland barley noodles.

[0093] The present invention establishes a comprehensive quality evaluation model for fresh highland barley noodles and performs regression analysis based on a support vector machine model. This method replaces traditional analysis methods, improves efficiency, provides scientific guidance for the industrial development of highland barley food, helps enterprises optimize raw material selection standards, achieves efficient utilization of high-quality raw materials, improves product quality, and better meets market demand.

Claims

1. A method for comprehensive quality evaluation of fresh highland barley noodles based on a SVM model, characterized in that: The following steps are involved: S1. Taking different varieties of highland barley raw materials, measuring the quality of the different varieties of highland barley raw materials to obtain raw material quality index data; S2. Prepare fresh highland barley noodles from different highland barley raw materials, and evaluate the cooking quality and edible quality of the corresponding fresh highland barley noodles to obtain finished product quality index data, and establish a comprehensive evaluation model using principal component analysis; S3, using correlation analysis to simplify the highland barley index of step S1 and the comprehensive evaluation obtained in S2, and then using the grey correlation method to select key indicator data; S4, using the key indicator data of step S3 to perform support vector machine model training to obtain a comprehensive quality evaluation model for fresh highland barley noodles; S5. Use the comprehensive quality evaluation model of fresh highland barley noodles to evaluate the quality of fresh highland barley noodles.

2. The method for comprehensive quality evaluation of fresh highland barley noodles based on the SVM model according to claim 1, wherein Step S1 specifically includes the following steps: S11, preparing whole highland barley flour; S12, determining the nutritional components of highland barley, including moisture content determination, ash content determination, fat content determination, protein content determination, amino acid content determination, total starch, amylose content determination, and β-glucan content determination; S13, determining the highland barley protein component; S14, determining the gelatinization properties of the whole highland barley flour.

3. The method for comprehensive quality evaluation of fresh highland barley noodles based on the SVM model according to claim 2, wherein The whole highland barley flour prepared in step S11 is passed through a 40-mesh sieve.

4. The method for comprehensive quality evaluation of fresh highland barley noodles based on the SVM model according to claim 1, wherein The cooking quality of the fresh highland barley noodles in step S2 includes noodle cooking characteristics, noodle texture characteristics, and noodle stretching characteristics.

5. The method for comprehensive quality evaluation of fresh highland barley noodles based on the SVM model according to claim 1, wherein The specific method for evaluating the edible quality is: using a sensory evaluation method to evaluate the color, appearance, palatability, elasticity, smoothness and taste of the fresh highland barley noodles.

6. The method for comprehensive quality evaluation of fresh highland barley noodles based on the SVM model according to claim 1, wherein In step S4, the support vector machine model uses a radial basis kernel function, a cost parameter of 1, a kernel function parameter Gramma of 0.1, and a hyperplane formula as follows: f(x)=ω T x+b, Where: ω is the transpose of the weight vector, b is the bias; The formula for finding the optimal hyperplane is as follows: In the formula, the value of the penalty factor c determines the weights of the two variables in the formula. ξi and ξi are slack variables used to relax the constraints on the optimal hyperplane. Sample data within the insensitive band defined by ε are not included in the loss. Only samples outside the insensitive band will affect the support vector regression model.

Images