CN118608106A - An intelligent agricultural supply chain supervision method - Google Patents

Abstract

The present invention discloses an intelligent agricultural supply chain supervision method, which includes data collection, data preprocessing, establishing an agricultural supply chain supervision model, model parameter optimization and agricultural supply chain supervision. The present invention belongs to the field of data processing technology, and specifically refers to an intelligent agricultural supply chain supervision method. This scheme introduces a convex combination based on two Laplace functions to define a robust entropy, and then defines a robust loss function; the loss function guided by mixed entropy is converted into a convex optimization problem, and a parameter update mechanism is formulated to handle constrained optimization problems, thereby increasing the reliability of model optimization; the lens imaging reverse learning method and the optimized circular chaos map are used to initialize and optimize individual positions; by designing an adaptive nonlinear decreasing weight factor, the search capability is dynamically adjusted and the search range is controlled; the parameters of the agricultural supply chain supervision model are efficiently and accurately optimized to improve the supervision capability and prediction accuracy of the supply chain.

Description

An intelligent agricultural supply chain supervision method

Technical Field

The present invention relates to the technical field of data processing, and in particular to an intelligent agricultural supply chain supervision method.

Background Art

The core purpose of agricultural supply chain supervision methods is to ensure that the entire process from production to sales of agricultural products can be carried out smoothly and efficiently, to help manage and optimize the agricultural supply chain, and to ensure food safety, quality and supply stability. However, general agricultural supply chain supervision methods have the problem that the model has poor ability to process noise data and is insensitive to samples that deviate from the central trend, resulting in poor model stability and serious risk of model overfitting; general agricultural supply chain supervision methods have the problem that the population initialization diversity and distribution uniformity are insufficient when optimizing model parameters, resulting in weak global search capabilities, slow search convergence speed and weak learning, which makes it impossible to quickly achieve accurate optimization.

Summary of the invention

In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent agricultural supply chain supervision method. In view of the problems that the general agricultural supply chain supervision method has poor model processing noise data capability and is insensitive to samples that deviate from the central trend, resulting in poor model stability and serious risk of model overfitting, this scheme introduces a convex combination based on two Laplace functions to define a robust entropy, and then defines a robust loss function, which can flexibly handle deviations, so that the model pays more attention to reducing the overall error and enhances the generalization ability of the model; by defining the upper and lower bounds of the robust entropy, a theoretical guarantee for the stability and convergence of the model is provided, and the introduction of Gaussian noise and the definition of boundary conditions help to analyze the stability performance of the model under different noise levels; the loss function guided by mixed entropy is converted into a convex optimization problem, and a parameter update mechanism is formulated to handle constrained optimization problems, thereby increasing the reliability of model optimization; in view of the general The agricultural supply chain supervision method has the problem of insufficient population initialization diversity and distribution uniformity when optimizing model parameters, resulting in weak global search ability, slow search convergence speed and weak learning ability, which makes it impossible to quickly achieve accurate optimization. This scheme uses the lens imaging reverse learning method and the optimized circular chaos map to initialize and optimize the individual position, so that the individual initial position is closer to the potential global optimal solution; by designing an adaptive nonlinear decreasing weight factor, the search ability is dynamically adjusted and the search range is controlled. In the early stage, more emphasis is placed on exploration, and in the later stage, more emphasis is placed on development, and the search is conducted in the local space in a more detailed manner; for individuals whose fitness values are lower than the average fitness value of the population, mobile strategy one is applied to improve the fitness of inferior individuals, and mobile strategy two is applied to individuals whose fitness values are not lower than the average fitness value, giving full play to the exploration ability of high-quality individuals; efficiently and accurately optimize the parameters of the agricultural supply chain supervision model to improve the supervision ability and prediction accuracy of the supply chain.

The technical solution adopted by the present invention is as follows: The present invention provides an intelligent agricultural supply chain supervision method, which includes the following steps:

Step S1: data collection;

Step S2: data preprocessing;

Step S3: Establish an agricultural supply chain supervision model;

Step S4: model parameter optimization;

Step S5: Agricultural supply chain monitoring.

Furthermore, in step S1, the data collection is to collect historical agricultural supply chain supervision data; the historical agricultural supply chain supervision data includes agricultural product production data, logistics and transportation data, environmental data, supply chain management data and satisfaction feedback level; the satisfaction feedback level is used as a data label.

Furthermore, in step S2, the data preprocessing is to clean the collected historical agricultural supply chain supervision data, convert the data and divide the data set; the data cleaning is to process missing values, duplicate values and outliers in the historical agricultural supply chain supervision data; the data conversion is to convert the cleaned data into a vector form and standardize it based on the maximum and minimum normalization method; the data set division is to randomly divide the historical agricultural supply chain supervision data after data conversion into a training set and a test set.

Further, in step S3, the agricultural supply chain supervision model is established by constructing a neural network and optimizing the training set and test set obtained by preprocessing the historical agricultural supply chain supervision data, wherein the training set is used to train the agricultural supply chain supervision model and the test set is used to evaluate the performance of the agricultural supply chain supervision model; specifically, the following steps are included:

Step S31: define robust entropy; based on the convex combination of two Laplace functions as the kernel, induce a robust entropy, expressed as:

;

Where U(·) is the robust entropy; E[·] is the expectation; n is the number of samples, i is the sample index; P is the probability distribution of the actually observed agricultural supply chain data; Q is the probability distribution of the agricultural supply chain data predicted by the model; is the deviation between the actual observed data and the model predicted data; is the balance weight parameter; and is the shape parameter;

Step S32: define a robust loss function; based on the robust entropy, a robust loss function is derived, which is expressed as:

;

In the formula, is the robust loss function;

Step S33: define the upper bound; set ; is Gaussian noise, satisfying ; The upper bound of the robust entropy is expressed as:

;

Step S34: define a lower bound; the lower bound of the robust entropy is expressed as:

;

Step S35: Optimizing the loss function; specifically including:

Step S351: Minimize the loss function. The robust loss function is introduced into the agricultural supply chain supervision model and minimized, which is expressed as:

;

;

Where M is the feature matrix used for supply chain data transformation; b is the bias term of the model; is the slack variable; is the square of the Frobenius norm; C is the regularization parameter; is the slack variable of sample (i,j); is the label of sample (i, j);

Step S352: By mixing the entropy guided loss convex conjugate function, the problem is transformed into:

;

;

;

Where g(·) is a convex conjugate function and v is a built-in parameter; and is an auxiliary variable used to transform the problem into a convex optimization problem;

Step S36: Update, specifically including:

Step S361: Fixing and , optimize M,b, ; expressed as:

;

;

;

In the formula, is the auxiliary Lagrange multiplier; G is the gradient term; is the input feature matrix;

Step S362: Fix M, b, ,optimization and ; expressed as:

;

;

Step S37: model determination; based on the BP neural network architecture, with the assistance of steps S31 to S36 for optimization, when the model converges on the training set, the agricultural supply chain supervision model training is completed; an accuracy threshold is set in advance, when the prediction accuracy of the trained agricultural supply chain supervision model for the test set is higher than the accuracy threshold, the agricultural supply chain supervision model is established; otherwise, the agricultural supply chain supervision model parameters are adjusted and retrained.

Furthermore, in step S4, the model parameter optimization is to adjust the parameters of the agricultural supply chain supervision model; specifically, the following steps are included:

Step S41: initialization; construct parameter optimization space based on shape parameters, balance weight parameters, regularization parameters, neural network architecture parameters and slack variables; introduce lens imaging reverse learning, and initialize and optimize individual positions based on optimized circular chaos diagrams; use the prediction accuracy of the test set of the agricultural supply chain supervision model trained based on individual positions as the individual fitness value; the formula used to initialize and optimize individual positions is as follows:

;

;

In the formula, and are the preparatory positions of the Jth dimension of the I+1th optimized individual and the Ith optimized individual respectively; mod(·,·) is the modulo operation; is the initialization position of the optimized individual J-th dimension; and are the upper and lower bound positions of the Jth dimension of the population, respectively; q is a hyperparameter used to fine-tune the individual position;

Step S42: Design a weight factor; introduce an adaptive nonlinear decreasing weight factor, which is expressed as follows:

;

In the formula, is the weight factor at the tth iteration; tmax is the maximum number of iterations;

Step S43: Design a mobile strategy 1; for individuals whose fitness value is lower than the average fitness value of the population, use mobile strategy 1 to update their position; mobile strategy 1 is expressed as follows:

;

In the formula, and are the positions of the I-th individual at the t+1-th iteration and the t-th iteration of the J-th dimension respectively; is the average position of the population in the Jth dimension; r is a random number between 0 and 1; is the position of the Jth dimension of the optimal individual in the population; is the attenuation factor;

Step S44: Design a second movement strategy; for individuals whose fitness value is not lower than the average fitness value of the population, use the second movement strategy to update the position; the second movement strategy is expressed as follows:

;

In the formula, Q _I is the individual fitness value; is the position of the worst individual in the population in the Jth dimension; is a random number between 0 and 1, independent of r; M is the adjustment factor; is the individual's historical optimal position in the Jth dimension;

Step S45: search judgment; a fitness threshold is set in advance. If there is an individual fitness value higher than the fitness threshold, an agricultural supply chain supervision model is established based on the individual position; if the maximum number of iterations is reached, the optimized individual position in the population is reinitialized; otherwise, the fitness value is updated to continue the mobile search.

Furthermore, in step S5, the agricultural supply chain supervision is based on the established agricultural supply chain supervision model, and real-time collection of agricultural product production data, logistics and transportation data, environmental data and supply chain management data; after pre-processing, the data is input into the agricultural supply chain supervision model, and a level threshold is set in advance. When the satisfaction feedback level output by the model is lower than the level threshold, the supply chain is judged to be abnormal, and early warning processing is performed.

The beneficial effects achieved by the present invention using the above scheme are as follows:

(1) In view of the problems that general agricultural supply chain supervision methods have poor model processing ability for noisy data and are insensitive to samples that deviate from the central trend, resulting in poor model stability and serious risk of model overfitting, this scheme introduces a convex combination of two Laplace functions to define a robust entropy, and then defines a robust loss function, which can flexibly handle deviations, so that the model pays more attention to reducing the overall error and enhances the generalization ability of the model; by defining the upper and lower bounds of the robust entropy, a theoretical guarantee for the stability and convergence of the model is provided. The introduction of Gaussian noise and the definition of boundary conditions help to analyze the stability performance of the model under different noise levels; the loss function guided by mixed entropy is transformed into a convex optimization problem, and a parameter update mechanism is formulated to deal with constrained optimization problems, thereby increasing the reliability of model optimization.

(2) In view of the problem that the general agricultural supply chain supervision method has insufficient population initialization diversity and distribution uniformity when optimizing model parameters, resulting in weak global search ability, slow search convergence speed and weak learning ability, which makes it impossible to quickly achieve accurate optimization, this scheme uses the lens imaging reverse learning method and the optimized circular chaos map to initialize and optimize the individual position, so that the individual initial position is closer to the potential global optimal solution; by designing an adaptive nonlinear decreasing weight factor, the search ability is dynamically adjusted and the search range is controlled. In the early stage, more emphasis is placed on exploration, and in the later stage, more emphasis is placed on development, and the search is conducted in the local space in a more detailed manner; for individuals whose fitness values are lower than the average fitness value of the population, the movement strategy one is applied to improve the fitness of inferior individuals, and for individuals whose fitness values are not lower than the average fitness value, the movement strategy two is applied to give full play to the exploration ability of high-quality individuals; the parameters of the agricultural supply chain supervision model are optimized efficiently and accurately, and the supervision ability and prediction accuracy of the supply chain are improved.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a flow chart of an intelligent agricultural supply chain supervision method provided by the present invention;

FIG2 is a schematic flow chart of step S3;

FIG. 3 is a schematic flow chart of step S4 .

The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

DETAILED DESCRIPTION

The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

Embodiment 1, referring to FIG1 , the present invention provides an intelligent agricultural supply chain supervision method, the method comprising the following steps:

Step S1: Data collection, collecting historical agricultural supply chain supervision data;

Step S2: Data preprocessing, data cleaning, data conversion and data set division of the collected historical agricultural supply chain supervision data;

Step S3: Establish an agricultural supply chain supervision model, introduce a convex combination based on two Laplace functions to define the robust entropy, and then define the robust loss function; by defining the upper and lower bounds of the robust entropy, introducing Gaussian noise and defining boundary conditions, the loss function guided by the mixed entropy is transformed into a convex optimization problem, and a parameter update mechanism is formulated; thereby realizing the establishment of the agricultural supply chain supervision model;

Step S4: Model parameter optimization, using the lens imaging reverse learning method and the optimized circular chaos map to initialize and optimize individual positions; by designing an adaptive nonlinear decreasing weight factor, applying movement strategy 1 to individuals whose fitness values are lower than the average fitness value of the population to improve the fitness of inferior individuals, and applying movement strategy 2 to individuals whose fitness values are not lower than the average fitness value; finally, the parameter optimization of the agricultural supply chain supervision model is achieved;

Step S5: Agricultural supply chain monitoring.

Embodiment 2, referring to FIG. 1 , is based on the above embodiment. In step S1 , historical agricultural supply chain supervision data includes agricultural product production data, logistics and transportation data, environmental data, supply chain management data, and satisfaction feedback level; the satisfaction feedback level is used as a data label.

Embodiment 3, referring to FIG1 , this embodiment is based on the above embodiment. In step S2, data cleaning is to process missing values, duplicate values and outliers in historical agricultural supply chain supervision data; data conversion is to convert the cleaned data into vector form and perform standardization based on the maximum and minimum normalization method; data set partitioning is to randomly divide the historical agricultural supply chain supervision data after data conversion into a training set and a test set.

Embodiment 4, referring to FIG. 1 and FIG. 2, this embodiment is based on the above embodiment. In step S3, the agricultural supply chain supervision model is established by constructing a neural network and optimizing the training set and test set obtained by preprocessing the historical agricultural supply chain supervision data. The training set is used to train the agricultural supply chain supervision model, and the test set is used to evaluate the performance of the agricultural supply chain supervision model. Specifically, the following steps are included:

Step S31: define robust entropy; based on the convex combination of two Laplace functions as the kernel, induce a robust entropy, expressed as:

;

Where U(·) is the robust entropy; E[·] is the expectation; n is the number of samples, i is the sample index; P is the probability distribution of the actually observed agricultural supply chain data; Q is the probability distribution of the agricultural supply chain data predicted by the model; is the deviation between the actual observed data and the model predicted data; is the balance weight parameter; and is the shape parameter;

Step S32: define a robust loss function; based on the robust entropy, a robust loss function is derived, which is expressed as:

;

In the formula, is the robust loss function;

Step S33: define the upper bound; set ; is Gaussian noise, satisfying ; The upper bound of the robust entropy is expressed as:

;

Step S34: define a lower bound; the lower bound of the robust entropy is expressed as:

;

Step S35: Optimizing the loss function; specifically including:

Step S351: Minimize the loss function. The robust loss function is introduced into the agricultural supply chain supervision model and minimized, which is expressed as:

;

;

Where M is the feature matrix used for supply chain data transformation; b is the bias term of the model; is the slack variable; is the square of the Frobenius norm; C is the regularization parameter; is the slack variable of sample (i,j); is the label of sample (i, j);

Step S352: By mixing the entropy guided loss convex conjugate function, the problem is transformed into:

;

;

;

Where g(·) is a convex conjugate function and v is a built-in parameter; and is an auxiliary variable used to transform the problem into a convex optimization problem;

Step S36: Update, specifically including:

Step S361: Fixing and , optimize M,b, ; expressed as:

;

;

;

In the formula, is the auxiliary Lagrange multiplier; G is the gradient term; is the input feature matrix;

Step S362: Fix M, b, ,optimization and ; expressed as:

;

;

Step S37: model determination; based on the BP neural network architecture, with the assistance of steps S31 to S36 for optimization, when the model converges on the training set, the agricultural supply chain supervision model training is completed; an accuracy threshold is set in advance, when the prediction accuracy of the trained agricultural supply chain supervision model for the test set is higher than the accuracy threshold, the agricultural supply chain supervision model is established; otherwise, the agricultural supply chain supervision model parameters are adjusted and retrained.

By performing the above operations, in view of the problems that general agricultural supply chain supervision methods have poor model processing noise data capabilities and are insensitive to samples that deviate from the central trend, resulting in poor model stability and serious model overfitting risks, this scheme introduces a convex combination based on two Laplace functions to define a robust entropy, and then defines a robust loss function, which can flexibly handle deviations, so that the model pays more attention to reducing the overall error and enhances the generalization ability of the model; by defining the upper and lower bounds of the robust entropy, a theoretical guarantee for the stability and convergence of the model is provided, and the introduction of Gaussian noise and the definition of boundary conditions help to analyze the stability performance of the model under different noise levels; the loss function guided by mixed entropy is transformed into a convex optimization problem, and a parameter update mechanism is formulated to deal with constrained optimization problems, thereby increasing the reliability of model optimization.

Embodiment 5, referring to FIG. 1 and FIG. 3 , this embodiment is based on the above embodiment. In step S4, the model parameter optimization is to adjust the parameters of the agricultural supply chain supervision model; specifically, the following steps are included:

Step S41: initialization; construct parameter optimization space based on shape parameters, balance weight parameters, regularization parameters, neural network architecture parameters and slack variables; introduce lens imaging reverse learning, and initialize and optimize individual positions based on optimized circular chaos diagrams; use the prediction accuracy of the test set of the agricultural supply chain supervision model trained based on individual positions as the individual fitness value; the formula used to initialize and optimize individual positions is as follows:

;

;

In the formula, and are the preparatory positions of the Jth dimension of the I+1th optimized individual and the Ith optimized individual respectively; mod(·,·) is the modulo operation; is the initialization position of the optimized individual J-th dimension; and are the upper and lower bound positions of the Jth dimension of the population, respectively; q is a hyperparameter used to fine-tune the individual position;

Step S42: Design a weight factor; introduce an adaptive nonlinear decreasing weight factor, which is expressed as follows:

;

In the formula, is the weight factor at the tth iteration; tmax is the maximum number of iterations;

Step S43: Design a mobile strategy 1; for individuals whose fitness value is lower than the average fitness value of the population, use mobile strategy 1 to update their position; mobile strategy 1 is expressed as follows:

;

In the formula, and are the positions of the I-th individual at the t+1-th iteration and the t-th iteration of the J-th dimension respectively; is the average position of the population in the Jth dimension; r is a random number between 0 and 1; is the position of the Jth dimension of the optimal individual in the population; is the attenuation factor;

Step S44: Design a second movement strategy; for individuals whose fitness value is not lower than the average fitness value of the population, use the second movement strategy to update the position; the second movement strategy is expressed as follows:

;

In the formula, Q _I is the individual fitness value; is the position of the worst individual in the population in the Jth dimension; is a random number between 0 and 1, independent of r; M is the adjustment factor; is the individual's historical optimal position in the Jth dimension;

Step S45: search judgment; a fitness threshold is set in advance. If there is an individual fitness value higher than the fitness threshold, an agricultural supply chain supervision model is established based on the individual position; if the maximum number of iterations is reached, the optimized individual position in the population is reinitialized; otherwise, the fitness value is updated to continue the mobile search.

By performing the above operations, in view of the problem that the general agricultural supply chain supervision method has insufficient population initialization diversity and distribution uniformity when optimizing model parameters, resulting in weak global search ability, slow search convergence speed and weak learning ability, which makes it impossible to quickly achieve accurate optimization, this scheme uses the lens imaging reverse learning method and the optimized circular chaos map to initialize and optimize the individual position, so that the individual initial position is closer to the potential global optimal solution; by designing an adaptive nonlinear decreasing weight factor, the search ability is dynamically adjusted, and the search range is controlled, with more emphasis on exploration in the early stage and development in the later stage, and a more detailed search in the local space; for individuals with fitness values lower than the average fitness value of the population, mobile strategy one is applied to improve the fitness of inferior individuals, and mobile strategy two is applied to individuals with fitness values not lower than the average fitness value, giving full play to the exploration ability of high-quality individuals; efficiently and accurately optimize the parameters of the agricultural supply chain supervision model to improve the supervision ability and prediction accuracy of the supply chain.

Embodiment 6, referring to FIG1, this embodiment is based on the above embodiment. In step S5, agricultural supply chain supervision is based on the established agricultural supply chain supervision model, and real-time collection of agricultural product production data, logistics and transportation data, environmental data and supply chain management data; after pre-processing, the data is input into the agricultural supply chain supervision model, and a level threshold is pre-set. When the satisfaction feedback level output by the model is lower than the level threshold, the supply chain is judged to be abnormal, and early warning processing is performed.

It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit of the invention.

The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims (6)

1. An intelligent agricultural supply chain supervision method is characterized in that: the method comprises the following steps:

step S1: collecting data;

step S2: preprocessing data;

Step S3: establishing an agricultural supply chain supervision model, and introducing a convex combination definition Lu Bangshang based on two Laplacian functions to further define a robust loss function; the method comprises the steps of converting a loss function guided by mixed entropy into a convex optimization problem by defining an upper bound and a lower bound of a robust entropy, introducing Gaussian noise and defining boundary conditions, and formulating a parameter updating mechanism; thereby realizing the establishment of an agricultural supply chain supervision model;

Step S4: model parameter optimization, namely initializing and optimizing individual positions by using a lens imaging reverse learning method and an optimized round chaos map; by designing a self-adaptive nonlinear decrementing weight factor, a first movement strategy is applied to individuals with fitness values lower than the average fitness value of the population to improve the fitness of poor individuals, and a second movement strategy is applied to individuals with fitness values not lower than the average fitness value; finally, parameter optimization of an agricultural supply chain supervision model is realized;

step S5: agricultural supply chain supervision;

Step S3 includes step S31: definition Lu Bangshang; based on the convex combination of two laplace functions as a kernel, a robust entropy is induced, expressed as:

；

Wherein U (·) is Lu Bangshang; e is desired; n is the number of samples, i is the sample index; p is the probability distribution of the actual observed agricultural supply chain data; q is the probability distribution of model predicted agricultural supply chain data; is the deviation between the actual observed data and the model predicted data; Is a balance weight parameter; And Is a shape parameter.

2. An intelligent agricultural supply chain supervision method according to claim 1, wherein: in step S3, the building of the agricultural supply chain supervision model is to build a neural network and optimize the neural network based on a training set and a testing set obtained by preprocessing historical agricultural supply chain supervision data, wherein the training set is used for training the agricultural supply chain supervision model, and the testing set is used for evaluating the performance of the agricultural supply chain supervision model; the method specifically comprises the following steps:

Step S31: definition Lu Bangshang;

step S32: defining a robust loss function; based on the robust entropy, a robust loss function is guided out, expressed as:

；

In the method, in the process of the invention, Is a robust loss function;

Step S33: defining an upper bound; is provided with ；Is Gaussian noise, satisfy; The upper bound of the robust entropy is expressed as:

；

step S34: defining a lower bound; the lower bound of the robust entropy is expressed as:

；

Step S35: optimizing a loss function; the method specifically comprises the following steps:

step S351: minimizing the loss function, introducing the robust loss function into the agricultural supply chain supervision model, and minimizing, expressed as:

；

；

Wherein M is a feature matrix for conversion of supply chain data; b is a bias term for the model; Is a relaxation variable; is the square of the Fr Luo Beini Usne norm; c is a regularization parameter; Is the relaxation variable of sample (i, j); A tag that is sample (i, j);

Step S352: the problem is translated into a mixed entropy guided loss convex conjugate function:

；

；

；

wherein g (·) is a convex conjugate function and v is a built-in parameter; And Is an auxiliary variable for converting the problem into a convex optimization problem;

step S36: the updating specifically comprises the following steps:

Step S361: fixing AndOptimizing the values of M, b,; Expressed as:

；

；

；

In the method, in the process of the invention, Is an auxiliary lagrangian multiplier; g is a gradient term; Is an input feature matrix;

Step S362: the weight of the fixed M, b, OptimizingAnd; Expressed as:

；

；

Step S37: judging a model; based on BP neural network architecture, optimizing with the aid of step S31 to step S36, and finishing the training of the agricultural supply chain supervision model when the model converges to the training set; presetting a correct rate threshold, and finishing the establishment of the agricultural supply chain supervision model when the predicted correct rate of the trained agricultural supply chain supervision model to the test set is higher than the correct rate threshold; otherwise, adjusting the parameters of the supervision model of the agricultural supply chain to retrain.

3. An intelligent agricultural supply chain supervision method according to claim 1, wherein: in step S4, the model parameter optimization is to adjust parameters of an agricultural supply chain supervision model; the method specifically comprises the following steps:

Step S41: initializing; constructing a parameter optimization space based on the shape parameters, the balance weight parameters, the regularization parameters, the architecture parameters of the neural network and the relaxation variables; introducing lens imaging reverse learning, and initializing and optimizing individual positions based on an optimized round chaotic map; taking the prediction accuracy of the agricultural supply chain supervision model trained based on the individual position to the test set as an individual fitness value; the formula for initializing and optimizing the individual position is as follows:

；

；

In the method, in the process of the invention, AndThe preparation positions of the J dimension of the I+1st optimized individual and the I optimized individual are respectively; mod (·, ·) is a modulo operation; Is an initialization location that optimizes the J-th dimension of the individual; And The upper limit position and the lower limit position of the J dimension of the population are respectively; q is a hyper-parameter for fine-tuning the individual's position;

step S42: designing a weight factor; an adaptive nonlinear decrementing weight factor is introduced, expressed as follows:

；

In the method, in the process of the invention, Is the weight factor at the t-th iteration; tmax is the maximum number of iterations;

Step S43: designing a first movement strategy; for individuals with fitness values lower than the population average fitness value, carrying out position update by adopting a first movement strategy; the movement policy one is expressed as follows:

；

In the method, in the process of the invention, AndThe positions of the ith individual in the jth dimension at the time of the (t+1) th iteration and the (t) th iteration are respectively; is the average position of the J-th dimension of the population; r is a random number from 0 to 1; Is the position of the J dimension of the population optimal individuals; Is an attenuation factor;

step S44: designing a second movement strategy; for individuals with fitness values not lower than the population average fitness value, carrying out position update by adopting a second movement strategy; the movement policy two is expressed as follows:

；

wherein Q _I is an individual fitness value; is the position of the J dimension of the worst individuals of the population; a random number of 0 to 1, independent of r; m is a regulatory factor; Is the location of the individual history optimal J dimension;

Step S45: searching and judging; presetting an fitness threshold, and if the individual fitness value is higher than the fitness threshold, establishing an agricultural supply chain supervision model based on the individual position; if the maximum iteration times are reached, the optimized individual positions in the population are reinitialized; otherwise, updating the fitness value and continuing the mobile search.

4. An intelligent agricultural supply chain supervision method according to claim 1, wherein: in step S1, the data collection is collection of historical agricultural supply chain regulatory data; the historical agricultural supply chain supervision data includes agricultural product production data, logistics transportation data, environmental data, supply chain management data, and satisfaction feedback levels; and taking the satisfaction feedback level as a data tag.

5. An intelligent agricultural supply chain supervision method according to claim 1, wherein: in step S5, the agricultural supply chain supervision is based on the established agricultural supply chain supervision model, and agricultural product production data, logistics transportation data, environmental data and supply chain management data are collected in real time; and inputting the processed result into an agricultural supply chain supervision model, presetting a grade threshold, and judging that the supply chain is abnormal when the satisfaction feedback grade output by the model is lower than the grade threshold, and performing early warning treatment.

6. An intelligent agricultural supply chain supervision method according to claim 1, wherein: in step S2, the data preprocessing is to perform data cleaning, data conversion and data set division on the collected historical agricultural supply chain supervision data; the data cleaning is to process missing values, repeated values and abnormal values of historical agricultural supply chain supervision data; the data conversion is to convert the cleaned data into a vector form and perform standardization processing based on a maximum and minimum normalization method; the data set division is to randomly divide the historical agricultural supply chain supervision data after data conversion into a training set and a testing set.