Fruit and vegetable preservative efficient screening method and system based on multi-criterion decision and random forest

Abstract

The invention provides a fruit and vegetable preservative efficient screening method and system based on a multi-criterion decision and a random forest. The method comprises the following steps: collecting a color change index, a rot index and a safety index of fruit and vegetable tissues subjected to preservation treatment; analyzing the importance of each index through a machine learning algorithm, and generating a dynamic weight coefficient; inputting the dynamic weight and the standardized index into a nonlinear calculation model, adjusting the contribution degree of each index through the difference of preset shape parameters, and outputting a comprehensive score; and grading and screening the preservative based on the comprehensive score.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of agricultural product preservation technology and artificial intelligence, specifically involving a method and system for efficient screening of fruit and vegetable preservatives based on multi-criteria decision-making and random forest. Background Technology

[0002] As people's demands for food safety and quality continue to rise, fruit and vegetable preservation technology has become an indispensable link in the agricultural industry chain. On the one hand, because horticultural crops are mostly grown in suburbs or mountainous areas far from cities, their ripening period is relatively concentrated; on the other hand, aging and decay are the main factors leading to a decline in the quality of horticultural crops. Statistics show that approximately 30% of fresh fruits and vegetables rot after harvest each year, causing economic losses exceeding 100 billion yuan. Therefore, there is an urgent need for efficient and reasonable storage and transportation technologies to resolve the market supply and demand contradiction between "concentrated market entry" and "long-term supply." Due to improper preservation measures, post-harvest losses of fruits and vegetables reach as high as 20%-30% annually, resulting in economic losses exceeding hundreds of billions of yuan.

[0003] As a crucial means of extending the shelf life of fruits and vegetables, the selection method of preservatives directly affects preservation effectiveness and food safety. Therefore, chemical preservatives have become the most commonly used preservation technology due to their low cost, ease of operation, and widespread adoption. With increasing public awareness of healthy eating, developing and screening more non-toxic, harmless, pollution-free, and low-cost chemical preservatives has become a key research focus. Over the past 20 years, several preservatives with market application potential have been successfully developed, such as 1-methylcyclopropene (1-MCP) and chlorpyrifos (CPPU), which can be used to preserve fruits and vegetables such as bananas, kiwifruit, cantaloupe, and broccoli, extending post-harvest storage time. However, the screening process for these preservatives faces problems such as inconsistent operating standards, poor repeatability, and high screening costs. Therefore, the following technical bottlenecks urgently need to be overcome in the screening of fruit and vegetable preservatives:

[0004] 1. High dependence on subjective experience: Traditional screening methods mainly rely on the experience of experts or single-indicator experiments, lacking a systematic and quantitative evaluation system, and making it difficult to effectively balance the trade-offs between multiple performance indicators.

[0005] 2. Limitations of static weights: Existing evaluation methods mostly use fixed weight coefficients, which cannot reflect the dynamic changes in the importance of various indicators among different fruit and vegetable varieties and storage environments.

[0006] 3. Ignoring nonlinear relationships: There are complex nonlinear interactions between yellowing, decay and safety indicators, and conventional linear weighted models are unable to accurately capture these complex relationships.

[0007] 4. Insufficient data utilization: A large amount of experimental data is only used for simple comparative analysis, failing to delve into potential patterns through advanced algorithms, resulting in low screening efficiency.

[0008] 5. The screening process is costly: the amount of fruit and vegetable samples required to screen a fruit and vegetable preservative can be tens or even hundreds of kilograms.

[0009] In the prior art, patent CN19534688B proposes a preservative based on a compound plant extract, but its screening process still relies on manual experiments; patent CN41580487A uses a compound formulation of water-soluble chitosan and a natural antibacterial agent, but the evaluation method is still the traditional sensory assessment. Currently, there are principles regarding the safety and effectiveness of fruit and vegetable preservative screening, but quantitative implementation methods are lacking. Although the random forest algorithm has been applied in the field of food safety (e.g., CN11522808A), there are no reports of combining it with multi-criteria decision-making for preservative screening. Furthermore, there are currently no reports or literature on methods for efficiently screening and evaluating the effects of preservatives based on standardized preparation of fruit and vegetable circular tissue slices and tissue culture. Summary of the Invention

[0010] To address the shortcomings and deficiencies of existing technologies, this invention provides a highly efficient screening method and system for fruit and vegetable preservatives that integrates intelligent decision-making and standardized experiments. The innovative tissue culture technology reconstructs the preservative screening paradigm, significantly reducing R&D costs while maintaining technical rigor. Specifically, standardized fruit and vegetable discs with a diameter of 0.5-1.2 cm are used as experimental carriers. The size is standardized using a precision punch, and the samples are then sterilized with 0.5% NaClO solution to ensure a sterile environment. This miniaturized design, combined with an innovative culture system—sterile filter paper perfectly adhering to the bottom of the culture dish and maintaining constant humidity through spraying sterile water—allows for the parallel processing of more than 50 disc samples in a single experiment, enabling simultaneous testing of multiple preservative concentrations and reducing raw material consumption by more than 90% compared to traditional methods.

[0011] At the data processing level, this invention breaks through the limitations of traditional static evaluation by constructing an adaptive decision engine. It captures apparent changes during storage in real time using image sensors, accurately quantifying the preservation effect based on the proportion of yellowing area (25%~50% corresponds to level 3 color change) and the distribution of rotten areas (50%~100% corresponds to level 3 rot). More importantly, this method innovatively integrates the random forest algorithm with multi-criteria decision theory: first, it combines the feature weights of three core indicators (color change, rot, and safety) based on the reduction of Gini impurity and permutation importance analysis; then, it uses the Softmax function to achieve dynamic weight allocation. This mechanism can automatically adjust constraint rules according to the type of fruit and vegetable; for example, it mandates a safety weight of ≥60% for berries, while setting a lower limit of ≥40% for vegetables, effectively responding to the needs of different storage environments.

[0012] To address the industry challenge of nonlinear interactions between indicators, this invention designs an evaluation model with a shape adjustment factor. This model amplifies the decision weight of safety by setting differentiated parameters (safety adjustment parameter c > decay parameter b ≥ color change parameter a ≥ 1), generating a comprehensive score of 0-100 points and classifying it into five levels. Validation of examples shows that preservatives achieving high ratings (such as 0.1 mM adenosine triphosphate) perform exceptionally well in inhibiting deterioration and ensuring safety, while schemes scoring below 60 points exhibit serious deficiencies. This quantitative evaluation method overcomes the limitations of traditional linear models, accurately capturing the complex correlation between yellowing, decay, and toxicity.

[0013] The resulting closed-loop technology integrates the ADMET safety assessment system, which constructs a safety index using parameters such as hepatotoxicity (elevated ALT), mutagenicity (Ames test), and acute toxicity (LD50). The system's overall workflow encompasses a complete innovation chain, from standardized sample preparation and image feature extraction to dynamic weight calculation and nonlinear evaluation, providing the first engineering solution for the fruit and vegetable preservation field that integrates cost reduction in experiments, intelligent decision-making, and quantitative assessment.

[0014] The specific technical solution adopted by this invention to solve its technical problem is as follows:

[0015] A method for screening fruit and vegetable preservatives includes:

[0016] The color change index, decay index, and safety index of vegetables after preservation treatment were collected.

[0017] The importance of each index is analyzed using machine learning algorithms to generate dynamic weighting coefficients.

[0018] The dynamic weights and standardized indices are input into the nonlinear calculation model, and the contribution of each indicator is adjusted by pre-setting shape parameters to output a comprehensive score.

[0019] Preservatives are selected based on a comprehensive scoring and grading system.

[0020] Furthermore, the shape parameters satisfy c>b≥a≥1, where a, b, and c correspond to the adjustment parameters of the color change index, decay index, and safety index, respectively.

[0021] Furthermore, the dynamic weighting coefficients are generated through the following steps:

[0022] The importance of the combined features of the color change index, decay index, and safety index is calculated based on the combination Gini impurity reduction and permutation importance using the random forest algorithm.

[0023] Weights are dynamically allocated using a normalized exponential function.

[0024] Furthermore, the safety index is obtained through prediction using the ADMET system and includes parameters for hepatotoxicity, mutagenicity, and acute toxicity.

[0025] Furthermore, the color change index and decay index are determined by the area ratio identified through image acquisition equipment and calculated according to a preset grading standard.

[0026] Furthermore, the fruit and vegetable tissue is a circular slice with a diameter of 0.5-1.2 cm, and its preparation includes:

[0027] Use a punch to cut leaves or fruit peel tissues of uniform size;

[0028] Disinfect with 0.5% NaClO solution for 3 minutes, then blot dry the surface moisture;

[0029] Each screening process includes at least 50 round tissue samples, with 3 replicates per preservative concentration.

[0030] Furthermore, the culture environment is constructed in the following manner:

[0031] Place sterile round filter paper flat on the bottom of the petri dish and spray with sterile water to make the filter paper completely adhere to the bottom of the dish;

[0032] The circular tissue samples were arranged in a 3×3 array on the filter paper.

[0033] Furthermore, the lower limit of the safety weight for preservative constraints on berries is 60% when dynamically weighting preservatives, and the lower limit of the constraint on vegetables is 40%.

[0034] Furthermore, a real-time monitoring mechanism is adopted in the dynamic weight allocation process, which automatically recalculates the weights after the data is updated and smooths the fluctuations through the sliding window mean.

[0035] And, a fruit and vegetable preservative screening system, comprising:

[0036] Data acquisition module: Configured to acquire the color change index, decay index, and safety index of vegetables and fruits after preservation treatment;

[0037] Dynamic weight calculation engine: Analyzes the importance of each index through machine learning algorithms and generates dynamic weight coefficients;

[0038] Nonlinear scoring processor: Receives dynamic weights and standardized indices, adjusts the contribution of indicators through preset shape parameters, and outputs a comprehensive score;

[0039] Decision output interface: Based on comprehensive scoring and grading, preservatives are screened and recommended solutions are generated.

[0040] And a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method described above.

[0041] A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0042] Compared to existing technologies, this invention and its preferred solution achieve multi-dimensional technological breakthroughs through systematic innovation. By reconstructing standardized tissue culture technology, it completely changes the high-cost operation mode of traditional preservative screening. Using fruit and vegetable discs with a diameter of 0.5-1.2 cm as miniature experimental carriers, combined with humidity control technology that uses sterilized filter paper attached to the bottom of the culture dish, more than 50 sample units can be processed simultaneously in a single experiment. This miniaturized parallel architecture significantly reduces raw material consumption, solving the industry pain point that traditional methods require hundreds of kilograms of samples.

[0043] At the data-driven decision-making level, this invention is the first to integrate a dynamic weighting mechanism with a nonlinear evaluation model. Based on the random forest algorithm, it analyzes the changes in importance of color change, decay, and safety indicators in real time, and uses the Softmax function to achieve adaptive weight allocation. For different fruit and vegetable characteristics (such as the high safety requirements of berries), the system automatically applies differentiated constraints, effectively solving the evaluation distortion problem caused by fixed weights. More significantly, it introduces a shape parameter adjustment mechanism, capturing the complex nonlinear relationships between indicators through parameterized design (prioritizing safety parameters), significantly improving the identification accuracy of high-safety preservatives.

[0044] The multi-source quantification technology integrated in the closed-loop technology provides solid support for decision-making: image sensors accurately identify the distribution of discoloration / rotten areas, replacing highly subjective sensory evaluations; the ADMET system integrates safety parameters such as hepatotoxicity and mutagenicity to establish an objective toxicity assessment benchmark. The resulting engineering system covers the entire process from sample preparation and feature extraction to intelligent decision-making, providing the first integrated solution for fruit and vegetable preservation that combines cost reduction, quantification, and adaptive design. Attached Figure Description

[0045] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0046] Figure 1 This is a flowchart illustrating steps S1 and S2 in an embodiment of the present invention.

[0047] Figure 2 This is a schematic diagram illustrating the standardized fabrication of circular tissue samples from Chinese cabbage leaves in an embodiment of the present invention.

[0048] Figure 3This is a schematic diagram of the 2D structure of an energy-related compound in an embodiment of the present invention.

[0049] Figure 4 These are comparative images of the appearance characteristics of the circular tissue of Chinese cabbage leaves during storage under different preservative treatments in this embodiment of the invention.

[0050] Figure 5 This is a flowchart illustrating the overall process of an embodiment of the present invention. Detailed Implementation

[0051] To make the features and advantages of the present invention more apparent and understandable, specific embodiments are described below in detail:

[0052] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0053] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0054] Traditional plant tissue culture technology refers to the cultivation of isolated plant organs, tissues, or cells in an artificially controlled environment to regenerate complete plants. Tissue disc culture technology can be used to study the growth patterns of cultured parts without interference from other parts of the plant, and to address theoretical and practical production problems by adjusting various conditions to influence their physiological activities. Therefore, considering accelerating the screening process of preservatives and reducing R&D costs, this invention cultivates tissue discs (0.5–1.2 cm) of Chinese cabbage leaves, longan, and lychee peels in ultrapure water to measure changes in conductivity before and after boiling. This standardized preparation of postharvest fruit and vegetable tissue discs enables simultaneous screening of multiple preservatives while effectively controlling R&D costs.

[0055] Building upon this foundation, this invention further acquires preservation characteristics during storage by standardizing the batch production of fruit and vegetable disc tissue and extracting images using image sensors. The safety index of preservatives is then evaluated using the ADMET system. This invention integrates three core indicators for evaluating fruit and vegetable preservation quality (color change index, decay index, and safety index) and innovatively proposes a non-linear comprehensive scoring standard through a dynamic weight allocation model constructed using machine learning algorithms. This enables precise assessment and efficient screening of the market potential for fruit and vegetable preservatives.

[0056] Therefore, this invention provides an efficient screening method for fruit and vegetable preservatives that integrates multi-criteria decision theory and machine learning algorithms, particularly an innovative technology based on random forest feature importance analysis and multi-attribute comprehensive evaluation. This method extracts preservation effect features (yellowing / browning index and decay index) during storage experiments using standardized batch production of fruit and vegetable micro-slices and batch extraction of images from image sensors. It evaluates the ADMET characteristics of preservatives to calculate a safety index. By constructing a dynamic weight model for the yellowing / browning index, decay index, and safety index, and combining it with machine learning algorithms for intelligent analysis of large amounts of preservative data, this method solves the problems of strong subjectivity, low efficiency, and difficulty in quantitative evaluation in traditional preservative screening processes. It provides a scientific, efficient, and quantifiable decision support tool for the field of fruit and vegetable preservation.

[0057] To verify the efficient screening method for postharvest preservatives for fruits and vegetables proposed in this invention, such as... Figure 5 As shown, taking the screening of preservatives for Chinese cabbage as an example, the solution was implemented according to the following steps:

[0058] Step S1: Standardized production of fruit and vegetable tissue slices and implementation of preservation solutions;

[0059] Step S2: Multi-dimensional data collection and standardization;

[0060] Step S3: Random Forest Feature Importance Analysis;

[0061] Step S4: Optimize the combined weights of fruit and vegetable preservation effects;

[0062] Step S5: Comparison of the nonlinear comprehensive evaluation model of preservatives with traditional evaluation methods.

[0063] like Figure 1 As shown, step S1 involves the standardized preparation of postharvest vegetable leaf tissue discs; disinfection and drying of the vegetable leaf disc tissues; selection and standardized preparation of several preservatives; preparation of tissue culture dishes; culture treatment of the vegetable disc tissues; and statistical evaluation of the preservation effects of different preservatives. This invention uses Chinese cabbage as an example. By preparing discs from Chinese cabbage leaf tissues and utilizing tissue culture technology, it simulates, screens, and analyzes the postharvest preservation effects of different preservatives on Chinese cabbage. It has advantages such as low cost and the ability to screen multiple concentrations of multiple preservatives in a single step.

[0064] Specifically, the steps include the following:

[0065] Step S11: Standardized preparation of fruit and vegetable tissue slices. Harvest Chinese cabbage (Brassica rapa var. parachinensis) on a sunny day, with a harvest period of about 35 days. The day of harvest is set as day 0. Select Chinese cabbage leaves that are free from mechanical damage, pests and diseases, and have uniform maturity and color. Select leaves of relatively uniform size and use a 1.2 cm diameter punch to take small round slices from each leaf and place them in a tissue culture dish.

[0066] Step S12: Disinfection and drying of fruit and vegetable round slices. Immerse the leaf round slices in 0.5% NaClO solution for 3 minutes, then quickly blot dry with paper towels.

[0067] Step S14: Exploring the dissolution method of the preservative. Due to the differences in the properties of the compounds, this embodiment attempts to initially dissolve the preservative compounds using acidic, alkaline, and ethanol methods. The initial solubility of each compound and the final concentration used as a preservative are shown in Table 1.

[0068] Table 1 Safety information of preservative compounds

[0069] Drug Name CAS Login Number English name solvent Solubility Final concentration Adenosine triphosphate 9000-83-3 Adenosine triphosphate water Easily soluble in water 0.01mM, 0.05mM, 0.1mM, 0.2mM, 0.5mM, 1mM 6-Benzylaminopurine 1214-39-7 6-Benzylaminopurine ethanol Slightly soluble in ethanol 0.01mM, 0.05mM, 0.1mM, 0.2mM, 0.5mM, 1mM Chlorpyrifos 68157-60-8 Forchlorfenuron ethanol Soluble in ethanol 0.01mM, 0.05mM, 0.1mM, 0.2mM, 0.5mM, 1mM choline chloride 67-48-1 Choline chloride water Easily soluble in water and alcohols 0.1g / L, 1g / L Sodium dichloroisocyanurate 2893-78-9 Sodium dichloroisocyanurate water Easily soluble in water, sparingly soluble in organic solvents 0.1g / L, 0.2g / L Sodium para-aminosalicylate 133-10-8 Sodium 4-aminosalicylate water It is readily soluble in water, slightly soluble in ethanol, and insoluble in ether. 0.1g / L, 0.3g / L

[0070] Step S14: Preparation of the preservative solution and preparation of the round tissue culture dishes. Based on the appropriate dissolution method of the compound, prepare a 50-fold stock solution (10 ml), and then dilute it again to a 1-fold solution (50 ml) to the target concentration before use. Sterilize the culture dishes, filter paper, forceps, and ultrapure water using high-temperature autoclaving. The autoclave used was an HVE-50 (Japan), set to 125℃ for 30 min. After the temperature dropped to 90℃, remove it from the autoclave and place it in a 60℃ oven to dry overnight until fully dry. Then, store it at room temperature for later use. Lay the round filter paper flat on the round culture dish and spray it with sterile water to ensure complete adhesion to the bottom of the dish. On the one hand, the white filter paper helps observe the morphological changes of the round slices of the Chinese cabbage leaves later; on the other hand, the moist filter paper helps maintain the microenvironmental humidity of the cultured Chinese cabbage leaf round slice tissue.

[0071] Step S15: Preservative treatment and culture of fruit and vegetable circular tissue slices. The solution diluted to the target concentration in Step S14 is poured into a glass petri dish or beaker; 50 choy sum circular tissue slices are placed in the solution and gently stirred to ensure they are completely submerged; after immersion for 5 minutes, the slices are removed with tweezers and laid flat on a dry paper towel. The remaining preservative solution is gently blotted with absorbent paper, taking care not to damage the slices; once the surface of the slices is dry, they are gently transferred with tweezers to the prepared petri dish from Step 3, arranged in a 3x3 pattern. The petri dish is covered and labeled accordingly. Biological replicates are performed in 3 petri dishes for each concentration of each preservative agent according to Table 2; the petri dish containing the choy sum circular tissue slices is then placed in a stable room temperature environment (25℃, 70% relative humidity) for 4 days.

[0072] Step S2 specifically includes the following steps:

[0073] Step S21: Collect safety data for the preservative. ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) pharmacokinetic methods are crucial in modern drug design and screening, and can be used as a safety assessment standard for screening compounds as preservatives. AmberSAR is the world's most up-to-date and comprehensive database of the absorption, distribution, metabolism, extraction, and toxicity assessment properties of various chemicals (http: / / lmmd.ecust.edu.cn / admetsar2). Based on literature searches, 14 compounds with similar structures or functions to adenosine triphosphate were identified, such as... Figure 3 As shown, most of their structures contain adenine structures (shaded areas), while sodium dichloroisocyanurate, chlorpyrifos, choline chloride, and sodium para-aminosalicylate contain precursor structures that generate adenine structures. Safety evaluation results were obtained by searching the CAS registry numbers of these compounds using tools such as ADMETlab 3.0, as shown in Table 2. Key parameters extracted from the ADMET analysis include: absorption (A): water solubility (Log S), intestinal permeability (Caco-2 value); distribution (D): plasma protein binding rate (PPB), blood-brain barrier penetration (log BB); metabolism (M): CYP450 enzyme metabolic tendency (e.g., CYP3A4 substrate probability); excretion (E): renal clearance (CL); toxicity (T): acute toxicity (LD50), mutagenicity (Ames test), hepatotoxicity (e.g., ALT elevation). The toxicity prediction results were normalized and dynamically weighted to comprehensively assess the safety risk value of the preservatives.

[0074] Table 2 Toxicity evaluation results of preservative compounds

[0075] Drug Name CAS Login Number Evaluation results Referenced research papers Adenosine triphosphate 9000-83-3 Drugs can induce liver damage Chem. Res.Toxicol. 2010,23, 171-183 6-Benzylaminopurine 1214-39-7 The LD50 for fish toxicity is 10,000-100,000 ug / L; it has no inhibitory effect on CYP4501A2 and CYP4502D6, with inhibition efficiencies pAC50 of 5.3 nM and 4.9 nM, respectively; it has no inhibitory effect on CYP4502C9, CYP4502C19, and CYP4503A4. QSAR Comb. Sci.28, 2009, 28:1418 - 1431; NatBiotechnol, 2009, 27: 1050-1055 Chlorpyrifos 68157-60-8 It does not exhibit antagonism against estrogen receptor α, androgen receptor, peroxisome proliferator-activated receptor γ, peroxisome proliferator-activated receptor δ, lipoprotein X receptor, glucocorticoid receptor, thyroid hormone receptor β, or vitamin D receptor. Environ HealthPersp, 2011,119:1142-1148; J.Chem. Inf.Model. 2007, 47,1395-1404 choline chloride 67-48-1 Toxicity LD50 to fish > 100,000 ug / L QSAR Comb. Sci.28, 2009, 28:1418 - 1431 Sodium dichloroisocyanurate 2893-78-9 The acute toxicity LD50 in rats was 2.227 mol / kg; no AMES toxicity was observed; the toxicity pIC50 in quail was -0.9 mmol / kg. Chem. Res. Toxicol. 2009, 22, 1913-1921; J. Chem. Inf. Model. 2009, 49, 2077-2081; Eur. J. Med. Chem, 2007, 42: 606-613 Sodium para-aminosalicylate 133-10-8 It does not inhibit P-glycoprotein, CYP450, or HERG; it is carcinogenic and non-biodegradable; it has no AMES toxicity; the acute LD50 in rats is 1.5761 mol / kg. In Vitro CellDev Biol Anim,2019, 55(5):368-375; PLoS One,2016, 11(3):e0149754

[0076] Step S22: Collection of the color change index (YI) of the circular tissue. This invention selects the color change index to statistically analyze the preservation effect of different preservatives and their concentrations on the circular tissue of Chinese cabbage leaves. The scoring and specific characteristics of the color change index are shown in Table 3.

[0077] Table 3. Scoring and Specific Characteristics of the Color Change Index

[0078] series Specific features 1 No obvious color change 2 The area of ​​color change accounts for 0-25% of the total area of ​​the disc. 3 The area of ​​color change accounts for 25% to 50% of the total area of ​​the disc. 4 The area of ​​color change accounts for 50-75% of the total area of ​​the disc. 5 The area of ​​color change accounts for 75% to 100% of the total area of ​​the disc. 6 The entire color changed.

[0079] Calculation formula:

[0080] Standardization process:

[0081] in and These are the maximum and minimum color change values ​​in the dataset, respectively.

[0082] Step S23: Collection of the decay index (RI) of the circular tissue. This invention uses the decay index to statistically analyze the preservation effects of different preservatives and their concentrations on the circular tissue of Chinese cabbage leaves. The scoring and specific characteristics of the decay index are shown in Table 4.

[0083] Table 4. Scoring and Specific Characteristics of the Decay Index

[0084] level of decay Specific features 1 No decay 2 The rotten area accounts for 0-50% of the total area of ​​the disc. 3 The rotten area accounts for 50% to 100% of the total area of ​​the disc. 4 Completely rotten

[0085] Calculation formula:

[0086] Standardization process:

[0087] in and These are the maximum and minimum decay index values ​​in the dataset, respectively.

[0088] The implementation of step S3 specifically includes the following steps;

[0089] Step 31: Setting model parameters, including:

[0090] (1) Number of decision trees: set to 500 to ensure stability (verified by OOB error curve);

[0091] (2) Feature subset size: The number of features considered when splitting each tree. (3 features in total)

[0092] (3) Splitting criterion: The principle of minimizing Gini impurity is adopted:

[0093] Where p(i|t) is the proportion of class i in node t, and c is the number of classes of samples in the current node (e.g., c=2 in a binary classification problem).

[0094] (4) Maximum depth of the tree: No limit is set until the node is pure or contains less than 5 samples;

[0095] (5) Other parameters: Minimum number of split samples is 2; minimum number of leaf samples is 1; Bootstrap sampling ratio: 100% (with replacement).

[0096] Step 32: The model training process includes:

[0097] (1) Bootstrap sampling: N sample subsets are drawn with replacement from the training set to construct each decision tree;

[0098] (2) Out-of-bag (OOB) estimation: Validating model performance using approximately 36.8% of the data that was not used in training:

[0099]

[0100] (3) Node splitting strategy: For each node, find the optimal splitting point from a randomly selected feature subset.

[0101] Step 33: Feature Importance Analysis Method.

[0102] (1) Calculate feature X based on the importance score of Gini impurity reduction j The average decrease in impurity due to splitting across all trees:

[0103]

[0104] in:

[0105] N trees The total number of trees in the forest; T represents a single tree; t represents the number of trees in T. Using X... j Split nodes; ΔGini(t,X) j Use X for node t j The reduction in Gini impurity due to splitting.

[0106] (2) Based on the importance of the arrangement

[0107] The importance of assessing the degree of model performance degradation by shuffling eigenvalues ​​is as follows:

[0108]

[0109] Where, N trees Given the total number of trees in the forest; calculate the raw error of each tree t on the OOB data. Randomly shuffle feature X j The value was recalculated and the error was recalculated. Take the average of the error increments on all trees as the feature X. j The importance of;

[0110] Step 34: Combinatorial Importance Scoring:

[0111] The results from the two methods are weighted and combined to obtain the final importance score:

[0112]

[0113] The weighting is based on empirical verification and emphasizes the stability of the Gini method in the evaluation of preservatives.

[0114] The implementation of step S4 specifically includes:

[0115] Step S41: Dynamic weight allocation strategy:

[0116] (1) Based on the feature importance results of the random forest, the normalized exponential function (Softmax) is used to enhance the discriminative power and stability of the weight allocation:

[0117]

[0118] in, This represents the dynamic weight of indicator j (color change index YI, decay index RI, safety index SI) in the comprehensive evaluation, and its value is calculated by the feature importance using the random forest model. This is obtained by normalization. The denominator is the sum of the importance of all indicators, ensuring that the total weight is 1.

[0119] (2) Considering the weight constraints of different application scenarios, and based on the differentiated needs of fruit and vegetable types and uses, segmented constraint rules are designed, as shown in Table 5:

[0120] Table 5 Examples of Segmentation Constraint Rules for Different Fruit and Vegetable Types

[0121] Fruit and vegetable types Basic constraints Dynamic adjustment Vegetables (such as leafy greens and root vegetables) <![CDATA[W SI ≥40%]]> <![CDATA[If the coefficient of variation of the decay index (RI) > 30%, then increase W RI to 35%]]> Berries (such as strawberries and blueberries) <![CDATA[W SI ≥60%]]> <![CDATA[When the daily change rate of the color change index (YI) > 5%, temporarily increase W YI to 25%]]> Processing uses <![CDATA[W RI ≤50%]]> <![CDATA[Simultaneously satisfy W SI : W RI ≤1.2 (safety first)]]>

[0122] (3) Dynamic Feedback and Iterative Optimization: A real-time monitoring mechanism is adopted to automatically recalculate feature importance and verify weight compliance after each batch of data is updated, triggering necessary automatic adjustments. At the same time, the weight adjustment trajectory is recorded through historical backtracking function, and the sliding window mean (such as the results of the last 5 calculations) is used to smooth sudden fluctuations and enhance weight stability. In addition, the system implementing this scheme provides a manual intervention interface, allowing experts to manually correct weight deviations and automatically record the reasons for corrections to be incorporated into subsequent model training, forming a closed-loop optimization process. This process combines automation and expert experience to ensure the dynamic, robust, and interpretable nature of weight allocation.

[0123] Step S42: Nonlinear comprehensive scoring:

[0124] This invention innovatively proposes a generalized average scoring model with shape parameters:

[0125]

[0126] Where a, b, and c are shape parameters (default a=1, b=1.2, c=1.5), used to amplify the impact on safety; k is an adjustment coefficient used to ensure TS∈[0,100]; + + =1.

[0127] Step S43: Overall rating level classification:

[0128] Based on the comprehensive score, the effectiveness of the preservative is divided into five levels:

[0129] Excellent (90-100 points, i.e., five stars): Excellent performance in all indicators, perfect safety score;

[0130] Good (80-89 points, or four stars): Major indicators meet the standards, with no obvious defects;

[0131] Above average (75-79 points, or three stars): Basically meets the requirements, but some indicators need improvement;

[0132] Below average (70-74 points, or two stars): Basically meets the requirements, but most indicators need improvement;

[0133] Pass (60-69 points, i.e., one star): Barely usable, but with several shortcomings;

[0134] Unqualified (<60 points, i.e., zero stars): Serious problems exist, application is not recommended.

[0135] Step S5: Comparison of the nonlinear comprehensive evaluation model and index evaluation method for preservatives. The joint verification and comparison results of the index evaluation scores and the nonlinear comprehensive evaluation model scores obtained in this embodiment are shown in Table 6. The yellowing index and decay index of leaf discs treated with different drugs after 4 days of storage showed significant differences, indicating that the preservation effects of different concentrations of different preservatives varied considerably. Therefore, these two indicators were selected as the basis for index scoring, and a lower total score indicates a better preservation effect. Figure 4 As shown, based on the evaluation results, it was found that 0.05 mM, 0.1 mM, or 0.2 mM chlorpyrifos and 0.05 mM 6-benzylaminopurine can inhibit the aging and spoilage of Chinese cabbage leaves. Considering their low cost and non-toxicity to humans, they can be regarded as ideal preservatives.

[0136] Table 6. Index scores and model scores for leaf discs treated with different drugs after 4 days of storage.

[0137]

[0138] As a means of verifying the reliability of the model evaluation, this embodiment uses the yellowing index and rot index as the index evaluation methods to statistically analyze the preservation effects of different preservatives and different concentrations on the circular tissue of Chinese cabbage leaves. The scores and specific characteristics of the yellowing index and rot index are shown in Table 7.

[0139] Table 7 Scoring and Specific Characteristics of Yellowing Index and Decay Index

[0140]

[0141] The process involves daily observation and statistical analysis of the yellowing and decay indices of the leaf discs of Chinese cabbage.

[0142] Yellowing Index =

[0143] Decay Index =

[0144] Overall score = Yellowing index 0.5+ Decay Index 0.5.

[0145] Based on the same inventive concept, this invention also provides a computer device, comprising: one or more processors, and a memory for storing one or more computer programs; the programs include program instructions, and the processor executes the program instructions stored in the memory. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, used to implement one or more instructions, specifically for loading and executing one or more instructions stored in a computer storage medium to implement the above-described method.

[0146] It should be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, performs the above-described method. This storage medium can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0147] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0148] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

[0149] This invention is not limited to the preferred embodiment described above. Anyone inspired by this invention can derive other forms of efficient screening methods and systems for fruit and vegetable preservatives based on multi-criteria decision-making and random forests. All equivalent variations and modifications made within the scope of the claims of this invention shall fall within the scope of this invention.

Claims

1. A method for screening fruit and vegetable preservatives, characterized in that: The color change index, decay index, and safety index of vegetables after preservation treatment were collected. The importance of each index is analyzed using machine learning algorithms to generate dynamic weighting coefficients. The dynamic weights and standardized indices are input into the nonlinear calculation model, and the contribution of each indicator is adjusted by pre-setting shape parameters to output a comprehensive score. Preservatives are selected based on a comprehensive scoring and grading system.

2. The method for screening fruit and vegetable preservatives according to claim 1, characterized in that: The shape parameters satisfy c>b≥a≥1, where a, b, and c correspond to the adjustment parameters of the color change index, decay index, and safety index, respectively.

3. The method for screening fruit and vegetable preservatives according to claim 1, characterized in that: The dynamic weighting coefficients are generated through the following steps: The importance of the combined features of the color change index, decay index, and safety index is calculated based on the combination Gini impurity reduction and permutation importance using the random forest algorithm. Weights are dynamically allocated using a normalized exponential function.

4. The method for screening fruit and vegetable preservatives according to claim 1, characterized in that: The safety index is predicted using the ADMET system and includes parameters for hepatotoxicity, mutagenicity, and acute toxicity.

5. The method for screening fruit and vegetable preservatives according to claim 1, characterized in that: The color change index and decay index are determined by the area ratio identified through image acquisition equipment and calculated according to a preset grading standard.

6. The method for screening fruit and vegetable preservatives according to claim 1, characterized in that: The fruit and vegetable tissue is a round slice with a diameter of 0.5-1.2 cm, and its preparation includes: Use a punch to cut leaves or fruit peel tissues of uniform size; Disinfect with 0.5% NaClO solution for 3 minutes, then blot dry the surface moisture; Each screening process includes at least 50 round tissue samples, with 3 replicates per preservative concentration.

7. The method for screening fruit and vegetable preservatives according to claim 1, characterized in that: The culture environment was constructed in the following manner: Place sterile round filter paper flat on the bottom of the petri dish and spray with sterile water to make the filter paper completely adhere to the bottom of the dish; The circular tissue samples were arranged in a 3×3 array on the filter paper.

8. The method for screening fruit and vegetable preservatives according to claim 3, characterized in that: When dynamically allocating weights, the lower limit for the safety weight of preservative constraints on berries is 60%, and the lower limit for constraints on vegetables is 40%.

9. The method for screening fruit and vegetable preservatives according to claim 3, characterized in that: The dynamic weight allocation process employs a real-time monitoring mechanism, automatically recalculating weights after data updates and smoothing fluctuations through a sliding window mean.

10. A fruit and vegetable preservative screening system, characterized in that, include: Data acquisition module: Configured to acquire the color change index, decay index, and safety index of vegetables and fruits after preservation treatment; Dynamic weight calculation engine: Analyzes the importance of each index through machine learning algorithms and generates dynamic weight coefficients; Nonlinear scoring processor: Receives dynamic weights and standardized indices, adjusts the contribution of indicators through preset shape parameters, and outputs a comprehensive score; Decision output interface: Based on comprehensive scoring and grading, preservatives are screened and recommended solutions are generated.

Images