A food sampling optimization method and system for antibiotic residue traceability

Abstract

The application discloses a food sampling optimization method and system for antibiotic residue traceability, wherein the method collects food sampling, food consumption and antibiotic related data, estimates the pollution probability and pollution concentration of antibiotics in food, simultaneously applies a Bayesian network model to predict the flow direction and probability of antibiotics in a 'production end-consumer end' supply chain, then calculates the pollution quantity caused by antibiotics and the average daily intake of people in different regions, finally constructs a target function and constraint condition, and solves the optimization problem by using a mixed integer programming method to determine the optimal sampling scheme under different resource allocation situations. The application aims to realize accurate evaluation of antibiotic residue risk and efficient allocation of resources.

Description

Technical Field

[0001] This invention relates to the field of food safety technology, and in particular to an optimized method and system for food sampling inspection oriented towards antibiotic residue traceability. Background Technology

[0002] As the livestock industry develops towards large-scale and intensive farming, animal output per unit area and per unit time has significantly increased. However, high-density farming also creates favorable conditions for the occurrence and spread of infectious diseases. To prevent and control disease risks in the farming process, farmers and production enterprises widely use antibiotics and other veterinary drugs in feed additives and treatments. Improper use of antibiotics, inadequate dosage control, or insufficient withdrawal period management can easily lead to antibiotic residues in animal-derived foods. These residues are often highly concealed and difficult to detect through sensory means. They can accumulate and spread throughout the food supply chain without timely identification, potentially posing health risks to consumers and causing economic losses such as product recalls, trade restrictions, and damage to brand reputation.

[0003] Currently, to address antibiotic resistance, routine sampling and testing of animal-derived foods throughout the entire supply chain are conducted. However, the food supply chain comprises thousands of nodes, with food sources covering diverse regions. Comprehensive sampling is financially and operationally infeasible, and sampling resources are generally limited. Furthermore, due to differences in the scale of production and operation entities, diversified distribution channels, and uneven levels of information technology, many foods still suffer from incomplete traceability chains, missing records of key nodes, or ineffective information sharing during actual distribution. Traceable and untraceable samples coexist for a long time. This traceability gap not only weakens the regulatory authorities' ability to accurately trace problematic batches and assess risks but also increases the difficulty of identifying high-risk varieties, high-risk regions, and high-risk distribution links when developing sampling plans, making it difficult to optimally allocate limited sampling resources within the complex food supply chain. Therefore, given the reality of incomplete traceability information and limited sampling resources, how to scientifically identify high-risk sources and flows within the complex food supply chain and achieve precise allocation and optimization of antibiotic residue sampling resources has become an urgent problem to be solved.

[0004] Existing invention patent CN114819529A discloses a system and method for monitoring and tracing the quality of live pigs based on blockchain and the Internet of Things. This system consists of four layers: a perception layer, a transmission layer, a data layer, and an application layer. It includes nine functional modules: enterprise information management, product information management, production and processing management, monitoring and early warning management, QR code management, statistical analysis management, traceability query management, product transaction management, and system management. Utilizing RFID technology, BDS, and sensor technology, it constructs a real-time monitoring system for the storage and transportation links of the live pig supply chain, providing data collection and analysis methods for real-time monitoring, anomaly warning, information recording, querying, and traceability functions during live pig transportation and storage. While it can achieve transparent supervision of the entire pig supply chain and has significant application value in decision-making regarding agricultural product quality and safety management, information traceability and response, and the recall of problematic products, it suffers from several drawbacks. Each level relies on external technology combinations (such as RFID, BDS, and blockchain), posing compatibility risks. Furthermore, it depends on specialized equipment (such as spectral analysis instruments and BDS positioning) and full-node blockchain deployment, resulting in high costs. Monitoring and early warning rely on data transmission to a central node, leading to latency risks. Additionally, the blockchain's private blockchain design makes it difficult to integrate with external systems and hinders effective scalability. In conclusion, while its technological integration achieves supply chain transparency and real-time control, it cannot achieve dynamic configuration with limited sampling resources. Summary of the Invention

[0005] The main objective of this invention is to provide an optimized method and system for food sampling inspection oriented towards antibiotic residue traceability, aiming to solve the problem of minimizing food safety risks in real-world scenarios with limited sampling resources, incomplete traceability information, and dynamic changes in the supply chain.

[0006] To achieve the above objectives, this invention provides an optimized method for food sampling inspection oriented towards antibiotic residue traceability, the method comprising the following steps:

[0007] Step 1: Obtain relevant characteristic data of the target food and characteristic data of the antibiotic. The relevant characteristic data of the target food includes food safety sampling data, food consumption data and population-related data. The characteristic data of the antibiotic includes microbiologically acceptable daily intake, toxicologically acceptable daily intake and detection limit data in food.

[0008] Step 2: Determine the probability and average concentration of antibiotic contamination in the target food based on the target food safety sampling data obtained in Step 1.

[0009] Step 3: Construct a Bayesian network model based on the target food safety sampling data obtained in Step 1, and use Monte Carlo simulation to perform probabilistic reasoning to obtain the probability of the flow of each antibiotic in the production-consumption supply chain.

[0010] Step 4: Based on the probability of antibiotic contamination and average concentration in the target food in Step 2, the flow probability in Step 3, and the food consumption and population of the target food in the preset area obtained in Step 1, calculate the number of foods contaminated with antibiotics and the average daily intake of antibiotics by people in different regions and the national population through food.

[0011] Step 5: Establish a resource allocation scenario based on the traceable and non-traceable production locations of the samples. The objective function is to maximize the number of foods contaminated with antibiotics. An optimization model is created with the constraints that the number of sampled batches does not exceed the historical maximum number of sampled samples of the target food in the preset area, and the average daily intake of antibiotics by residents nationwide through food does not exceed the microbiologically acceptable daily intake and the toxicologically acceptable daily intake. The optimal sample size and sampling frequency for the target food are determined using mixed integer programming.

[0012] Optionally, in step 1, the relevant characteristic data of the target food are cleaned and standardized to unify the units and numerical ranges and ensure data consistency and comparability. Data cleaning includes identifying and processing missing values, outliers and duplicate entries. Data standardization uses the Z-Score method to transform the data into a standard form with a mean of 0 and a standard deviation of 1.

[0013] Optionally, in step 2, the formula for calculating the pollution probability is:

[0014] , ;

[0015] Among them, PC j,k and PC k NC represents the probability of antibiotic residue k in food at the consumption location j and the country level, respectively. j,k Indicates the number of food samples in consumption location j contaminated with a specific antibiotic residue k; NT j This represents the total number of food samples randomly inspected in consumption location j;

[0016] The formula for calculating the average concentration is as follows:

[0017] , ;

[0018] and These represent the average concentrations of antibiotic k in the corresponding contaminated egg samples at the consumption location j and the country level, respectively. Indicates the concentration of antibiotic K in a specific contaminated sample C; NC j,k and NC k These represent the number of samples contaminated with antibiotic k at the consumption location j and the country level, respectively.

[0019] Optionally, step 3, the step of constructing the Bayesian network model, specifically includes:

[0020] Based on the food safety sampling data obtained in step 1, a dataset of non-compliant samples is created. A Bayesian network model is constructed using food category, antibiotic type, production time, consumption time, production location, consumption location, and supply chain node characteristics as network nodes. The belief propagation method is then used to obtain the probability distribution of the flow of each antibiotic in the production-consumption supply chain.

[0021] Optionally, in step 4, the formula for calculating the average daily intake of antibiotics is:

[0022] ;

[0023] in, This represents the average estimated daily intake of antibiotic k in consumption location j. DI represents the average concentration of antibiotic k in consumption location j. j BW represents the average daily food intake per person in consumption location j, and BW represents the population weight.

[0024] Optionally, in step 5, the objective function is:

[0025] ;

[0026] Among them, PI k,i CE represents the probability of detecting contamination of antibiotic k from production site i. k,i The quantity of food produced at location i that was contaminated with antibiotic k, nb k,i The number of samples is the decision variable.

[0027] Optionally, in step 1, the antibiotic includes at least one of enrofloxacin, ofloxacin, chloramphenicol, ciprofloxacin, metronidazole, doxycycline, florfenicol, furazolidone, metronidazole, sulfamethoxazole, and thiamphenicol.

[0028] Optionally, in step 5, the resource allocation scenarios include simultaneously allocating sampled resources between traceable and non-traceable food sources, allocating only to traceable food sources, allocating only to non-traceable food sources, or allocating resources by antibiotic type.

[0029] In addition, to achieve the above objectives, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any of the above claims.

[0030] In addition, to achieve the above objectives, the present invention also provides a food sampling optimization system for antibiotic residue traceability, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0031] Beneficial effects:

[0032] (1) By using Bayesian networks and mixed integer programming, the sampling strategy can prioritize the coverage of high-risk sources (such as specific production sites or antibiotic types), thereby maximizing the detection of contaminated food under the same resources. The optimized strategy can detect additional contaminated samples and reduce daily intake, directly improving the targeting and effectiveness of supervision.

[0033] (2) By comparing daily intake assessments with safety thresholds, the system can identify areas of excessive risk and types of antibiotics, and optimize sampling accordingly. This helps with early intervention and reduces population exposure. The document mentions that the "degree of health risk reduction" can be used as a benefit evaluation indicator, reflecting the value of public health protection.

[0034] (3) The multi-scenario model allows regulators to flexibly allocate resources based on traceability levels (e.g., traceable vs. non-traceable), avoiding resource waste. Furthermore, benefit assessments show that the optimization strategy can achieve higher efficiency under budget constraints, is applicable to different regulatory conditions, and improves the sustainable use of resources.

[0035] (4) The integrated system provides structured outputs (such as pollution probability parameters and flow distribution), enabling the sampling plan to be based on data rather than experience, thus reducing subjective bias.

[0036] (5) Through probabilistic traceability analysis, the method can identify weak links in the supply chain, indirectly promote the completion of traceability information, and in the long run, help to build a more transparent food supply chain, thereby effectively promoting the improvement of the traceability system. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating an optimized food sampling method for antibiotic residue traceability according to the present invention.

[0039] Figure 2 This is a statistical chart showing the probability of antibiotic contamination in poultry eggs from different consumption areas (provinces) in the examples;

[0040] Figure 3 This is a statistical chart showing the probability of antibiotic contamination in poultry eggs nationwide, as illustrated in the example.

[0041] Figure 4 Box-line chart showing the average antibiotic concentration in poultry eggs from various consumption regions (provinces) from 2015 to 2022;

[0042] Figure 5 A statistical graph of the marginal probabilities of the variables predicted by the Bayesian network;

[0043] Figure 6 A statistical graph showing the posterior probability of antibiotic transmission in untraceable poultry eggs in various consumption areas, predicted by a Bayesian network.

[0044] Figure 7 A statistical graph showing the posterior probability of antibiotic transmission in traceable poultry eggs across various consumption areas, predicted by a Bayesian network.

[0045] Figure 8 A statistical chart showing the number of poultry eggs contaminated with antibiotics from traceable and untraceable sources nationwide;

[0046] Figure 9 A statistical chart showing the spread of contaminated poultry eggs across different consumption areas;

[0047] Figure 10-11 This chart shows the distribution of average daily intake of various antibiotics in different consumption areas and the statistics of average daily intake of various antibiotics nationwide.

[0048] Figure 12-13 This chart shows the optimized statistics on the number of batches sampled for food safety testing of various antibiotics, the optimized statistics on the number of poultry eggs contaminated with antibiotics that can be detected, and the optimized statistics on the average daily intake that can be reduced.

[0049] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0050] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0051] See Figure 1-2 This invention provides an optimized food sampling strategy for antibiotic residue traceability, the method comprising the following steps:

[0052] Step 1: Data Acquisition and Standardization. The acquired raw data includes relevant characteristic data of food and characteristic data of antibiotics. The relevant characteristic data of food includes food safety sampling inspection data, food consumption data, and population-related data. The food safety sampling inspection data covers the name, category, time, location, pass / fail assessment, and information on excessive levels of harmful substances detected during food sampling. The food consumption data includes the average annual consumption of food in each region. The population-related data includes the number of permanent residents in each region at the end of each year. Generally, it is necessary to collect a wide range of multi-source data, including food safety sampling inspection records published by the 31 provincial-level market supervision administrations (data for Hong Kong, Macao, and Taiwan are currently unavailable), covering product name, sampling time, location, pass / fail assessment, and information on excessive levels of harmful substances; and average annual food consumption data for each province provided by the National Bureau of Statistics of China and the China Health and Nutrition Survey. The data includes specific information on antibiotics, such as the names and classifications of major contaminating antibiotics like enrofloxacin, ofloxacin, and chloramphenicol; maximum residue limits (MRLs) in the corresponding foods; limits of detection (LODs) for the detection methods; and health baseline parameters such as the toxicologically acceptable daily intake (TADI) and the microbiologically acceptable daily intake (MADI). This antibiotic data can be derived from the assessment results of the WHO / FAO Expert Committee, relevant Chinese national standards, and authoritative literature. It is compiled into an antibiotic characteristics database that can be matched with food sampling data and consumption data, providing a foundation for subsequent daily intake calculations and health risk assessments.

[0053] Step 2: Calculation of antibiotic residue contamination probability and concentration.

[0054] Specifically, based on the data obtained in step 1, the food sampling data are grouped and statistically analyzed according to consumption location and antibiotic type. Frequency estimation methods are used to calculate the conditional contamination probability of each antibiotic within the consumption location and the overall contamination probability in all samples. Furthermore, based on the detected concentrations of non-compliant samples, the regional average concentration and the national overall average concentration are calculated to form structured contamination probability and concentration parameters. Here, contamination probability refers to the probability that the antibiotic concentration in food exceeds the national legal standard. Contaminated samples refer to food samples where the antibiotic concentration exceeds the national legal standard. Simultaneously, a dataset of non-compliant samples is constructed based on the collected food safety sampling data and used as input to a Bayesian network model to estimate the flow and probability of each antibiotic in the target food along the "production location-consumption location" supply chain.

[0055] Details are as follows:

[0056] S2.1 Regional Grouping Statistical Sub-step: Group and count the total number of samples and the number of contaminated samples by consumption location and antibiotic type, obtain the total number of samples and the number of samples contaminated by each antibiotic for each consumption location and antibiotic type, and record the detection concentration information of the corresponding contaminated samples.

[0057] S2.2 Contamination Probability Estimation Sub-step: Calculate the conditional contamination probability and the overall contamination probability. Specifically, based on the grouping statistics, the frequency estimation method is used to calculate the probability of antibiotic k contamination in samples from consumption location j (i.e., the conditional contamination probability under the condition that "the sample comes from consumption location j"), which is the ratio of the number of samples contaminated with this antibiotic in that region to the total number of samples sampled in that region; at the same time, the overall contamination probability of antibiotic k in all samples is calculated, which is the ratio of the number of samples contaminated with this antibiotic in all consumption locations to the total number of samples sampled in all consumption locations.

[0058] The formula for calculating the probability of regional pollution is: ,(1);

[0059] The formula for calculating the overall pollution probability is: (2);

[0060] in, and These represent the contamination probability of antibiotic residue k in food at the consumption location j and the national level, respectively. This indicates the number of food samples in consumption location j that are contaminated with a specific antibiotic residue k; This represents the total number of food samples sampled from consumption location j, where j is the consumption location index (from 1 to J), representing the specific consumption location region, and J represents the total number of consumption locations.

[0061] S2.3 Concentration Parameter Estimation Sub-step: Calculate the regional average concentration and the national average concentration based on the detected concentrations of non-compliant samples. Specifically, based on the detected concentration data of contaminated samples in each consumption location, and grouped by consumption location and antibiotic type, calculate the average concentration of samples contaminated with a certain antibiotic in each consumption location and / or other statistics that can represent the residual level of that region; and at the national level, calculate the overall average concentration of the antibiotic in all contaminated samples and other representative statistics. Use the obtained contamination probability and concentration parameters at the consumption location and national levels as input parameters for subsequent daily dietary intake assessment and sampling strategy optimization steps.

[0062] The formulas for calculating the average concentration of antibiotic k in the sample from consumption site j, and the overall average concentration of antibiotic k in all samples are as follows:

[0063] The formula for calculating the regional average concentration is: (3);

[0064] The formula for calculating the overall average concentration is: (4);

[0065] in, and These represent the average concentrations of antibiotic k in the corresponding contaminated egg samples at the consumption location j and the country level, respectively. This indicates the concentration of antibiotic K in a specific contaminated (c) sample; and These represent the number of egg samples contaminated with antibiotic k at the consumption location j and national levels, respectively; C represents the upper limit of the number of specific contaminated samples that can be accumulated. For provinces lacking antibiotic concentration data, the maximum residue limit (MRL) or limit of detection (LOD) for that antibiotic in food will be used to supplement the data.

[0066] Step 3: Predicting the probability of antibiotic residue flow based on Bayesian networks. Based on the food safety sampling data collected in Step 1, a dataset of non-compliant samples is created. A Bayesian network model is constructed using food category, antibiotic type, production time, consumption time, production location, consumption location, and supply chain node characteristics as network nodes. The belief propagation method is used to obtain the probability distribution of each antibiotic's flow along the "production location-consumption location" supply chain. Specifically, a conditional probability table is estimated using a parameter learning method based on historical sampling data, and Monte Carlo simulation is used for probabilistic inference to obtain the probability distribution of each antibiotic's flow along the production location-consumption location supply chain.

[0067] Details are as follows:

[0068] S3.1 Non-compliant sample screening and modeling dataset construction sub-steps: Screen non-compliant samples from food sampling data and extract relevant fields; specifically, screen the sampling samples that are judged to be non-compliant from the food sampling data, and extract fields such as food category, type of contamination, production time, consumption time, production location, consumption location and supply chain node characteristics to form a non-compliant sample dataset for modeling.

[0069] S3.2 Bayesian Network Structure Construction Sub-step: A structure learning algorithm is used to construct the network structure. Food type, contamination type, production time, consumption time, production location, consumption location, and supply chain node characteristics are used as network nodes. A Bayesian network structure learning algorithm and a tree-based Bayesian network structure learning algorithm are employed to autonomously learn the directed connections between nodes on the non-compliant sample dataset, constructing a Bayesian network structure to characterize the source and flow of antibiotic residues under non-compliant sample conditions. The Tree-Augmented Naive Bayes algorithm is used for the Bayesian network structure learning algorithm.

[0070] S3.3 Conditional Probability Table Parameter Learning Sub-step: Parameter learning is performed using the non-compliant sample dataset; specifically, the conditional probability of each node in the Bayesian network is learned using the non-compliant sample dataset under the given parent node value. A Bayesian estimation method based on frequency statistics and incorporating Dirichlet prior is used to smoothly estimate each conditional probability, thereby obtaining the conditional probability table of each node under the condition of excessive antibiotic residues.

[0071] S3.4 Evidence Setting and Reasoning Scenario Construction Sub-step: Set up observational evidence for reasoning; specifically, based on the type of antibiotic to be analyzed and the consumption area, select variables such as antibiotic type, sampling location or sample non-compliance results as observational evidence, input the observational evidence into the trained Bayesian network model, and construct a reasoning scenario for tracing the source and predicting the flow of antibiotic residues.

[0072] S3.5 Flow Probability Calculation and Pollution Source Tracing Sub-step: Monte Carlo simulation is used to calculate the posterior probability distribution. Specifically, in the constructed inference scenario, based on the conditional probability table, the Monte Carlo simulation inference algorithm is used to perform multiple random samplings on the Bayesian network to trace the source of antibiotic pollution, and to calculate the posterior probability distribution of each antibiotic under different "production site-consumption site" combinations under the condition that the sample is unqualified, in preparation for the subsequent assessment of daily dietary intake.

[0073] Among them, Bayesian networks use Bayes' theorem to estimate the value of another event (e.g., ...). In the event of a certain occurrence (e.g.) The probability of something occurring is called the posterior probability. .

[0074] (5);

[0075] in, It is an event The prior probability, In known events Under the conditions conditional probability, yes The probability. A Bayesian network model includes: (1) a set of variables (nodes) (1) a set of directed arcs between variables; (2) a finite set of states for each discrete variable; and (3) a set of conditional and unconditional probabilities. If there exists a node... To the node An arc is called a node. For nodes The parent node, the node For nodes The child nodes.

[0076] The following formula presents the Bayesian network formula, which is the joint probability distribution of all variables. This formula is given by the product of all conditional probability tables specified in the Bayesian network: (6);

[0077] in, In variables Known parent node events under conditions The probability of.

[0078] In this step, all non-compliant samples from the food data are used to build a Bayesian network model. The data is automatically and randomly divided into an 80% training set and a 20% test set. Antibiotic types are used as the target variable, and Tree-Augmented NaiveBayes (TAN) are employed for structure learning. After the structure is determined, the parameters are estimated using the Bayes method, resulting in three Bayesian network models. The predicted antibiotic types are then compared with the true categories using the test set, and accuracy, sensitivity, and specificity are calculated. The next step, based on the known BN structure and marginal probabilities, uses the Belief Propagation method to infer the posterior probability of egg contamination given specific evidence conditions (formula below). Finally, Monte Carlo simulation is used to calculate the probability of each antibiotic's flow along the "production-consumption" supply chain.

[0079] ;

[0080] in, It is the posterior probability that antibiotic k is detected at consumption location j and originates from production source i (traceable or untraceable source); It is the combined probability that food at consumption location j carries individual antibiotic k from production source i; It is the marginal probability that food is contaminated with antibiotic K; This represents the marginal probability that antibiotic residues originate from production source i. Let represent the conditional probability of detecting antibiotic k in food given an animal species l; Let represent the conditional probability that food is contaminated at consumption location j, given an animal species l. Let i represent the conditional probability that the contaminated food originated from animal l, given that the food production source is i.

[0081] Step 4: Calculation of Contamination Quantity and Assessment of Daily Intake. Based on the aforementioned contamination probability, concentration, food consumption, and flow probability, calculate the daily antibiotic intake (traceable and untraceable) and the quantity of antibiotic-contaminated food (traceable and untraceable) for different population groups in various regions, and analyze the contamination transmission pathways.

[0082] Details are as follows:

[0083] S4.1 Dataset Construction Sub-step: Match flow probability, contamination probability, consumption data, etc. to supply chain combinations; Specifically, based on the obtained supply chain flow of each antibiotic, match flow probability, actual contamination probability, consumption data, concentration data, and weight data to each supply chain to construct a dataset for contamination quantity calculation and daily intake assessment.

[0084] S4.2 Unit Contamination Quantity Estimation Sub-step: Estimate the contamination quantity using "production location, consumption location, and antibiotic type" as units; specifically, in the daily intake assessment dataset, using "production location, consumption location, and antibiotic type" as basic units, and combining the corresponding flow probability, actual contamination probability, and consumption information, estimate the quantity of food expected to be contaminated with antibiotics in each unit, which is used to reflect the quantity level of potentially contaminated food under different combinations.

[0085] For example, by combining the overall probability of antibiotic contamination in food sampling data, the flow probability predicted by Bayesian networks, per capita consumption, and population size in each region, we can calculate the quantity of food contaminated with antibiotic k from production site i to consumption site j; the quantity of food contaminated with antibiotic k produced from production site i; and the quantity of food contaminated with antibiotic k detected at consumption site j.

[0086] ;

[0087] ;

[0088] ;

[0089] in, Total food consumption in consumption location j; CE k,i,j The quantity of food potentially contaminated by antibiotics k at the production source i in the consumption location j. The quantity of potentially contaminated food produced in region i nationwide due to antibiotic k. This refers to the amount of food potentially contaminated due to antibiotic K.

[0090] S4.3 Sub-step for calculating daily intake of unit: Calculate the contribution of each unit to the daily intake of the population in the consumption area; specifically, using parameters related to flow probability, residual concentration, consumption and population weight in the same dataset, and referring to the recommended methods of the Food and Agriculture Organization of the United Nations and the World Health Organization on dietary exposure assessment, the contribution of each unit to the daily dietary intake of the corresponding population in the consumption area is calculated independently, with "production area, consumption area, antibiotic type" as the intake unit, to obtain the impact of different combinations.

[0091] S4.4 Provincial and National Summary and Risk Assessment Sub-step: Summarize the results and compare them with acceptable daily intake to analyze the risks. Specifically, based on the calculation results of unit pollution quantity and unit daily dietary intake, the contributions of pollution quantity and intake from different production sources and different types of antibiotics within the same consumption area are summed to obtain the daily dietary intake (traceable and untraceable) and expected pollution quantity (traceable and untraceable) of each antibiotic and all antibiotics for the population in each province and city; and on this basis, the results of each province and city are further summarized to obtain the assessment results of daily dietary intake and pollution quantity for the national population.

[0092] The average daily intake of antibiotics from a certain food source is calculated using the following formula for people in different regions.

[0093] The formula for calculating daily intake is: ,(11)

[0094] ;

[0095] ;

[0096] in, The average estimated daily intake of antibiotic K in consumption location J (µg / kgbw / day). The average estimated daily intake (µg / kgbw / day) of antibiotic k from production site i at the national level. The average estimated daily intake of antibiotic K at the national level (µg / kgbw / day). The average concentration level (μg / kg) of antibiotic K in consumption location J; DI j BW represents the average daily food intake per person in the consumption area (kg / day / person); BW represents the weight of an adult in China (60kg / person).

[0097] After calculating the national average estimated daily intake, this step introduces two safety thresholds to compare with the calculated average estimated daily intake in order to determine the risk of antibiotic residues to human health. The two thresholds are the Microbiological Acceptable Daily Intake (MADI) and the Toxicological Acceptable Daily Intake (TADI). The TADI uses data on the acceptable daily intake of various antibiotics published in China's national food safety standards and literature, while the MADI will be re-estimated using the following formula.

[0098] ;

[0099] Among them, MIC 50 The minimum concentration required to inhibit 50% of the most sensitive human symbiotic microbiota is indicated; MCC refers to the mass of colonic contents (220g); EF indicates the effective fraction of the oral dose reaching the colonic microbiota (20%); BW is the body weight of a Chinese adult (60kg / person).

[0100] Step 5, Sampling strategy optimization steps based on mixed integer programming.

[0101] S5.1 Scenario Division and Parameter Preparation Sub-steps: Construct four types of sampling resource allocation scenarios and retrieve relevant parameters; Specifically, based on the traceability of antibiotic-contaminated food, construct multiple sampling resource allocation scenarios, including: scenarios where sampling resources are allocated simultaneously between traceable and non-traceable source foods, scenarios where resources are allocated only to traceable source foods, scenarios where resources are allocated only to non-traceable source foods, and scenarios where traceable and non-traceable source foods are considered comprehensively based on antibiotic type as the allocation unit; and retrieve the antibiotic residue contamination probability and residue concentration parameters obtained above, as well as the daily dietary exposure assessment results, as input parameters for the sampling optimization model.

[0102] S5.2 Decision Variables and Mixed Integer Programming Model Construction Sub-steps: Establish a model with the number of samples as the decision variable; specifically, under each sampling scenario, the number of samples divided according to antibiotic type and traceability status is used as the decision variable, and the objective function is to maximize the number of detected antibiotic-contaminated foods. Under the constraints of total sample size, health risk, and minimum sample size, a corresponding mixed integer programming model is established.

[0103] S5.3 Model Solving and Sampling Scheme Determination Sub-steps: The optimal sampling configuration is obtained by using a solution algorithm; specifically, a mixed integer programming solution algorithm is used to solve the model under each sampling scenario to obtain the optimal sampling sample quantity and sampling frequency for each antibiotic under different traceability states, thus forming a food sampling strategy configuration scheme under multiple scenarios.

[0104] S5.4 Regulatory Benefit Assessment Sub-step: Calculate the additional number of contaminated samples detected and the reduction in health risk. Under each sampling scenario, after the sampling strategy is formulated, this step uses the same set of daily intake and contamination quantity assessment methods to calculate the additional number of contaminated samples that can be detected by the optimized strategy under the same total sample size, and the resulting reduction in EDI. Based on this, the degree of health risk reduction is determined, serving as the comprehensive evaluation result of the regulatory benefits of each sampling strategy.

[0105] For example, based on the risk-based sampling requirements of food safety regulatory authorities, the goal is to maximize the number of contaminated eggs identified through sampling strategies, while being constrained by current sampling capacity limitations and acceptable health risk thresholds. Sampling resources are allocated under four combined scenarios. The model's constraint is the maximum permissible number of sampling batches per year, derived from historical data. For each scenario, cost-effectiveness is evaluated based on the number of contaminated eggs under reduced EDI (Electronic Data Interference) at different resource levels. Specifically, the formula is as follows:

[0106] The objective function is: (15);

[0107] Detection probability (16);

[0108] (17);

[0109] Constraints include: total sample size constraint (18);

[0110] in, ; ; ;

[0111] ; ;

[0112] in, The effectiveness of the sampling strategy is indicated by the number of contaminated food items identified by the sampling strategy. This represents the probability that the sampling strategy detects at least one batch of contaminated antibiotic k from production site i, and this probability follows a binomial distribution. This indicates the quantity of food products from production location i that were contaminated with antibiotic k. This indicates the number of batches from production site i that were contaminated with antibiotic k and were used for random sampling. This represents the national probability that food contains antibiotic k. This represents the probability that a product originating from location i is contaminated with antibiotic k at the national level. It is the posterior probability of tracing antibiotic k back to consumption location j, calculated according to formula (1); This is an estimate of the maximum annual sampling capacity (batch) for food at the national level, based on historical sampling records. This is the minimum sample size requirement set for each pair of "antibiotic-production location" to ensure statistical reliability; and These represent the microbiological and toxicological acceptable daily intake (ADRI) of antibiotic k, respectively, and are used as safety thresholds. Based on the above objective function and constraints, a mixed-integer programming method is used to optimize and solve it, ultimately yielding the number of batches from production site i contaminated with antibiotic k.

[0113] Furthermore, the specific processing steps for the acquired data include:

[0114] S1.1 Data Cleaning Sub-step: Perform quality checks on the raw data, identify and process missing values, outliers, and duplicate entries; for missing data, use data imputation algorithms and related data matching methods to fill in the missing data; for sample data of duplicate records, delete them; preferably, use the K-nearest neighbor algorithm to imput missing values ​​and use the 3σ criterion to identify and correct outliers.

[0115] S1.2 Data Standardization Sub-step: Standardize the cleaned data to unify the units and numerical ranges. Data standardization uses the Z-Score method to transform the data into a standard form with a mean of 0 and a standard deviation of 1. The purpose is to unify the units and numerical ranges of various variables, eliminate scale differences between different indicators, improve data consistency and comparability, and provide standardized input for subsequent model calculations.

[0116] S1.3 Data Update Sub-step: This involves periodically or in real-time updating the dataset to include the latest data on food sampling, food trade, and consumption. Preferably, at the end of each year, food safety sampling records published by 31 provincial market supervision administrations (data for Hong Kong, Macao, and Taiwan is currently unavailable) are collected and added to the existing dataset. Parameters such as contamination probability, contamination concentration, and intake are re-estimated, and the optimization problem is re-solved to obtain the optimal sampling plan for the following year, ensuring the timeliness and accuracy of the model.

[0117] Furthermore, to better illustrate the above-described method of the present invention, specific embodiments and comparative examples are provided below:

[0118] Taking Chinese poultry eggs as an example. Table 1 shows the results of two random inspections of peanuts conducted by the Beijing Municipal Administration for Market Regulation (the place of consumption) in 2018 and 2019. Batch 1 indicates that the Beijing Municipal Administration for Market Regulation conducted a random inspection on October 23, 2018, on a batch of poultry eggs from Hebei Province. The monitoring results showed that the enrofloxacin concentration was 5.6 (µg / kg), which is less than the national legal standard of 10 (µg / kg), indicating that this batch of poultry eggs is qualified. Batch 2 indicates that the Beijing Municipal Administration for Market Regulation conducted a random inspection on July 19, 2019, on a batch of poultry eggs from Beijing. The monitoring results showed that the enrofloxacin concentration was 129.5 (µg / kg), which is greater than the national legal limit of 10 (µg / kg), indicating that this batch of poultry eggs is unqualified.

[0119]

[0120] Example 1

[0121] Step 1: Based on food consumption data, we can obtain the average and median daily consumption of poultry eggs per person in various consumption areas (provinces) in my country (data for Hong Kong, Macao and Taiwan are currently unavailable) from 2015 to 2022. As shown in Table 2, there are significant differences in poultry egg consumption among various consumption areas in my country from 2015 to 2022. Tianjin has the highest average daily consumption of poultry eggs per person at 51.64 grams / person / day, while Tibet Autonomous Region has the lowest average daily consumption of poultry eggs per person at 8.53 grams / person / day. The median consumption is also highest in Tianjin and lowest in Tibet Autonomous Region.

[0122] Table 2 - Average and median daily poultry and egg consumption per person in various consumption regions (data for Hong Kong, Macao and Taiwan is currently unavailable) from 2015 to 2022 (g / person / day)

[0123] Step 2: Based on food consumption data, we can obtain the mean and median population of each consumption region (province) in my country from 2015 to 2022. As can be seen from Table 3, there are significant differences in the population of each consumption region (province) in my country from 2015 to 2022. Guangdong Province has the highest mean population at 123,161,300, while Xinjiang Uygur Autonomous Region has the lowest mean population at 3,537,500. Similarly, Guangdong Province has the highest median population, while Xinjiang Uygur Autonomous Region has the lowest.

[0124]

[0125] Step 3: Based on the antibiotic data, the acceptable daily intake (ADI) for microbiology and toxicology of 11 major contaminating antibiotics, including enrofloxacin, ofloxacin, and chloramphenicol, can be obtained, as shown in Table 4. The ADI for toxicology can be used directly, while the ADI for microbiology will be recalculated using formula (15). The results show that among the ADIs for toxicology, sulfamethoxazole has the highest value of 50 (µg / kg body weight / day), while enrofloxacin and ofloxacin have the lowest values ​​of 0.3 (µg / kg body weight / day). Among the calculated ADIs for microbiology, chloramphenicol has the highest value of 0.917 (µg / kg body weight / day), while ciprofloxacin has the lowest value of 0.014 (µg / kg body weight / day).

[0126]

[0127] Step 4: Based on the antibiotic data, the regulatory limits and detection limits for 11 major contaminating antibiotics, including enrofloxacin, ofloxacin, and chloramphenicol, can be obtained, as shown in Table 5. Among the legal limits, 6 antibiotics, including chloramphenicol and ciprofloxacin, are prohibited from detection; 4 antibiotics, including doxycycline and enrofloxacin, have a legal limit of 10 (µg / kg); and ofloxacin has a limit of 2 (µg / kg). In the detection limit data, thiamphenicol has the highest detection limit at 0.59 (µg / kg), while florfenicol has the lowest at 0.02 (µg / kg). The detection limits will be used to fill in any gaps in the concentration of non-compliant samples at a particular consumption location.

[0128]

[0129] Step 5: Based on food safety sampling data (2015-2022), and using formula (1), the contamination probability of antibiotic residue k in each consumption area (province) can be obtained. Specific data are shown in Table 6. The data in Table 6 are then used to create heat maps for East China, South China, North China, Central China, Southwest China, Northwest China, and Northeast China, as shown below. Figure 2 As shown in the figure, there are significant differences in the probability of contamination of various antibiotics in different consumption areas in my country. Enrofloxacin and florfenicol are both found to be contaminated in poultry eggs in all consumption areas, with a high probability of contamination. The contaminated areas are mainly concentrated in the southern / southern regions (such as Guangdong, Guangxi, and Yunnan), followed by the central / eastern regions (Hubei and Hunan; Jiangsu and Zhejiang).

[0130]

[0131] Step 6: Based on food safety sampling data (2015-2022), and using formula (2), the contamination probability of 11 antibiotics nationwide can be obtained, as shown in Table 7. Figure 3As shown, there are significant differences in the contamination rates of various antibiotics in poultry eggs across my country. Florfenicol has the highest contamination rate at approximately 0.0065, followed by enrofloxacin at approximately 0.0043, while thiamphenicol has the lowest contamination rate.

[0132]

[0133] Step 7: Based on food safety sampling data (2015-2022), and using formulas (3) and (4), the average concentrations of 11 antibiotics in various consumption areas and nationwide can be calculated. Specific data are shown in Table 8-16. Among them, metronidazole has the highest national average concentration of contamination at 6948.79 (µg / kg), followed by enrofloxacin at 6711.58 (µg / kg), and the lowest is thiamphenicol at 60.87 (µg / kg). Metronidazole has the highest average concentration of contamination in Zhejiang at 639.73 (µg / kg), and the lowest average concentration in Shanghai at 4.18 (µg / kg) (excluding the case of LOD filling).

[0134]

[0135]

[0136]

[0137]

[0138]

[0139]

[0140]

[0141]

[0142]

[0143] Step 8: Based on food safety sampling data (2015-2022), construct a dataset of non-compliant samples. Using a Bayesian network model, obtain the probability of the flow of each antibiotic in poultry eggs along the national "production-consumption" supply chain. Organize the probabilities of traceable / untraceable sources in Table 17, and plot the specific marginal probabilities and posterior probabilities. Figure 5 , Figure 6 , Figure 7 ,Depend on Figure 6-7As shown in Table 17, the probability of untraceable sources for all antibiotics is greater than that of traceable sources, with a ratio of approximately 8:2. The posterior probability of untraceable sources in various regions and compounds is significantly higher than that of traceable sources (peak value 30-34%). Figure 6 The posterior probability of traceable origin is compressed to 9% (), while the posterior probability is compressed to 9%. Figure 7 In both scenarios, the eastern region exhibited the highest posterior probabilities, with many compounds exceeding 25% when the source was untraceable, decreasing to 6%-9% when the source was traceable. This was followed by the central and southern regions; the northern, northeastern, and northwestern regions maintained lower probabilities. High-risk signals were primarily driven by doxycycline, enrofloxacin, and florfenicol, while chloramphenicol and thiamphenicol appeared less frequently and had weaker signals.

[0144]

[0145] Step 9: Based on the pollution probability, average pollution concentration, and predicted flow probability calculated in Steps 5-8, and combined with the consumption and population data obtained in Steps 1-2, the number of poultry eggs contaminated by the 11 antibiotics in each "production-consumption" supply chain is calculated using formulas (8)-(10). The number of contaminated poultry eggs in each consumption area is compiled into Table 18, and the number of poultry eggs contaminated by each antibiotic is plotted as follows: Figure 8 ,Depend on Figure 8 As shown in Table 18, nationwide, enrofloxacin and florfenicol are the two largest categories of antibiotics in contaminated eggs, accounting for 102,000 tons and 67,000 tons annually, respectively; furazolidone and thiamphenicol contribute less. Spatially, the number of antibiotic-contaminated eggs is concentrated in eastern and southern provinces (e.g., Shandong and Guangdong), while the northwest and northeast regions have relatively lower levels; antibiotic residues are mainly dominated by enrofloxacin and florfenicol, with most contaminated eggs originating from untraceable sources, and traceable sources accounting for the remainder. Transfer analysis shows ( Figure 9 The consumption of contaminated eggs is highest in the East (89.44 thousand tons per year), South (50.42 thousand tons per year), and Southwest (34.41 thousand tons per year), with untraceable contaminated eggs accounting for the majority in each region, peaking in the Southwest (90.7%), while the South (85.1%) and Northwest (84.5%) also maintain high levels. Several producing regions export a significant share of their contaminated output to other regions (e.g., approximately 31.8% in the Northeast and approximately 10.8% in the Central), indicating substantial inter-regional redistribution.

[0146]

[0147] Step 10: Based on the pollution probability, average pollution concentration, and predicted flow probability calculated in Steps 5-8, and combined with the consumption and population data obtained in Steps 1-2, the average intake of 11 antibiotics in various consumption areas and nationwide is calculated based on formulas (11)-(13). The national average intake of the 11 antibiotics is compiled into Table 19. The detailed average intake of each antibiotic in each consumption area (divided into East China, South China, North China, Central China, Southwest China, Northwest China, and Northeast China) is plotted as follows. Figure 10-11 The calculation results are then compared with the safety thresholds in Table 4. Figure 10-11 As shown in Table 19, the average daily intake of many antibiotics is close to their maximum acceptable daily intake (MADI) for microbiology. The national average daily intake of enrofloxacin is 0.1 μg / kg body weight / day, slightly exceeding its MADI (0.11 μg / kg body weight / day) in both the central and eastern regions and nationwide, indicating a potential risk of resistance selection in human Escherichia coli. Nationwide ( Figure 11 In terms of dietary exposure, metronidazole (average daily intake of 0.20 μg / kg body weight / day) and enrofloxacin were the primary sources, followed by sulfamethoxazole (average daily intake of 0.15 μg / kg body weight / day) and doxycycline (average daily intake of 0.12 μg / kg body weight / day). In various regions ( Figure 10 a–f、 Figure 11 In the GK (General Data Points), the eastern region had the highest average daily intake of most antibiotics, while the northern region saw peak intakes of florfenicol and furazolidone, and the central region also had high levels of metronidazole. Levels close to the MADI (Maximum Intake Level) included ciprofloxacin in the eastern region (MADI = 0.014 μg / kg body weight / day), ofloxacin in the northeastern region (MADI = 0.11 μg / kg body weight / day), and doxycycline in the southern region (MADI = 0.24 μg / kg body weight / day).

[0148]

[0149] Step 11: Based on the pollution probability, average pollution concentration, predicted flow probability, number of contaminated poultry eggs, and average daily intake calculated in Steps 5-10, and combined with the consumption and population data obtained in Steps 1-2, four optimization scenarios are established based on formulas (13)-(21). By solving the optimization problem, the optimized sampling plan is obtained, and the reduction in average daily intake after optimization is calculated as the evaluation of optimization benefits. The specific optimization data for each scenario are compiled into Tables 13-21, and the overall results are visualized. Figure 12-13 .

[0150] Scenario 1: Sampling resources are simultaneously allocated between traceable and non-traceable food sources. A budget of 5000–50000 is set, and the optimization is performed to determine the number of batches required to monitor each antibiotic in poultry eggs under different traceability statuses within each budget. Variable descriptions are shown in Table 20, and specific data are shown in Tables 21-23. As the figures and tables show, with increasing budget (30,000–50,000 batches), the monitoring combination expands to a medium detection rate target, adding metronidazole (traceable), ofloxacin, and dimethicone (non-traceable), while low-detection-rate residues remain at a marginal level. Figure 12 a). As input increases, the number of contaminated eggs detected increases, but tends to stabilize at around 170–230 thousand tons per year, with untraceable sources consistently exceeding traceable sources. Figure 12 b). The average daily intake of all nine antibiotics showed a decreasing trend; at 50,000 batches, metronidazole (0.19 μg / kg body weight / day), enrofloxacin (0.16 μg / kg body weight / day), and sulfamethoxazole (0.13 μg / kg body weight / day) showed the largest decreases. Figure 12 c). Under a medium budget (5,000–15,000 batches), enrofloxacin resulted in the largest reduction in average daily intake, followed by metronidazole (traceable), ofloxacin, and sulfamethoxazole (non-traceable), while low-detection residues remained at a marginal level. Figure 12 a). As input increases, the number of contaminated eggs detected increases, but tends to stabilize at around 170–230 thousand tons per year, with untraceable sources consistently exceeding traceable sources. Figure 12 b). The average daily intake of all nine antimicrobial drugs showed a decreasing trend; at 50,000 batches, metronidazole (0.19 μg / kg body weight / day), enrofloxacin (0.16 μg / kg body weight / day), and sulfamethoxazole (0.13 μg / kg body weight / day) showed the largest decreases. Figure 12 c). At a medium budget (5,000–15,000 lots), enrofloxacin resulted in the largest reduction in average daily intake, after which metronidazole became dominant (>15,000 lots).

[0151]

[0152]

[0153]

[0154]

[0155] Scenario 2: Sampling resources are allocated only to traceable food sources. A budget of 5000-50000 is set, and the number of batches required to monitor the traceability status of each antibiotic in poultry eggs is optimized under different budgets. Variable descriptions are shown in Table 24, and specific data are shown in Tables 25-27. As the figures and tables show, with increasing budget, the allocation scheme increasingly favors sulfamethoxazole, ofloxacin, and metronidazole, while low-prevalence antibiotics remain marginal. The detection volume increases with investment but tends to stabilize at a level of 30-50 thousand tons / year. The six antibiotics show a decrease in average daily intake.

[0156]

[0157]

[0158]

[0159]

[0160] Scenario 3: Sampling resources are allocated only to food products from untraceable sources. A budget of 5000-50000 is set, and the number of batches required to monitor the untraceability status of each antibiotic in poultry eggs is optimized under different budgets. Variable descriptions are shown in Table 28, and specific data are shown in Tables 29-31. As the graphs and tables show, with increasing budgets, the detection limit remains similar (30-50 thousand tons per year), but the average daily intake of the eight antimicrobial drugs decreases; additional batches are particularly effective for doxycycline and ciprofloxacin.

[0161]

[0162]

[0163]

[0164]

[0165] Scenario 4: Resource allocation is based on antibiotic type, comprehensively considering both traceable and untraceable food sources. A budget of 500-5000 is set, and the number of batches required to monitor different traceability statuses of a single antibiotic under different budgets is optimized. Variable descriptions are shown in Table 32, and specific data are shown in Tables 33-37. As the graphs and tables show, early resources are mainly concentrated on high-incidence and untraceable residues, subsequently expanding to traceable florfenicol and enrofloxacin, while low-incidence residues only receive a small number of batches for testing. With increasing budget, the number of detectable contaminated eggs and the average daily intake reduction both show an upward trend, subsequently stabilizing. Compared to multi-residue testing combinations, single-residue testing items reach a stable state later and at a higher level, consistent with the slower saturation rate of leading single-residue items.

[0166]

[0167]

[0168]

[0169]

[0170]

[0171]

[0172] Furthermore, this invention also provides a food sampling optimization system for antibiotic residue traceability, used to implement the methods described above. The system includes: a data acquisition module for acquiring food safety sampling data, consumption data, and population data; connecting to a market supervision database to update sampling data in real time; a data processing module for cleaning and standardizing data; a model prediction module for performing Bayesian network flow prediction and contamination parameter estimation; deploying Bayesian networks and optimization algorithms to support cloud computing; a risk assessment module for calculating daily intake and contamination levels and performing risk analysis; and finally generating a risk level report (high / medium / low); and an optimization module for constructing a multi-scenario mixed integer programming model and optimizing the sampling strategy.

[0173] The system integrates a cloud computing platform to enable real-time data updates and dynamic model adjustments. Data is updated monthly, and the sampling scheme is re-optimized.

[0174] Furthermore, specific embodiments of the food sampling optimization system for antibiotic residue traceability and a computer storage medium of the present invention are basically the same as the above-described method embodiments, and will not be described in detail here.

[0175] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. All equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. An optimized method for food sampling inspection aimed at antibiotic residue traceability, characterized in that, The method includes the following steps: Step 1: Obtain relevant characteristic data of the target food and characteristic data of the antibiotic. The relevant characteristic data of the target food includes food safety sampling data, food consumption data and population-related data. The characteristic data of the antibiotic includes microbiologically acceptable daily intake, toxicologically acceptable daily intake and detection limit data in food. Step 2: Determine the probability and average concentration of antibiotic contamination in the target food based on the target food safety sampling data obtained in Step 1. Step 3: Construct a Bayesian network model based on the target food safety sampling data obtained in Step 1, and use Monte Carlo simulation to perform probabilistic reasoning to obtain the probability of the flow of each antibiotic in the production-consumption supply chain. Step 4: Based on the probability of antibiotic contamination and average concentration in the target food in Step 2, the flow probability in Step 3, and the food consumption and population of the target food in the preset area obtained in Step 1, calculate the number of foods contaminated with antibiotics and the average daily intake of antibiotics by people in different regions and the national population through food. Step 5: Establish a resource allocation scenario based on the traceable and non-traceable production locations of the samples. The objective function is to maximize the number of foods contaminated with antibiotics. An optimization model is created with the constraints that the number of sampled batches does not exceed the historical maximum number of sampled samples of the target food in the preset area, and the average daily intake of antibiotics by residents nationwide through food does not exceed the microbiologically acceptable daily intake and the toxicologically acceptable daily intake. The optimal sample size and sampling frequency for the target food are determined using mixed integer programming.

2. The method according to claim 1, characterized in that, In step 1, the relevant characteristic data of the target food are cleaned and standardized to unify the units and numerical ranges and ensure data consistency and comparability. Data cleaning includes identifying and processing missing values, outliers and duplicate entries. Data standardization uses the Z-Score method to transform the data into a standard form with a mean of 0 and a standard deviation of 1.

3. The method according to claim 1, characterized in that, In step 2, the formula for calculating the pollution probability is: ， ； in, and Let $k$ represent the probability of antibiotic residue $k$ contamination in food at the consumption location $j$ and the country level, respectively. This indicates the number of food samples in consumption location j that are contaminated with a specific antibiotic residue k; This represents the total number of food samples randomly inspected in consumption location j; The formula for calculating the average concentration is as follows: ， ； and These represent the average concentrations of antibiotic k in the corresponding contaminated egg samples at the consumption location j and the country level, respectively. This indicates the concentration of antibiotic K in a specific contaminated sample C; and These represent the number of samples contaminated with antibiotic k at the consumption location j and the country level, respectively.

4. The method according to claim 1, characterized in that, Step 3, specifically the step of constructing the Bayesian network model, includes: Based on the food safety sampling data obtained in step 1, a dataset of non-compliant samples is created. A Bayesian network model is constructed using food category, antibiotic type, production time, consumption time, production location, consumption location, and supply chain node characteristics as network nodes. The belief propagation method is then used to obtain the probability distribution of the flow of each antibiotic in the production-consumption supply chain.

5. The method according to claim 1, characterized in that, In step 4, the formula for calculating the average daily intake of antibiotics is as follows: ； in, This represents the average estimated daily intake of antibiotic k in consumption location j. The average concentration of antibiotic k in consumption location j. BW represents the average daily food intake per person in consumption location j, and BW represents the population weight.

6. The method according to claim 1, characterized in that, In step 5, the objective function is: ； in, This represents the probability of detecting contamination of antibiotic k from production location i. This refers to the quantity of food produced at location i that was contaminated with antibiotic k. The number of samples is the decision variable.

7. The method according to claim 1, characterized in that, In step 1, the antibiotics include at least one of enrofloxacin, ofloxacin, chloramphenicol, ciprofloxacin, metronidazole, doxycycline, florfenicol, furazolidone, metronidazole, sulfamethoxazole, and thiamphenicol.

8. The method according to claim 1, characterized in that, In step 5, resource allocation scenarios include simultaneously allocating sampled resources between traceable and non-traceable food sources, allocating only to traceable food sources, allocating only to non-traceable food sources, or allocating resources based on antibiotic type.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1-8.

10. A food sampling optimization system for antibiotic residue traceability, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-8.

Images