Agricultural product quality and safety dual-dimension comprehensive evaluation method and system

Abstract

The application provides a kind of agricultural product quality safety double-dimension comprehensive evaluation method and system, method includes: adopt localization data set to build double-dimension evaluation model, calculate product inherent risk index, quantize enterprise score;Combined with product inherent risk index and enterprise score to generate comprehensive evaluation index;According to the comprehensive evaluation index and the preset risk classification threshold, the classification evaluation result is generated;With differentiating supervision strategy to generate high-frequency sampling plan, enterprise interview instruction and consumption guide star level identification.The application integrates data, synchronizes multiple types of data to form a unified set, realizes multi-source fusion to break fragmentation;The model is dynamic, the threshold and weight are adjusted dynamically to adapt to regional industry, seasonal consumption and enterprise size difference, and the classification evaluation is generated by combining real-time data correction and linking supervision strategy.This improves the evaluation accuracy and supervision efficiency, and strengthens the consumer's right to know.

Description

Technical Field

[0001] This invention relates to the field of agricultural product quality and safety technology, and in particular to a two-dimensional comprehensive evaluation method and system for agricultural product quality and safety. Background Technology

[0002] Current quality and safety assessments of edible agricultural products primarily employ single-dimensional testing and static threshold-based judgment models. For example, traditional methods often focus on quantitative laboratory analysis of heavy metal / pesticide / veterinary drug residues, obtaining pollutant concentration data through sampling and then setting fixed thresholds (e.g., lead ≤ 0.2 mg / kg, cadmium ≤ 0.1 mg / kg) based on national or local standards for compliance determination. Some regions supplement this with resident consumption surveys, but these often use static per capita consumption averages (e.g., leafy vegetables 0.3 kg / person·day), failing to consider seasonal fluctuations (e.g., a surge in leafy vegetable consumption in summer) or regional differences (e.g., differences in consumption habits between industrial and agricultural areas). Assessing enterprise management capabilities largely relies on on-site inspections and scoring, using a binary "qualified / unqualified" judgment or a simple weighted average score, lacking a dynamic correlation mechanism with product risk.

[0003] The existing technology has the following shortcomings: First, data silos are serious. Data such as heavy metal / veterinary drug testing, consumption statistics, and enterprise management capability assessment are scattered across different systems and lack an effective integration mechanism, resulting in fragmented assessment results. Second, the model is highly static. Threshold settings and weight allocation are mostly based on historical experience or general standards, which cannot dynamically adapt to regional industrial characteristics (such as high background values ​​of heavy metals in industrial areas), seasonal consumption patterns (such as the increased proportion of root and tuber consumption in winter), and differences in enterprise size (such as the low rate of testing equipment configuration in small and medium-sized enterprises), resulting in inaccurate assessments. Summary of the Invention

[0004] This invention aims to at least solve the technical problems existing in the prior art, and innovatively proposes a two-dimensional comprehensive evaluation method and system for agricultural product quality and safety.

[0005] To achieve the above-mentioned objectives of this invention, this invention provides a two-dimensional comprehensive evaluation method for agricultural product quality and safety, the method comprising: S1. Collect data on heavy metal and pesticide residue testing in the target urban area, per capita daily consumption data of residents, and survey data on enterprise management capabilities to form a local dataset; S2. Construct a two-dimensional evaluation model based on the localized dataset; S3. Based on the aforementioned dual-dimensional evaluation model, the inherent risk index of the product is calculated using the weighted toxicity equivalent scoring model, and the enterprise score is quantified based on the enterprise management capability index system. S4. Based on the product's inherent risk index and the company's score, a dynamically corrected comprehensive evaluation index is generated using a nonlinear correction function. S5. Based on the comprehensive evaluation index and the preset risk classification threshold, generate a classification evaluation result using the risk classification matrix; S6. Based on the aforementioned graded assessment results and real-time consumption data, generate high-frequency sampling plans, enterprise interview instructions, and consumer guidance star ratings using differentiated regulatory strategies.

[0006] On the other hand, the present invention also provides a two-dimensional comprehensive evaluation system for agricultural product quality and safety, the system comprising: processor; Memory used to store processor-executable instructions; The processor is configured to implement the dual-dimensional comprehensive evaluation method for agricultural product quality and safety when executing the executable instructions.

[0007] The beneficial effects of this invention are as follows: This invention effectively solves the core defects of data silos and static models in traditional assessments through localized data integration and dynamic model design. At the data integration level, it simultaneously collects heavy metal / veterinary drug residue detection data, resident consumption data, and enterprise management capability data to form a unified dataset. Through dimensional mapping, standardization processing, and nested dual-dimensional models, it achieves deep fusion of multi-source data, breaking the fragmentation dilemma caused by scattered data in traditional assessments. At the model dynamic level, through a dynamic weight adjustment mechanism, spatiotemporal dimension expansion, and nonlinear correction functions, the threshold setting and weight allocation can dynamically adapt to regional industrial characteristics (e.g., high heavy metal background values ​​in industrial areas), seasonal consumption patterns (e.g., increased consumption of root vegetables in winter), and differences in enterprise size (e.g., low equipment configuration rate for small and medium-sized enterprises). Combined with real-time consumption data superimposed and corrected by the risk grading matrix, it ultimately generates a graded assessment result and links it to differentiated regulatory strategies (high-frequency sampling, enterprise interviews, and star-rating consumer guidance). This not only improves the accuracy of assessments and the efficiency of regulatory resource allocation but also strengthens the protection of consumers' right to know, achieving a leap from "static threshold judgment" to "dynamic intelligent assessment." Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0008] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of a two-dimensional comprehensive evaluation method for agricultural product quality and safety according to the present invention. Detailed Implementation

[0009] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0010] Example 1 like Figure 1 As shown, a two-dimensional comprehensive evaluation method for agricultural product quality and safety is proposed, the method comprising: S1. Collect data on heavy metal and pesticide residue testing in the target urban area, per capita daily consumption data of residents, and survey data on enterprise management capabilities to form a local dataset; In step S1, it should be noted that the localized dataset covers all large-scale edible agricultural product production enterprises (including family farms and cooperatives) and major retail terminals within the target urban area. The data collection frequency is dynamically adjusted according to the characteristics of the product categories: perishable agricultural products such as leafy vegetables and berries are collected once a week, while root vegetables and grains are collected once a month. Resident consumption data is obtained by integrating the annual consumption report of the urban statistics bureau, sales data of large supermarkets, and order data from community group buying platforms, and the consumption structure weight is updated quarterly. The enterprise management capability survey data is collected through a dual-channel approach of "on-site verification + system integration." On-site verification focuses on recording key control points in the production process (such as pesticide use records and pre-harvest testing records). System integration obtains real-time data from the enterprise quality management system (QMS) and traceability platform through open API interfaces to ensure data timeliness (test data ≤ 72 hours, enterprise data ≤ 15 days). Quality control is performed simultaneously during the data collection process, including parallel sample verification of test data (relative deviation ≤ 10%), outlier removal of consumption data (using the 3σ criterion), and cross-validation of enterprise data.

[0011] S2. Construct a two-dimensional evaluation model based on the localized dataset; S3. Based on the aforementioned dual-dimensional evaluation model, the inherent risk index of the product is calculated using the weighted toxicity equivalent scoring model, and the enterprise score is quantified based on the enterprise management capability index system. S4. Based on the product's inherent risk index and the company's score, a dynamically corrected comprehensive evaluation index is generated using a nonlinear correction function. S5. Based on the comprehensive evaluation index and the preset risk classification threshold, generate a classification evaluation result using the risk classification matrix; S6. Based on the aforementioned graded assessment results and real-time consumption data, generate high-frequency sampling plans, enterprise interview instructions, and consumer guidance star ratings using differentiated regulatory strategies.

[0012] In step S6, it is necessary to explain in detail that the generation logic of the differentiated regulatory strategy adopts a dual-axis linkage mechanism of "risk level - consumer enthusiasm". For products with an extremely high risk level (comprehensive evaluation index ≥ 75), regardless of consumer enthusiasm, the highest level of regulatory response is triggered: sales of the batch of products are immediately suspended, full-chain traceability is initiated (tracing back to the planting base / breeding site and all downstream distribution nodes), and daily spot checks are implemented on the production enterprise for 3 months; simultaneously, a meeting instruction letter is generated for the enterprise, specifying rectification requirements (such as increasing the testing frequency to mandatory testing for every batch, changing high-risk input suppliers, etc.), and the meeting participants include the legal representative and quality manager of the enterprise. For products with a comprehensive evaluation index of 60-74, if the real-time consumer enthusiasm (sales volume in the past 7 days accounting for ≥ 20% of the market share of similar products) reaches the warning threshold, the next higher level of regulation is implemented: key spot checks are implemented 3 times a week, and the spot check items cover all relevant parameters of the seven-dimensional indicator system; a risk warning letter is sent to the enterprise, requiring the submission of a risk control plan within 5 working days, and a dedicated person is assigned to the factory to supervise the rectification. For medium-risk products with a comprehensive evaluation index of 40-59, the intensity of supervision will be dynamically adjusted according to consumer demand: when consumer demand is high (market share 10%-20%), random inspections will be conducted weekly; when consumer demand is low (market share <10%), random inspections will be conducted every two weeks. At the same time, suggestions for improving quality management will be provided to guide companies in optimizing weak areas (such as increasing employee training frequency to once a month and improving emergency response plans). For low-risk products with a comprehensive evaluation index <40, routine supervision will be implemented: random inspections will be conducted quarterly, focusing on verifying the operation of the traceability system. Companies that maintain a low-risk status for six consecutive months may be included in a "whitelist" for management, and the frequency of random inspections will be appropriately reduced. In terms of consumer guidance, star ratings are generated based on the tiered assessment results (5 stars for the lowest risk and 1 star for the highest risk). These ratings are displayed in real time through government websites, supermarket electronic screens, and third-party consumer platforms. The rating information includes the product name, manufacturer, comprehensive evaluation index, and explanation of the main risk points (e.g., "2 stars: cadmium residue is close to the threshold, it is recommended to consume no more than twice a month"). At the same time, SMS warnings are sent to consumers for high-risk products, suggesting alternative choices (e.g., "Leafy vegetables are currently at higher risk, root vegetables are recommended").

[0013] As an optional embodiment of the present invention, optionally, constructing a two-dimensional evaluation model based on the localized dataset in step S2 includes: S201. Based on the localized dataset, a data mapping result between the product risk dimension and the enterprise management capability dimension is generated using a dimensionality partitioning method; wherein, the product risk dimension is formed by integrating heavy metal and pesticide residue detection data and per capita daily consumption data of residents to form a pollutant exposure risk sub-dimensional, and the enterprise management capability dimension is formed by integrating enterprise management capability survey data to form a process control capability sub-dimensional. In step S201, it is necessary to explain in detail that the dimension division method adopts a dual-track mechanism of "hierarchical nesting + feature mapping". For the product risk dimension, firstly, the heavy metal detection data (such as cadmium, lead, arsenic, etc.) are converted into individual toxicity equivalents according to the "National Food Safety Standard Limits for Contaminants in Food" (e.g., toxicity equivalent factor TEF=1.0 for cadmium, TEF=0.01 for lead). Then, the pollutant exposure is calculated by combining the per capita daily consumption data of residents (exposure = detection concentration × consumption). Finally, the basic score of the pollutant exposure risk sub-dimension is generated by weighted summation. The weight allocation is dynamically adjusted according to the proportion of health impact of each pollutant in the local epidemiological data (e.g., the weight of cadmium exposure in industrial areas is set to 0.35, and the weight of organophosphorus pesticides in agricultural areas is set to 0.4). For the enterprise management capability dimension, the survey data was broken down into 5 primary indicators (production standard compliance, testing capability configuration, traceability system integrity, personnel training frequency, and emergency response plan) and 18 secondary indicators (such as production standard compliance including "pesticide use record completeness rate" and "pre-harvest self-inspection rate"). The Analytic Hierarchy Process (AHP) was used to determine the indicator weights (testing capability configuration has the highest weight, accounting for 0.3), and a 0-100 point scale was used for quantification scoring to form the original score matrix for the process control capability sub-dimensional. Cross-dimensional correlation verification was performed simultaneously during the data mapping process. For example, Pearson correlation analysis was conducted between the enterprise's "advanced testing equipment" indicator and the "pesticide and veterinary drug residue detection rate" of the product risk dimension. For indicator pairs with an absolute correlation coefficient > 0.6, a linkage correction rule was established to ensure the inherent logical consistency of the two-dimensional data.

[0014] S202. Based on the data mapping results, a standardized dimensional dataset is generated using data preprocessing and standardization methods; wherein, the data preprocessing includes cleaning detection data, correcting consumption data, and quantifying enterprise capability indicators, and the standardization methods include Min-Max normalization and KNN interpolation to fill in missing values; In step S202, it is necessary to explain in detail that the data preprocessing stage adopts differentiated processing strategies for different types of data: the detection data cleaning is achieved by constructing an outlier identification model, and the detection results that deviate from the historical mean by 3 times the standard deviation (such as a sudden increase of 5 times in the pesticide residue value of a batch of leafy vegetables) are marked. Combined with the laboratory quality control records (such as instrument calibration certificates and operator qualifications), it is determined whether they are real outliers. If they are confirmed to be systematic errors, they are replaced with the mean of parallel samples of the same batch; the consumption data correction introduces a seasonal fluctuation coefficient. For example, the consumption in the two weeks before the Spring Festival is adjusted upward by 15% according to historical data, and downward by 8% during extreme weather (three consecutive days of rain). The random fluctuation of daily consumption data is eliminated by the moving average method. The quantification of enterprise capability indicators adopts a dual approach of "quantification of qualitative indicators + standardization of quantitative indicators": For qualitative indicators such as "completeness of traceability system", three levels of values ​​are assigned: "complete traceability (100 points), partial traceability (60 points), and no traceability (0 points)"; For quantitative indicators such as "number of testing equipment", the industry benchmark value is first determined by the quartile method (e.g., the benchmark value for testing equipment of large enterprises is 8 units), and then the score is calculated according to the ratio of the actual value to the benchmark value (e.g., if an enterprise is equipped with 12 units of equipment, the score is 12 / 8×100=150 points, and the upper limit is set at 120 points). In the standardization process, Min-Max normalization compresses all indicators to the [0,1] interval. The formula is: Standardized value = (Original value - Minimum value) / (Maximum value - Minimum value), where the maximum and minimum values ​​are based on local 3-year historical data and national industry standard thresholds (e.g., the maximum value of heavy metal cadmium is taken as 1.5 times the national standard limit). When using KNN interpolation to fill missing values, feature neighbors are selected according to the data type. For example, when enterprise testing capability data is missing, testing equipment configuration data of enterprises of the same size and category are matched first. Samples with a missing rate of more than 20% are removed as a whole to ensure the validity of the dataset.

[0015] S203. Based on the standardized dimensional dataset, a framework structure for a two-dimensional evaluation model is generated using a model architecture design method; wherein, the framework structure includes a matrix framework of X-axis product risk dimension and Y-axis enterprise management capability dimension, as well as nested pollutant exposure risk sub-model and process control capability sub-model. In step S2003, it is necessary to explain in detail that the model architecture design adopts a composite structure of "matrix base + nested sub-model". The X-axis product risk dimension takes the contaminant exposure risk sub-model as the core, and the risk level increases from left to right along the axis (0-100 points), specifically divided into 5 risk intervals: 0-20 points (extremely low exposure risk), 21-40 points (low exposure risk), 41-60 points (medium exposure risk), 61-80 points (high exposure risk), and 81-100 points (extremely high exposure risk). The interval boundary values ​​are dynamically calibrated based on local food safety incident statistics over the past 5 years (e.g., the score corresponding to the contaminant exposure value that historically caused mass incidents is set to 85 points). The Y-axis enterprise management capability dimension is based on the process control capability sub-model, with capability levels increasing from bottom to top along the axis (0-100 points). It is also divided into 5 capability intervals: 0-20 points (extremely weak management capability), 21-40 points (weak management capability), 41-60 points (medium management capability), 61-80 points (strong management capability), and 81-100 points (extremely strong management capability). The interval division references the enterprise risk level classification standards in the "Measures for Risk Classification Management of Food Production Enterprises," and takes into account the actual management level distribution characteristics of local enterprises (e.g., when small and medium-sized enterprises account for 60%, the lower limit of the medium management capability interval is adjusted to 35 points). Each cell in the matrix framework represents a specific combination of "product risk - enterprise capability," for example, (X=75 points, Y=30 points) corresponds to a high-risk combination of "high product risk - weak enterprise management." The pollutant exposure risk sub-model is nested on the X-axis and includes a pollutant identification module (building a local pollutant database based on the national standard GB2762-2022, including 32 key pollutants such as cadmium, lead, and chlorpyrifos), an exposure calculation module (integrating daily consumption and detection concentration, considering consumption differences among different populations (adults / children / pregnant women), and setting population-specific correction coefficients), and a toxicity weighting module (using the WHO-recommended baseline dose method (BMD) to calculate the health risk weight of each pollutant). The process control capability sub-model is nested on the Y-axis and includes an indicator acquisition module (connecting to the enterprise's QMS system to obtain key control point data in real time), a weight calculation module (using the entropy weight method to dynamically adjust the weights of 18 secondary indicators; for example, when a certain type of enterprise experiences continuous traceability system failures, the weight of the "traceability system integrity" indicator is increased from 0.15 to 0.25), and a capability scoring module (generating a comprehensive enterprise capability score through fuzzy comprehensive evaluation). The framework of the two-dimensional model is implemented in Python, and the matrix operation core is built using the TensorFlow framework. It supports parallel evaluation processing of 1,000+ samples per second, ensuring that it still maintains sub-second response speed when the local data volume reaches 100,000.

[0016] S204. Based on the aforementioned framework, a final two-dimensional evaluation model is generated using dynamic parameter configuration and model validation optimization methods; wherein, the dynamic parameter configuration includes a dynamic weight adjustment mechanism and spatiotemporal dimension expansion, and the model validation optimization includes historical data backtesting, sensitivity analysis, and expert calibration mechanisms.

[0017] In step S204, it is necessary to explain in detail that the dynamic parameter configuration achieves multi-scenario adaptation by constructing a "three-dimensional parameter pool": In the time dimension, a quarterly weight adjustment window is set to adjust the weight ratio of each pollutant in the pollutant exposure risk sub-dimensional according to the agricultural production cycle (such as spring plowing and autumn harvest). For example, during the vigorous growth period of leafy vegetables in summer, the weight of organophosphorus pesticides is increased from 0.2 to 0.3; In the spatial dimension, a geographical correction coefficient is set for different regional characteristics (such as the area around industrial zones and water source protection areas), and heavy metal detection data of enterprises around industrial zones are assigned a risk weight of 1.2 times; In the category dimension, for special categories such as infant complementary food ingredients, the weight of the "pollutant detection frequency" indicator is increased by 20%. The dynamic weight adjustment mechanism adopts a dual engine of "data-driven + rule-triggered". When the detection rate of a certain type of pollutant exceeds 1.5 times the historical average for three consecutive months, the emergency weight adjustment procedure is automatically initiated and an early warning notification is sent to the system administrator. The spatiotemporal dimension expansion supports the integration of GIS geographic information data, refining the assessment unit from the city level to the street / township level. It optimizes the risk spatial distribution model by combining land use type data (such as cultivated land and orchards). For example, for areas with excessive background heavy metal values ​​in the soil, the "soil improvement measure implementation rate" indicator is added to the enterprise management capability assessment. Model validation and optimization first involves backtesting with historical data. 120,000 data points on edible agricultural product testing from the past three years and corresponding quality and safety incident records are selected. The prediction results of the dual-dimensional assessment model are compared with the actual occurrence of events, calculating accuracy (≥85%), precision (≥80%), and recall (≥75%). Features are extracted from misjudged samples (such as low-risk events predicted by the model but actual occurrences of excessive levels), and the sub-model algorithm is optimized. Sensitivity analysis uses the controlled variable method to test the impact of fluctuations in key parameters (such as toxicity equivalent factor and consumption weight) on the assessment results. When parameters change within ±20%, the impact is assessed. During operation, the fluctuation range of the comprehensive evaluation index is required to be ≤10% to ensure model stability. An expert calibration mechanism is established, comprising a 15-member expert committee of food safety regulators, agricultural technology experts, and public health scholars. A calibration meeting is held every six months to review the high-risk product list and enterprise management capability scoring standards output by the model. The model parameters are manually fine-tuned by incorporating the latest scientific research findings (such as research on the toxicity of emerging pollutants) and feedback from regulatory practice (such as tracking data on the effectiveness of enterprise rectification). For example, high-risk pesticide varieties identified by expert consensus are included in the priority evaluation sequence, and their detection data are given a 1.3-fold weight in the exposure calculation. Through dynamic parameter configuration and multi-dimensional verification and optimization, the final two-dimensional evaluation model can adapt to changes in localized data characteristics and regulatory needs.

[0018] As an optional embodiment of the present invention, optionally, in step S3, based on the dual-dimensional evaluation model, the inherent risk index of the product is calculated using a weighted toxicity equivalent scoring model, and the enterprise score is quantified based on the enterprise management capability index system, including: S301. Based on the product risk dimension in the dual-dimensional assessment model, a product inherent risk index is generated using a weighted toxicity equivalent scoring model. The weighted toxicity equivalent scoring model integrates heavy metal and pesticide / veterinary drug residue detection data and per capita daily consumption data, and uses toxicity equivalent coefficients and consumption weight coefficients for weighted calculation to generate a quantitative product inherent risk index that includes a pollutant exposure risk correction process. In step S301, it is necessary to explain in detail that the core algorithm of the weighted toxicity equivalent scoring model adopts a two-level calculation architecture of "weighted summation of individual pollutant risks + exposure correction". First, for each edible agricultural product sample, the detection data of heavy metals (cadmium, lead, arsenic, mercury, etc.) and pesticides and veterinary drugs (organophosphates, carbamates, pyrethroids, etc.) are extracted, and the limit values ​​of each pollutant are determined with reference to the "National Food Safety Standard Limits for Contaminants in Food" and the "National Food Safety Standard Maximum Residue Limits for Pesticides in Food". For pollutants whose detected concentration exceeds the limit value, they are directly marked as "seriously exceeding the limit", and their individual risk score is assigned a base score of 100 points; for pollutants that do not exceed the limit value, their "exceedance multiple ratio" is calculated as the initial quantitative basis for individual risk.

[0019] Secondly, a toxicity equivalent factor (TEF) is introduced to standardize the toxicity intensity of different pollutants. The TEF values ​​are based on the toxicity classification standards of the International Agency for Research on Cancer and the World Health Organization, and are adjusted locally using local epidemiological data. For example, the TEF is set at 1.0 (baseline value) for cadmium, 0.01 for lead, 0.1 for arsenic, 0.05 for chlorpyrifos (organophosphorus pesticide), and 0.2 for carbofuran (carbamate pesticide).

[0020] Secondly, exposure risk correction is performed by integrating per capita daily consumption data (CR). Consumption data is derived from local residents' dietary consumption survey reports, stratified by age group (children, adults, pregnant women), gender, and seasonal characteristics to generate a matrix of average daily consumption for specific population groups. For a certain product category (such as leafy greens), its consumption weight coefficient Wc is dynamically adjusted based on the product's proportion in the local residents' dietary structure. For example, if leafy greens account for 40% of total vegetable consumption, then Wc = 0.4. The exposure correction coefficient (Ec) is calculated as: Ec = 1 + (CR / CRavg - 1) × 0.3, where CRavg is the average consumption of this product among the population. When the actual consumption is higher than the average, Ec > 1, amplifying the risk score; conversely, it is reduced. For example, if a children's group consumes an average of 80g of strawberries per day, higher than the population average of 50g, then Ec = 1 + (80 / 50 - 1) × 0.3 = 1.18, meaning the risk score for this group's strawberry product needs to be multiplied by a correction coefficient of 1.18.

[0021] Finally, a weighted summation is used to obtain the product's inherent risk index. Simultaneously, risk level thresholds are set: an inherent risk index <20 indicates "extremely low risk," 20-40 indicates "low risk," 41-60 indicates "medium risk," 61-80 indicates "high risk," and >80 indicates "extremely high risk." This model, by deeply coupling pollutant toxicity, exposure levels, and consumption characteristics, achieves a refined and dynamic quantitative assessment of the product's inherent risk.

[0022] S302. Based on the enterprise management capability dimension in the dual-dimensional evaluation model, a quantitative score for the enterprise is generated using the enterprise management capability indicator system. The enterprise management capability indicator system integrates indicator data on production process standard compliance, quality management system certification, and testing equipment configuration, and uses a weighted average calculation with a dynamic weight adjustment mechanism and a five-level scoring system to generate a quantitative score result for the enterprise that includes the scores of each indicator and the total score.

[0023] The expression for generating a quantitative score for a company using the corporate governance capability indicator system is as follows: in, This represents the enterprise's quantitative score (total score, ranging from 0 to 100, calculated based on a weighted average of seven indicators). Indicates the first The item has a five-level rating (1-5 points, based on on-site survey data, such as "completely non-compliant" = 1 point, "completely compliant" = 5 points). Indicates the first The dynamic weighting coefficients of the indicators (adjusted based on enterprise size / regional industry characteristics, such as the weighting of traceability capabilities for SMEs) =0.25), Indicates the first The compliance correction factor for each indicator (dynamically adjusted based on the quality management system certification results, such as ISO22000 certified companies) =1.2); Detailed Seven-Dimensional Indicator System =1: Production process standard compliance rate (e.g., GAP certification implementation rate) =2: Quality management system certification status (e.g., ISO22000 / HACCP certification status) =3: Level of testing equipment configuration (e.g., gas chromatograph configuration rate) =4: Traceability system coverage (e.g., the proportion of blockchain traceability applications) =5: Completeness of emergency response plans (e.g., completeness of response procedures for sudden pollution incidents) =6: Employee training frequency (e.g., annual food safety training hours) =7: Environmental management compliance (such as the compliance rate of wastewater and exhaust gas emissions).

[0024] In step S302, it is necessary to explain in detail that the quantification process of the enterprise management capability indicator system first requires the collection of raw data for the seven-dimensional indicators. Production process standard compliance is assessed through on-site verification of whether the enterprise strictly implements Good Agricultural Practices (GAP), such as soil improvement records, irrigation water monitoring frequency, and pesticide use registration and filing for planting enterprises. The percentage of compliant items is converted to a score of 1-5, with 5 points for complete compliance and 1 point for serious non-compliance. Quality management system certification status is determined based on the certification type and validity period: ISO 22000 certification within its validity period earns 5 points, HACCP certification only earns 4 points, application for certification earns 3 points, and no certification earns 1 point. Testing equipment configuration level (i=3) assesses whether the enterprise has testing equipment commensurate with its production scale. For example, vegetable production enterprises need to have a pesticide residue rapid tester (basic configuration earns 3 points), a liquid chromatograph (advanced configuration earns 4 points), and a gas chromatography-mass spectrometry (GC-MS) system (advanced configuration earns 5 points); no equipment at all earns 1 point. The traceability system coverage rate is calculated based on the coverage of traceability information for each stage of a company's products from planting / breeding to sales. A 100% coverage rate with full blockchain traceability earns 5 points; only origin traceability earns 3 points; no traceability system earns 1 point. The emergency response plan completeness review examines whether the plan includes emergency scenarios such as pollutant exceedances and epidemic prevention and control. A plan with all necessary elements and at least two drills per year earns 5 points; a plan that exists but has not been drilled earns 3 points; no plan earns 1 point. Employee training frequency is calculated based on the total annual food safety training time. ≥40 hours of training per employee per year earns 5 points; 20-40 hours earns 3 points; <20 hours earns 1 point. Environmental management compliance requires reviewing the company's wastewater and exhaust gas emission test reports for the past year. All reports must meet standards with no exceedances to earn 5 points; one minor exceedance earns 3 points; multiple exceedances or penalties from environmental protection departments earn 1 point.

[0025] The determination of dynamic weighting coefficients needs to be combined with the size of the enterprise and the characteristics of the regional industry. For small and medium-sized enterprises (with fewer than 100 employees), the weight of traceability system coverage is increased to 0.25, which is higher than that of large enterprises (0.15), because small and medium-sized enterprises generally have weak traceability capabilities and need to be given special attention. For enterprises around industrial parks, the weight of environmental management compliance is increased from 0.10 to 0.18 to strengthen the control and assessment of their production environment risks.

[0026] As an optional embodiment of the present invention, optionally, the expression for generating the product's inherent risk index using the weighted toxicity equivalent scoring model in step S301 is as follows: in, This indicates the inherent risk index of the product (quantified value, ranging from 0 to 100). Indicates the total number of pollutants. Indicates the first Measured concentrations of various pollutants (based on localized detection data). Indicates the first The toxicity equivalent coefficients of the pollutants (based on authoritative standards such as WHO / JECFA, e.g., lead = 1.0, cadmium = 2.5). This represents the consumption weighting coefficient (dynamically adjusted based on category / season / region, e.g., leafy vegetables = 1.2 in winter). Indicates the first The average daily consumption of agricultural products by residents corresponding to each pollutant (based on localized consumption data).

[0027] As an optional embodiment of the present invention, optionally, in step S4, generating a dynamically corrected comprehensive evaluation index based on the product's inherent risk index and the enterprise score using a nonlinear correction function includes: S401. Based on the inherent risk index of the product and the quantitative score of the enterprise, a nonlinear function is constructed; wherein, the nonlinear function adopts a hyperbolic tangent and exponential decay composite form, integrating regional industrial characteristics, seasonal factors and enterprise size parameters. As an optional embodiment of the present invention, the expression of the nonlinear correction function is optionally: in, This represents the dynamically adjusted comprehensive evaluation index (range 0-100, after Min-Max normalization). This represents the scaling factor (adjusted based on regional industry characteristics, such as industrial zones). =1.2, agricultural area =0.9), Represents the hyperbolic tangent function. This represents the risk amplification factor (adjusted for seasonal factors, such as leafy vegetables in summer). =1.3, winter rhizomes =0.8), This indicates the inherent risk index of the product. This represents the enterprise capability adjustment coefficient (adjusted based on enterprise size, such as for small and medium-sized enterprises). =0.2, large enterprises =0.1), This represents the company's quantitative score. Indicates the decay rate (based on historical data weighting adjustment, such as...) =0.15 corresponds to a decay to e-0.9≈0.4 after 6 months. This represents a time variable (in months, reflecting the difference between the data collection time and the current time). This represents the time decay compensation coefficient (adjusted based on data timeliness, such as data from the last 3 months). =5.0).

[0028] In step S401, it should be noted that the nonlinear function achieves dynamic correction of the comprehensive evaluation index through multi-parameter coupling. The scaling factor is adjusted according to regional industrial characteristics. For example, in areas with high industrial pollution risk, the scaling factor is set to 1.2 to amplify the impact of potential risks; while in major agricultural production areas, the scaling factor is set to 0.9 to weaken non-productive pollution factors. The hyperbolic tangent function tanh is used for nonlinear mapping of the product's inherent risk index. When the inherent risk index value is low (e.g., <30 points), tanh (inherent risk index) increases slowly, avoiding distortion of evaluation results due to small fluctuations in low-risk products. When the inherent risk index value enters the medium-to-high risk range (e.g., >50 points), the slope of the function curve increases, strengthening the sensitivity to high-risk products. The risk amplification factor is dynamically set based on seasonal consumption characteristics. For example, in summer, leafy vegetables are more susceptible to pests and diseases, leading to increased pesticide use; the risk amplification factor is set to 1.3 to increase the risk warning level. In winter, root and tuber products have a longer storage period, and pollutants degrade more fully; the risk amplification factor is set to 0.8 to reduce over-assessment risk. The enterprise capability adjustment coefficient is adjusted differently according to the enterprise size. Small and medium-sized enterprises (SMEs) have larger fluctuations in management capabilities due to resource constraints. The enterprise capability adjustment coefficient is set to 0.2 to make the correction effect of the enterprise score on the comprehensive index more significant. Large enterprises have relatively complete management systems, and the enterprise capability adjustment coefficient is set to 0.1 to maintain the stability of the evaluation.

[0029] S402. Based on the nonlinear function, a set of regionally adapted correction coefficients is generated using a dynamic parameter configuration mechanism; wherein, the set of correction coefficients is determined through historical data backtesting and sensitivity analysis, and includes industrial / agricultural zone weight adjustment parameters, summer / winter consumption amplification coefficients, and compensation factors for small and medium-sized enterprises / large enterprises; As explained in step S402, the dynamic parameter configuration mechanism first constructs an initial candidate set of correction coefficients through backtesting of historical data. Specifically, it collects local data on the quality and safety of edible agricultural products, enterprise regulatory records, and consumption characteristics from the past 3-5 years, grouping them into three dimensions: regional type (industrial zone, agricultural zone), season (summer, winter), and enterprise size (small and medium-sized enterprises, large enterprises). For each group of data, the controlled variable method is used to test the impact of different combinations of correction coefficients on the comprehensive evaluation index. For example, in the industrial zone scenario, the scaling factor is gradually adjusted from 1.0 to 1.5, and the fit between the comprehensive evaluation index and the actual occurrence rate of safety incidents is observed. The coefficient value that maximizes the prediction accuracy is selected as the initial value. Sensitivity analysis is performed using Monte Carlo simulation, applying a random perturbation of ±20% to each parameter (such as the scaling factor and risk amplification factor), calculating the fluctuation range of the comprehensive evaluation index, and prioritizing the optimization of key parameters that have a significant impact on the results (fluctuation range > 5%). For parameters with a smaller impact (fluctuation range < 2%), industry default values ​​are used to simplify the calculation. For example, in the analysis of leafy vegetables, it was found that the risk amplification coefficient was much more sensitive than the time decay compensation coefficient. Therefore, the risk amplification coefficient for summer leafy vegetables was selected as a key control parameter. By comparing the correlation between the pesticide residue exceedance rate of summer leafy vegetables in different years and the risk amplification coefficient, 1.3 was ultimately determined to be the optimal value. After the modified coefficient set is generated, a quarterly update mechanism needs to be established. This mechanism should incorporate the latest regional pollution survey data (such as newly added pollution sources in industrial areas), changes in consumption trends (such as the increased proportion of root and tuber consumption in winter), and adjustments in enterprise scale and structure (such as the increase in the number of small and medium-sized enterprises) to dynamically calibrate the coefficient set, ensuring that it always remains adapted to local conditions.

[0030] S403. Based on the set of correction coefficients, calculate the risk-capability nonlinear correction value using the interaction term; wherein, the interaction term amplifies the risk of high-risk-low-capability scenarios through the product term of the product risk index and the enterprise score, and uses the hyperbolic tangent function to constrain the boundary of the correction value. In step S403, it should be noted that the core function of the interaction item is to capture the dynamic coupling relationship between the inherent risks of the product and the enterprise's control capabilities, avoiding misjudgments of risk caused by simple weighting. Specifically, in the calculation, the inherent risk index of the product is first constructed. Enterprise quantitative scores The product term ( × When the product is in a high-risk situation (such as...) (>60 points) and the company's capabilities are low (e.g., When the score is less than 3, this product term will significantly amplify the risk contribution value, for example... =70、 When =2, the interaction item value is 140, which is much higher than in low-risk-high-capability scenarios (such as...). =30、 =150 when =5), highlighting the safety hazards of the high-risk-low-capability combination through this nonlinear amplification effect. Simultaneously, to prevent the interaction term's value from expanding infinitely and causing the comprehensive evaluation index to overflow the reasonable range, a hyperbolic tangent function tanh is introduced to constrain the product term's boundaries, compressing the interaction term result to the [-1,1] interval before multiplying it with the correction coefficient set. For example, when... × When tanh = 200, tanh(200) ≈ 1, and the interaction term correction value is mainly determined by parameters such as scaling factor and risk amplification factor; when × When = 50, tanh(50) ≈ 1, still maintaining a high contribution; while when = 50, tanh(50) ≈ 1, it still maintains ... × When =10, tanh(10)≈1, ensuring that the interaction terms do not excessively suppress the comprehensive index in low-risk scenarios. Furthermore, the interaction terms incorporate a time decay factor. For enterprise scores or pollutant concentration data collected more than 6 months ago, the decay coefficient e(-λt) is calculated using the time variable t and the decay rate λ (e.g., λ=0.15). For example, the decay coefficient for data from 12 months ago is e(-1.8)≈0.165. At this time, the interaction terms ( × ×e(-λt)) will automatically reduce the weight of historical data, increasing the sensitivity of the evaluation results to the current security status.

[0031] S404. Based on the nonlinear correction value and the time decay term, the final comprehensive evaluation index is generated using the normalization method; wherein, the time decay term is dynamically weighted by the exponential function to reduce the weight of historical data, and the normalization process uses the Min-Max method to map the correction result to the 0-100 range, and generates a standardized index report containing details of the correction process and parameter configuration instructions.

[0032] In step S404, it should be noted that the core function of the time decay term is to dynamically adjust the weight of historical data, ensuring that the evaluation results prioritize reflecting the recent quality and safety status. Specifically, the old data is weighted down using the exponential function e(-λt), where λ is the decay rate (e.g., λ=0.15) and t is the difference between the data collection time and the current time (in months). For time-sensitive parameters such as pesticide residue detection values ​​and recent rectification records of enterprises, the time decay effect is particularly important, as it can avoid evaluation lag caused by applying historical data indiscriminately.

[0033] The normalization process uses the Min-Max method to map the nonlinear correction values ​​to a standard range of 0-100 points.

[0034] The standardized index report should include three core parts: First, a detailed explanation of the correction process, listing the calculation process of the product's inherent risk index (including the toxicity equivalent of each pollutant and the consumption weight), the composition of the enterprise score (including the scores and weights of indicators such as traceability system coverage and environmental management compliance), the values ​​of each parameter of the nonlinear function (such as scaling factor and risk amplification factor), and the calculation results of the interaction term; Second, a parameter configuration description, explaining the impact logic of regional industrial characteristics (such as industrial zone / agricultural zone), seasonal factors (such as summer / winter), and enterprise size (small and medium-sized enterprises / large enterprises) on the correction factor, for example, "The scaling factor of the industrial zone was adjusted to 1.2 this quarter because the background value of heavy metals in the surrounding soil increased due to the addition of new chemical enterprises"; Third, an interpretation of the index results, providing targeted governance suggestions to regulatory authorities through risk level classification (such as 0-20 points for low risk, 21-50 points for medium risk, and 51-100 points for high risk) and the ranking of key influencing factors (such as "The current high risk is mainly due to excessive pesticide residues (contribution of 62%) and insufficient enterprise traceability system coverage (contribution of 28%)"). The report should also include a statement on the timeliness of the data, indicating the collection time of each original data point and its actual weight after attenuation, to ensure the transparency and traceability of the evaluation process.

[0035] As an optional embodiment of the present invention, optionally, in step S5, generating a classification assessment result based on the comprehensive evaluation index and the preset risk classification threshold using a risk classification matrix includes: S501. Based on the comprehensive evaluation index and the preset risk classification threshold, a classification matrix framework is generated; wherein, the classification matrix framework includes an X-axis comprehensive evaluation index range (0-100), a Y-axis risk level dimension (low / medium / high / extremely high), and a nested threshold dynamic adjustment mechanism (e.g., raising the high-risk threshold for industrial zones to...). ≥65) and policy linkage rules (e.g., high risk corresponds to "high-frequency spot checks + enterprise interviews"); In step S501, it is necessary to explain in detail that the construction of the hierarchical matrix framework uses the comprehensive evaluation index as the core horizontal axis, dividing the index range of 0-100 points into four continuous and non-overlapping sub-ranges, corresponding to the low, medium, high, and extremely high risk levels on the Y-axis, respectively. Specifically, the division criteria are: 0-20 points correspond to low risk, 21-50 points to medium risk, 51-75 points to high risk, and 76-100 points to extremely high risk. This basic threshold can be adaptively adjusted according to regional characteristics through a dynamic threshold adjustment mechanism. For example, in areas with high industrial pollution risk, to enhance the sensitivity of risk warnings, the high-risk threshold is raised from the default 51 points to 65 points; that is, when the comprehensive evaluation index reaches 65 points or above, it is determined to be a high-risk level. In major agricultural production areas, if the overall quality and safety level is high, the high-risk threshold can be appropriately lowered to 45 points to avoid overreacting to minor risks. The Y-axis risk level dimension not only represents the degree of risk but also incorporates corresponding strategy linkage rules for each level, forming a three-in-one mapping relationship of "index range - risk level - control strategy". For example, a low-risk level corresponds to the basic regulatory strategy of "routine spot checks + annual review"; a medium-risk level triggers a mechanism of "quarterly special spot checks + enterprise self-inspection reports"; a high-risk level initiates enhanced control measures of "high-frequency spot checks (at least once a month) + interviews with enterprise leaders + time-limited rectification of potential risks"; and an extremely high-risk level will directly adopt emergency response measures of "suspension of sales + comprehensive investigation + tracing the source + administrative penalties". In addition, the tiered matrix framework also supports a dynamic transition and decline mechanism for risk levels. When the comprehensive evaluation index drops from the high-risk range to the medium-risk range in two consecutive assessments and remains stable for more than 3 months, its risk level can be downgraded and the control strategy adjusted accordingly. Conversely, if the index of a medium-risk product rises significantly and exceeds the high-risk threshold within an assessment cycle, the control measures will be upgraded immediately to ensure the timeliness and accuracy of risk control.

[0036] S502. Based on the aforementioned hierarchical matrix framework and real-time consumption data, a hierarchical assessment result is generated using a dynamic mapping method. The dynamic mapping method determines the initial risk level by directly comparing the comprehensive evaluation index with the preset risk level threshold, and generates the final hierarchical assessment result by overlaying and correcting the real-time consumption data (e.g., if the consumption exceeds the average by 20%, the risk level is increased by one level). In step S502, it is necessary to explain in detail that the dynamic mapping method first performs an initial risk level determination. The comprehensive evaluation index generated in step S404 is directly compared with the preset basic threshold in the grading matrix framework. For example, if the comprehensive evaluation index is 45 points, compared with the default threshold (21-50 points for medium risk), it is initially determined to be at a medium risk level. Subsequently, real-time consumption data is introduced for overlay correction to reflect the amplification effect of the actual consumption scale of the product on the risk. Specifically, regional consumption data of the edible agricultural product within the current assessment period (e.g., the last 7 days) is collected, and its ratio to the average consumption of the last 6 months is calculated. When the actual consumption exceeds the average by 20%, the risk level adjustment mechanism is triggered. For example, if the comprehensive evaluation index of a certain leafy vegetable is 50 points (at the upper limit of medium risk), and its recent consumption reaches 130% of the historical average, its risk level is upgraded from medium risk to high risk; conversely, if the consumption is below 50% of the average, and the comprehensive evaluation index is at the lower limit of medium risk (e.g., 25 points), the risk level can be considered for downgrading to low risk. Real-time consumption data is collected from various sources, including transaction records from farmers' markets, sales data from supermarkets, order information from e-commerce platforms, and statistics from community group buying. This data is aggregated and dynamically updated in real time via a data interface. For situations where consumption data is missing or fluctuates abnormally (such as sudden changes in consumption behavior due to public health emergencies), the system will automatically activate a smoothing algorithm, using the average consumption over the past three assessment periods as a reference benchmark to avoid interference from single-period data anomalies in the correction results. After dynamic mapping, the generated tiered assessment results not only include the final risk level but also indicate the initial level, the basis for consumption data correction, and the specific adjustment range. For example, "Initial level: Medium risk (50 points); Due to consumption exceeding the average by 30%, upgraded by one level; Final level: High risk," ensuring the interpretability of the risk level adjustment process.

[0037] Example 2 A dual-dimensional comprehensive evaluation system for agricultural product quality and safety includes a processor and a memory for storing executable instructions. The processor is configured to implement a dual-dimensional comprehensive evaluation method for agricultural product quality and safety when executing the executable instructions. It should be noted that the computer device includes a processor and memory, and may also include one or more of a multimedia component, an input / output (I / O) interface, and a communication component. The processor controls the overall operation of the computer device, completing all or part of the steps of the dual-dimensional comprehensive evaluation method for agricultural product quality and safety. The memory stores various types of data to support device operation and can be implemented by volatile or non-volatile storage devices or combinations thereof, such as SRAM, EEPROM, etc. The multimedia component includes a screen (such as a touchscreen) and an audio component. The audio component has a microphone to receive external audio signals and at least one speaker to output audio signals. The I / O interface provides an interface for the processor and other interface modules (such as a keyboard, mouse, buttons, etc.). The communication component is used for wired or wireless communication between devices. Wireless communication includes Wi-Fi, Bluetooth, etc., and the communication component includes a Wi-Fi module, etc. As a preferred embodiment, the computer device can be implemented using electronic components such as ASICs and DSPs to execute the dual-dimensional comprehensive evaluation method for agricultural product quality and safety.

[0038] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A two-dimensional comprehensive evaluation method for agricultural product quality and safety, characterized in that, The method includes: S1. Collect data on heavy metal and pesticide residue testing in the target urban area, per capita daily consumption data of residents, and survey data on enterprise management capabilities to form a local dataset; S2. Construct a two-dimensional evaluation model based on the localized dataset; S3. Based on the aforementioned dual-dimensional evaluation model, the inherent risk index of the product is calculated using the weighted toxicity equivalent scoring model, and the enterprise score is quantified based on the enterprise management capability index system. S4. Based on the product's inherent risk index and the company's score, a dynamically corrected comprehensive evaluation index is generated using a nonlinear correction function. S5. Based on the comprehensive evaluation index and the preset risk classification threshold, generate a classification evaluation result using the risk classification matrix; S6. Based on the aforementioned graded assessment results and real-time consumption data, generate high-frequency sampling plans, enterprise interview instructions, and consumer guidance star ratings using differentiated regulatory strategies.

2. The method for comprehensive evaluation of agricultural product quality and safety from two dimensions as described in claim 1, characterized in that, Step S2, which involves constructing a two-dimensional evaluation model based on the localized dataset, includes: S201. Based on the localized dataset, a data mapping result between the product risk dimension and the enterprise management capability dimension is generated using a dimensionality partitioning method; wherein, the product risk dimension is formed by integrating heavy metal and pesticide residue detection data and per capita daily consumption data of residents to form a pollutant exposure risk sub-dimensional, and the enterprise management capability dimension is formed by integrating enterprise management capability survey data to form a process control capability sub-dimensional. S202. Based on the data mapping results, a standardized dimensional dataset is generated using data preprocessing and standardization methods; wherein, the data preprocessing includes cleaning detection data, correcting consumption data, and quantifying enterprise capability indicators, and the standardization methods include Min-Max normalization and KNN interpolation to fill in missing values; S203. Based on the standardized dimensional dataset, a framework structure for a two-dimensional evaluation model is generated using a model architecture design method; wherein, the framework structure includes a matrix framework of X-axis product risk dimension and Y-axis enterprise management capability dimension, as well as nested pollutant exposure risk sub-model and process control capability sub-model. S204. Based on the aforementioned framework, a final two-dimensional evaluation model is generated using dynamic parameter configuration and model validation optimization methods; wherein, the dynamic parameter configuration includes a dynamic weight adjustment mechanism and spatiotemporal dimension expansion, and the model validation optimization includes historical data backtesting, sensitivity analysis, and expert calibration mechanisms.

3. The method for comprehensive evaluation of agricultural product quality and safety from two dimensions as described in claim 1, characterized in that, In step S3, based on the aforementioned dual-dimensional assessment model, the inherent risk index of the product is calculated using a weighted toxicity equivalent scoring model, and the enterprise score is quantified based on the enterprise management capability indicator system, including: S301. Based on the product risk dimension in the dual-dimensional assessment model, a product inherent risk index is generated using a weighted toxicity equivalent scoring model. The weighted toxicity equivalent scoring model integrates heavy metal and pesticide / veterinary drug residue detection data and per capita daily consumption data, and uses toxicity equivalent coefficients and consumption weight coefficients for weighted calculation to generate a quantitative product inherent risk index that includes a pollutant exposure risk correction process. S302. Based on the enterprise management capability dimension in the dual-dimensional evaluation model, a quantitative score for the enterprise is generated using the enterprise management capability indicator system. The enterprise management capability indicator system integrates indicator data on production process standard compliance, quality management system certification, and testing equipment configuration, and uses a weighted average calculation with a dynamic weight adjustment mechanism and a five-level scoring system to generate a quantitative score result for the enterprise that includes the scores of each indicator and the total score.

4. The dual-dimensional comprehensive evaluation method for agricultural product quality and safety as described in claim 3, characterized in that, The expression for generating the product's inherent risk index using the weighted toxicity equivalent scoring model in step S301 is as follows: in, This indicates the inherent risk index of the product. Indicates the total number of pollutants. Indicates the first The measured concentrations of the pollutants, Indicates the first The toxicity equivalent coefficient of the pollutant, This represents the consumption weighting coefficient. Indicates the first The average daily consumption of agricultural products by residents corresponding to each type of pollutant.

5. The method for comprehensive evaluation of agricultural product quality and safety from two dimensions as described in claim 1, characterized in that, In step S4, based on the product's inherent risk index and the company's score, a dynamically corrected comprehensive evaluation index is generated using a nonlinear correction function, including: S401. Based on the inherent risk index of the product and the quantitative score of the enterprise, a nonlinear function is constructed; wherein, the nonlinear function adopts a hyperbolic tangent and exponential decay composite form, integrating regional industrial characteristics, seasonal factors and enterprise size parameters. S402. Based on the nonlinear function, a set of regionally adapted correction coefficients is generated using a dynamic parameter configuration mechanism; wherein, the set of correction coefficients is determined through historical data backtesting and sensitivity analysis, and includes industrial / agricultural zone weight adjustment parameters, summer / winter consumption amplification coefficients, and compensation factors for small and medium-sized enterprises / large enterprises; S403. Based on the set of correction coefficients, calculate the risk-capability nonlinear correction value using the interaction term; S404. Based on the nonlinear correction value and the time decay term, the final comprehensive evaluation index is generated using the normalization method; wherein, the time decay term is dynamically weighted using an exponential function to reduce the weight of historical data.

6. A dual-dimensional comprehensive evaluation method for agricultural product quality and safety as described in claim 1 or 5, characterized in that, The expression for the nonlinear correction function is: in, This represents the dynamically adjusted comprehensive evaluation index. Indicates the scaling factor. Represents the hyperbolic tangent function. This represents the risk amplification factor. This indicates the inherent risk index of the product. This represents the enterprise capability adjustment coefficient. This represents the company's quantitative score. Indicates the attenuation rate. Represents a time variable. This represents the time decay compensation coefficient.

7. The method for comprehensive evaluation of agricultural product quality and safety from two dimensions as described in claim 1, characterized in that, In step S5, based on the comprehensive evaluation index and the preset risk grading threshold, the risk grading matrix is ​​used to generate a grading assessment result, including: S501. Based on the comprehensive evaluation index and the preset risk classification threshold, generate a classification matrix framework; S502. Based on the aforementioned hierarchical matrix framework and real-time consumption data, a hierarchical assessment result is generated using a dynamic mapping method; wherein, the dynamic mapping method determines the initial risk level by directly comparing the comprehensive evaluation index with the preset risk level threshold, and generates the final hierarchical assessment result by superimposing and correcting the real-time consumption data.

8. A comprehensive evaluation system for agricultural product quality and safety based on two dimensions, characterized in that, The system includes: processor; Memory used to store processor-executable instructions; The processor is configured to implement the dual-dimensional comprehensive evaluation method for agricultural product quality and safety as described in any one of claims 1 to 7 when executing the executable instructions.

Images