Agricultural supply chain demand prediction method and system based on digital driving

By slicing processing and feature representation of agricultural supply chain demand data, combining adjustable parameters and mobile weight adjustment mechanism, the problems of low accuracy of demand prediction and slow convergence of model in the existing technology are solved, and more efficient agriculture is achieved Supply chain demand forecasting and decision support.

CN119026875AActive Publication Date: 2024-11-26JILIN SCI & TECH INNOVATION PLATFORM MANAGEMENT CENT

Patent Information

Application Number
CN202411496084.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-25
Publication Date
2024-11-26
Estimated Expiration
2044-10-25

AI Technical Summary

Technical Problem

The existing agricultural supply chain demand prediction methods have poor ability to capture demand data and are allergic to outliers, resulting in low prediction accuracy; poor randomness and widespreadness when adjusting model parameters, resulting in slow convergence of training models and inability to adapt to demand fluctuations. This will affect the prediction efficiency.

Method used

By slicing the original data, serial operation based on feature representation, we introduce adaptive kernel functions and loss functions of adjustable parameters, and dynamically adjust the model training process; divide the parameter interval into sub-intervals, design the initialization strategy and mobile weight adjustment mechanism, Improve the adaptability and search efficiency of the model in the search space.

Benefits of technology

Improve the accuracy and adaptability of agricultural supply chain data, enhance the flexibility of the model in demand fluctuations, provide more accurate demand forecasts, and help supply chain managers make better decisions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026875A_ABST
    Figure CN119026875A_ABST
Patent Text Reader

Abstract

The invention discloses an agricultural supply chain demand prediction method and system based on digital driving. The method comprises the steps of data acquisition, data preprocessing, establishment of an agricultural supply chain demand prediction model, model parameter adjustment and agricultural supply chain demand prediction based on digital driving. The invention belongs to the field of demand prediction, and particularly relates to an agricultural supply chain demand prediction method and system based on digital driving, and the method comprises the steps: carrying out the slicing processing of original data, and carrying out the series operation based on feature representation; based on an adaptive kernel function and a loss function which introduce adjustable parameters, decision support of supply chain management is improved; a unique initialization strategy is designed by dividing a parameter interval into a plurality of sub-intervals, so that the demand capability of an agricultural supply chain is improved; the search efficiency is enhanced by a moving weight adjusting mechanism; through superior determination and calculation of the solution density, it is ensured that the model adapts to demand fluctuation in an agricultural supply chain, and a supply chain manager is helped to make a better decision.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of demand forecasting, and specifically to a method and system for forecasting agricultural supply chain demand based on digital drive. Background Art

[0002] The digital-driven agricultural supply chain demand forecasting method uses modern digital technology, data analysis tools and algorithms to predict the demand of each link in the agricultural supply chain to optimize the forecasting model of production, inventory, distribution and other decisions. This method combines technologies such as big data, artificial intelligence and machine learning to make more accurate demand forecasts by collecting, analyzing and processing data from the agricultural supply chain. However, the general agricultural supply chain demand forecasting method has the problem that the model has poor ability to capture agricultural supply chain demand data and is too sensitive to outliers, resulting in low demand forecasting accuracy; the general agricultural supply chain demand forecasting method has poor randomness and extensiveness when adjusting model parameters, and cannot flexibly adjust the movement of each individual in the search space, which leads to slow convergence of the training model and cannot adapt to the demand fluctuation data in the agricultural supply chain, resulting in low prediction efficiency. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a method and system for forecasting agricultural supply chain demand based on digital drive. In view of the problem that the general agricultural supply chain demand forecasting method has poor model capture ability of agricultural supply chain demand data and is too sensitive to outliers, resulting in low demand forecasting accuracy, this solution slices the original data, and improves the capture ability of time series information based on the series operation of feature representation; based on the introduction of adaptive kernel functions and loss functions with adjustable parameters, the model training process can be dynamically adjusted according to the prediction error, so that the model can flexibly respond to different types of demand fluctuations, thereby improving the prediction accuracy of agricultural supply chain data, being more adaptable to complex agricultural supply chain data, and enhancing the decision support for supply chain management; for general agricultural supply The chain demand forecasting method has the problems of poor randomness and extensiveness when adjusting the model parameters, and cannot flexibly adjust the movement of each individual in the search space, which leads to slow convergence of the training model and cannot adapt to the demand fluctuation data in the agricultural supply chain, resulting in low prediction efficiency. This scheme divides the parameter interval into multiple sub-intervals and designs a unique initialization strategy to ensure randomness and extensiveness in the search space, which helps to avoid local optimal solutions and improve the ability of agricultural supply chain demand; the adjustment mechanism of mobile weights improves the adaptability of the model in different parameter dimensions, enhances the search efficiency, and enables the model to better adapt to supply chain changes; through superior judgment and calculation of solution density, it ensures that the model adapts to demand fluctuations in the agricultural supply chain, thereby providing accurate agricultural supply chain demand forecasts and helping supply chain managers make better decisions.

[0004] The technical solution adopted by the present invention is as follows: The agricultural supply chain demand forecasting method based on digital drive provided by the present invention comprises the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data preprocessing;

[0007] Step S3: Establishing an agricultural supply chain demand forecasting model;

[0008] Step S4: model parameter adjustment;

[0009] Step S5: Agricultural supply chain demand forecasting based on digital drive.

[0010] Furthermore, in step S1, the data collection is to collect historical agricultural supply chain data; the historical agricultural supply chain data includes crop growth data, climate data, environmental data, soil data, time data and historical demand assessment levels; the historical demand assessment levels are used as data labels to obtain a time series data set.

[0011] Furthermore, in step S2, the data preprocessing is to perform data cleaning, data conversion and data set division on the collected data; the data cleaning is to process missing values, duplicate values ​​and outliers; the data conversion is to convert the data into vector form and perform standardization processing; the data set division is to divide the data set into a test set and a training set.

[0012] Further, in step S3, the establishment of the agricultural supply chain demand forecasting model specifically includes the following steps:

[0013] Step S31: Slicing: Slice the time series data. The slicing operation at each scale is as follows:

[0014] ;

[0015] Where X is the input time series data; P k is the data block after the k-th scale slice; k is the scale; S (p,k) is the slice width; S (s,k) is the slicing step size; Patching(·) is the slicing operation function;

[0016] Step S32: Feature integration: Use K time encoders to learn feature representation, and the output feature representation of each time encoder is: ; The output features of all time encoders are concatenated together and used as input features for prediction, expressed as: ; Among them, h 1 、h2 and h k are the output feature representations of the 1st, 2nd, and kth time encoders respectively; Encoder k is the kth temporal encoder; H is the concatenated feature representation;

[0017] Step S33: constructing an agricultural supply chain demand forecasting model architecture; the model architecture includes: an input layer for accepting feature representation of a data set; an LSTM layer for capturing the dynamic characteristics of a time series; a fully connected layer for outputting a forecast value; and an activation function, using a ReLU activation function;

[0018] Step S34: Designing a model loss function; specifically including:

[0019] Step S341: define an adaptive kernel function and introduce an adjustable parameter p. The formula used is as follows:

[0020] ;

[0021] Where k(·) is the adaptive kernel function; u and v are the built-in parameters of the kernel function;

[0022] Step S342: construct an adaptive loss function, the formula used is as follows:

[0023] ;

[0024] Where Lk(·) is the adaptive loss function; is the predicted label; y is the true label; is the adjustment parameter; c is the smoothing parameter;

[0025] Step S343: construct the final loss function, the formula used is as follows:

[0026] ;

[0027] Where L(·) is the loss function used by the model; α1 is the weight coefficient;

[0028] Step S35: model judgment; when the agricultural supply chain demand prediction model converges for the training set, the agricultural supply chain demand prediction model training is completed; a prediction threshold is set in advance, when the prediction accuracy of the trained agricultural supply chain demand prediction model for the test set is higher than the prediction threshold, the agricultural supply chain demand prediction model is established, otherwise the model parameters are adjusted, the data set is re-divided and the model is retrained.

[0029] Furthermore, in step S4, the model parameter adjustment specifically includes the following steps:

[0030] Step S41: Search initialization; based on slice width, slice step, adjustable parameter p, adjustment parameter , smoothing parameter c and weight coefficient α to establish parameter search space; divide the parameter interval of each dimension into n subintervals, initialize the individual position, and use the prediction accuracy of the agricultural supply chain demand prediction model trained based on the individual position on the test set as the individual fitness value; the formula used to initialize the individual position is as follows:

[0031] ;

[0032] In the formula, x id (·) is the position of the dth dimension of the i-th individual, 0 refers to the initialization position, i.e., the 0th iteration; UP da and LOW da are the starting and ending positions of the ath subinterval in the dth dimension respectively; k is the index of the partition interval; α is the parameter for adjusting the distribution of sampling points; m is the number of subintervals; n is the number of large intervals; m is the population size;

[0033] Step S42: Design the movement weight; the formula used is as follows:

[0034] ;

[0035] Where D is the total dimension of the search space; t is the number of iterations; x maxd and x mind are the maximum and minimum values ​​of the position of the dth dimension in the population; f i is the fitness value of the i-th individual; g d (·) is the individual's historical optimal position in the dth dimension; w id (·) is the moving weight of the dth dimension of the ith individual;

[0036] Step S43: Superiority determination; for each non-superior solution, calculate the distance from the optimal solution in each dimension and accumulate to obtain the solution density; non-superior solutions are those solutions that do not have any other solution that performs better on all objectives, that is, the positions of all dimensions are poor; sort the non-dominated solutions in descending order according to the solution density, and reinitialize the individual positions of 10% of the non-dominated solutions in each update;

[0037] Step S44: Location update; the formula used is as follows:

[0038] ;

[0039] Where b d (·) is the position of the optimal individual in the population in the dth dimension; c 1 is the global factor;

[0040] Step S45: search judgment; a fitness threshold is set in advance. When there is an individual fitness value higher than the fitness threshold, the agricultural supply chain demand forecasting model obtained based on the individual position is used as the established agricultural supply chain demand forecasting model; otherwise, if the maximum number of iterations is reached, return to step S41, otherwise return to step S42.

[0041] Furthermore, in step S5, the digitally driven agricultural supply chain demand forecasting is based on the established agricultural supply chain demand forecasting model, and real-time crop growth data, climate data, environmental data, soil data and time data are collected, and input into the agricultural supply chain demand forecasting model after preprocessing, and the demand assessment level output by the model is used as the agricultural supply chain demand forecasting result.

[0042] The agricultural supply chain demand forecasting system based on digital drive provided by the present invention comprises a data acquisition module, a data preprocessing module, an agricultural supply chain demand forecasting model establishment module, a model parameter adjustment module and an agricultural supply chain demand forecasting module;

[0043] The data collection module collects historical agricultural supply chain data and sends the data to the data preprocessing module;

[0044] The data preprocessing module performs data cleaning, data conversion and data set division on the collected data; and sends the data to the agricultural supply chain demand forecasting model establishment module;

[0045] The agricultural supply chain demand forecasting model building module slices the original data, based on the serial operation of the feature representation; based on the introduction of the adaptive kernel function and loss function with adjustable parameters, finally realizes the establishment of the agricultural supply chain demand forecasting model; and sends the data to the model parameter adjustment module;

[0046] The model parameter adjustment module divides the parameter interval into multiple sub-intervals to design a unique initialization strategy; designs a mobile weight adjustment mechanism; finally realizes the model parameter adjustment through superiority judgment and calculation of solution density; and sends the data to the agricultural supply chain demand forecasting module;

[0047] The agricultural supply chain demand forecasting module realizes agricultural supply chain demand forecasting based on the established agricultural supply chain demand forecasting model.

[0048] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0049] (1) In view of the problem that general agricultural supply chain demand forecasting methods have poor model capture ability of agricultural supply chain demand data and are too sensitive to outliers, resulting in low demand forecasting accuracy, this scheme slices the original data and improves the ability to capture time series information by serializing feature representations; based on the introduction of adaptive kernel functions and loss functions with adjustable parameters, the model training process can be dynamically adjusted according to the prediction error, so that the model can respond flexibly to different types of demand fluctuations, thereby improving the prediction accuracy of agricultural supply chain data, being more adaptable to complex agricultural supply chain data, and enhancing the decision support for supply chain management.

[0050] (2) The general agricultural supply chain demand forecasting method has the problems of poor randomness and extensiveness when adjusting model parameters, and cannot flexibly adjust the movement of each individual in the search space, which leads to slow convergence of the training model and inability to adapt to the demand fluctuation data in the agricultural supply chain, resulting in low prediction efficiency. This scheme divides the parameter interval into multiple sub-intervals and designs a unique initialization strategy to ensure randomness and extensiveness in the search space, which helps to avoid local optimal solutions and improve the ability of agricultural supply chain demand; the adjustment mechanism of mobile weights improves the adaptability of the model in different parameter dimensions, enhances the search efficiency, and enables the model to better adapt to supply chain changes; through superior judgment and calculation of solution density, it ensures that the model adapts to demand fluctuations in the agricultural supply chain, thereby providing accurate agricultural supply chain demand forecasts and helping supply chain managers make better decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic diagram of the flow chart of the agricultural supply chain demand forecasting method based on digital drive provided by the present invention;

[0052] Figure 2 A schematic diagram of a digitally driven agricultural supply chain demand forecasting system provided by the present invention;

[0053] Figure 3 is a schematic flow chart of step S3;

[0054] Figure 4 It is a schematic diagram of the process of step S4.

[0055] 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

[0056] 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.

[0057] In the description of the present invention, it is necessary to understand 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 referred system or element must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0058] Example 1, see Figure 1 The present invention provides a method for predicting agricultural supply chain demand based on digital drive, which comprises the following steps:

[0059] Step S1: Data collection, collecting historical agricultural supply chain data;

[0060] Step S2: Data preprocessing, data cleaning, data conversion and data set division of the collected data;

[0061] Step S3: Establishing an agricultural supply chain demand forecasting model, slicing the original data, and performing a series operation on the feature representation; introducing an adaptive kernel function and a loss function with adjustable parameters, and finally realizing the establishment of an agricultural supply chain demand forecasting model;

[0062] Step S4: Model parameter adjustment, divide the parameter interval into multiple sub-intervals to design a unique initialization strategy; design a mobile weight adjustment mechanism; finally achieve model parameter adjustment through superiority judgment and solution density calculation;

[0063] Step S5: Based on digital-driven agricultural supply chain demand forecasting, agricultural supply chain demand forecasting is realized based on the established agricultural supply chain demand forecasting model.

[0064] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the historical agricultural supply chain data includes crop growth data, climate data, environmental data, soil data, time data and historical demand assessment levels; the historical demand assessment levels are used as data labels to obtain a time series data set.

[0065] Example 3, see Figure 1This embodiment is based on the above embodiment. In step S2, data cleaning is to process missing values, duplicate values ​​and outliers; data conversion is to convert data into vector form and perform standardization; data set division is to divide the data set into a test set and a training set.

[0066] Example 4, see Figure 1 and Figure 3 Based on the above embodiment, in step S3, establishing an agricultural supply chain demand forecasting model specifically includes the following steps:

[0067] Step S31: Slicing: Slice the time series data. The slicing operation at each scale is as follows:

[0068] ;

[0069] Where X is the input time series data; P k is the data block after the k-th scale slice; k is the scale; S (p,k) is the slice width; S (s,k) is the slicing step size; Patching(·) is the slicing operation function;

[0070] Step S32: Feature integration: Use K time encoders to learn feature representation, and the output feature representation of each time encoder is: ; The output features of all time encoders are concatenated together and used as input features for prediction, expressed as: ; Among them, h 1 、h 2 and h k are the output feature representations of the 1st, 2nd, and kth time encoders respectively; Encoder k is the kth temporal encoder; H is the concatenated feature representation;

[0071] Step S33: constructing an agricultural supply chain demand forecasting model architecture; the model architecture includes: an input layer for accepting feature representation of a data set; an LSTM layer for capturing the dynamic characteristics of a time series; a fully connected layer for outputting a forecast value; and an activation function, using a ReLU activation function;

[0072] Step S34: Designing a model loss function; specifically including:

[0073] Step S341: define an adaptive kernel function and introduce an adjustable parameter p. The formula used is as follows:

[0074] ;

[0075] Where k(·) is the adaptive kernel function; u and v are the built-in parameters of the kernel function;

[0076] Step S342: construct an adaptive loss function, the formula used is as follows:

[0077] ;

[0078] Where Lk(·) is the adaptive loss function; is the predicted label; y is the true label; is the adjustment parameter; c is the smoothing parameter;

[0079] Step S343: construct the final loss function, the formula used is as follows:

[0080] ;

[0081] Where L(·) is the loss function used by the model; α1 is the weight coefficient;

[0082] Step S35: model judgment; when the agricultural supply chain demand prediction model converges for the training set, the agricultural supply chain demand prediction model training is completed; a prediction threshold is set in advance, when the prediction accuracy of the trained agricultural supply chain demand prediction model for the test set is higher than the prediction threshold, the agricultural supply chain demand prediction model is established, otherwise the model parameters are adjusted, the data set is re-divided and the model is retrained.

[0083] By performing the above operations, in view of the problem that the general agricultural supply chain demand forecasting method has poor model capture ability of agricultural supply chain demand data and is too sensitive to outliers, resulting in low demand forecasting accuracy, this solution slices the original data and improves the capture ability of time series information based on the series operation of feature representation; based on the introduction of adaptive kernel functions and loss functions with adjustable parameters, the model training process can be dynamically adjusted according to the prediction error, so that the model can flexibly respond to different types of demand fluctuations, thereby improving the prediction accuracy of agricultural supply chain data, being more adaptable to complex agricultural supply chain data, and enhancing the decision support for supply chain management.

[0084] Example 5, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S4, the model parameter adjustment specifically includes the following steps:

[0085] Step S41: Search initialization; based on slice width, slice step, adjustable parameter p, adjustment parameter , smoothing parameter c and weight coefficient α to establish parameter search space; divide the parameter interval of each dimension into n subintervals, initialize the individual position, and use the prediction accuracy of the agricultural supply chain demand prediction model trained based on the individual position on the test set as the individual fitness value; the formula used to initialize the individual position is as follows:

[0086] ;

[0087] In the formula, x id (·) is the position of the dth dimension of the i-th individual, 0 refers to the initialization position, i.e., the 0th iteration; UP da and LOW da are the starting and ending positions of the ath subinterval in the dth dimension respectively; k is the index of the partition interval; α is the parameter for adjusting the distribution of sampling points; m is the number of subintervals; n is the number of large intervals; m is the population size;

[0088] Step S42: Design the movement weight; the formula used is as follows:

[0089] ;

[0090] Where D is the total dimension of the search space; t is the number of iterations; x maxd and x mind are the maximum and minimum values ​​of the position of the dth dimension in the population; f i is the fitness value of the i-th individual; g d (·) is the individual's historical optimal position in the dth dimension; w id (·) is the moving weight of the dth dimension of the ith individual;

[0091] Step S43: Superiority determination; for each non-superior solution, calculate the distance from the optimal solution in each dimension and accumulate to obtain the solution density; non-superior solutions are those solutions that do not have any other solution that performs better on all objectives, that is, the positions of all dimensions are poor; sort the non-dominated solutions in descending order according to the solution density, and reinitialize the individual positions of 10% of the non-dominated solutions in each update;

[0092] Step S44: Location update; the formula used is as follows:

[0093] ;

[0094] Where b d (·) is the position of the optimal individual in the population in the dth dimension; c 1 is the global factor;

[0095] Step S45: search judgment; a fitness threshold is set in advance. When there is an individual fitness value higher than the fitness threshold, the agricultural supply chain demand forecasting model obtained based on the individual position is used as the established agricultural supply chain demand forecasting model; otherwise, if the maximum number of iterations is reached, return to step S41, otherwise return to step S42.

[0096] By performing the above operations, the general agricultural supply chain demand forecasting method has the problems of poor randomness and extensiveness when adjusting the model parameters, and cannot flexibly adjust the movement of each individual in the search space, which leads to slow convergence of the training model and cannot adapt to the demand fluctuation data in the agricultural supply chain, resulting in low prediction efficiency. This scheme divides the parameter interval into multiple sub-intervals and designs a unique initialization strategy to ensure randomness and extensiveness in the search space, which helps to avoid local optimal solutions and improve the ability of agricultural supply chain demand; the adjustment mechanism of mobile weights improves the adaptability of the model in different parameter dimensions, enhances the search efficiency, and enables the model to better adapt to supply chain changes; through superior judgment and calculation of solution density, it ensures that the model adapts to demand fluctuations in the agricultural supply chain, thereby providing accurate agricultural supply chain demand forecasts and helping supply chain managers make better decisions.

[0097] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the digitally driven agricultural supply chain demand forecasting is based on the established agricultural supply chain demand forecasting model, and real-time crop growth data, climate data, environmental data, soil data and time data are collected. After preprocessing, they are input into the agricultural supply chain demand forecasting model, and the demand assessment level output by the model is used as the agricultural supply chain demand forecasting result.

[0098] Embodiment 7, see Figure 2 , this embodiment is based on the above embodiment, the agricultural supply chain demand forecasting system based on digital drive provided by the present invention includes a data acquisition module, a data preprocessing module, an agricultural supply chain demand forecasting model establishment module, a model parameter adjustment module and an agricultural supply chain demand forecasting module;

[0099] The data collection module collects historical agricultural supply chain data and sends the data to the data preprocessing module;

[0100] The data preprocessing module performs data cleaning, data conversion and data set division on the collected data; and sends the data to the agricultural supply chain demand forecasting model establishment module;

[0101] The agricultural supply chain demand forecasting model building module slices the original data, based on the serial operation of the feature representation; based on the introduction of the adaptive kernel function and loss function with adjustable parameters, finally realizes the establishment of the agricultural supply chain demand forecasting model; and sends the data to the model parameter adjustment module;

[0102] The model parameter adjustment module divides the parameter interval into multiple sub-intervals to design a unique initialization strategy; designs a mobile weight adjustment mechanism; finally realizes the model parameter adjustment through superiority judgment and calculation of solution density; and sends the data to the agricultural supply chain demand forecasting module;

[0103] The agricultural supply chain demand forecasting module realizes agricultural supply chain demand forecasting based on the established agricultural supply chain demand forecasting model.

[0104] 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.

[0105] 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 variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0106] 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

1. A digital-driven agricultural supply chain demand forecasting method, characterized by: The method comprises the following steps: Step S1: Data collection, collecting historical agricultural supply chain data; Step S2: data preprocessing; Step S3: Establishing an agricultural supply chain demand forecasting model, slicing the original data, and performing a series operation on the feature representation; introducing an adaptive kernel function and a loss function with adjustable parameters, and finally realizing the establishment of an agricultural supply chain demand forecasting model; Step S4: Model parameter adjustment, divide the parameter interval into multiple sub-intervals to design a unique initialization strategy; design a mobile weight adjustment mechanism; finally achieve model parameter adjustment through superiority judgment and solution density calculation; Step S5: realizing agricultural supply chain demand forecasting based on digital drive and based on the established agricultural supply chain demand forecasting model; Step S3 includes step S32: feature integration; using K time encoders to learn feature representation, the output feature representation of each time encoder is: ; The output features of all time encoders are concatenated together and used as input features for prediction, expressed as: ; Among them, h1, h2 and h k are the output feature representations of the 1st, 2nd, and kth time encoders respectively; Encoder k is the kth temporal encoder; H is the concatenated feature representation; P k It is the data block after slicing at the kth scale.

2. The method for predicting agricultural supply chain demand based on digital drive according to claim 1 is characterized by: In step S3, the establishment of the agricultural supply chain demand forecasting model specifically includes the following steps: Step S31: Slicing: Slice the time series data. The slicing operation at each scale is as follows: ; Where X is the input time series data; P k is the data block after the k-th scale slice; k is the scale; S (p,k) is the slice width; S (s,k) is the slicing step size; Patching(·) is the slicing operation function; Step S32: feature integration; Step S33: constructing an agricultural supply chain demand forecasting model architecture; the model architecture includes: an input layer for accepting feature representation of a data set; an LSTM layer for capturing the dynamic characteristics of a time series; a fully connected layer for outputting a forecast value; and an activation function, using a ReLU activation function; Step S34: designing a model loss function; Step S35: model judgment; when the agricultural supply chain demand prediction model converges for the training set, the agricultural supply chain demand prediction model training is completed; a prediction threshold is set in advance, when the prediction accuracy of the trained agricultural supply chain demand prediction model for the test set is higher than the prediction threshold, the agricultural supply chain demand prediction model is established, otherwise the model parameters are adjusted, the data set is re-divided and the model is retrained.

3. The method for predicting agricultural supply chain demand based on digital drive according to claim 2 is characterized in that: In step S34, the design model loss function specifically includes: Step S341: define an adaptive kernel function and introduce an adjustable parameter p. The formula used is as follows: ; Where k(·) is the adaptive kernel function; u and v are the built-in parameters of the kernel function; Step S342: construct an adaptive loss function, the formula used is as follows: ; Where Lk(·) is the adaptive loss function; is the predicted label; y is the true label; is the adjustment parameter; c is the smoothing parameter; Step S343: construct the final loss function, the formula used is as follows: ; Where L(·) is the loss function used by the model; α1 is the weight coefficient.

4. The method for predicting agricultural supply chain demand based on digital drive according to claim 1 is characterized in that: In step S4, the model parameter adjustment specifically includes the following steps: Step S41: Search initialization; based on slice width, slice step, adjustable parameter p, adjustment parameter , smoothing parameter c and weight coefficient α to establish parameter search space; divide the parameter interval of each dimension into n subintervals, initialize the individual position, and use the prediction accuracy of the agricultural supply chain demand prediction model trained based on the individual position on the test set as the individual fitness value; the formula used to initialize the individual position is as follows: ; In the formula, x id (·) is the position of the dth dimension of the i-th individual, 0 refers to the initialization position, i.e., the 0th iteration; UP da and LOW da are the starting and ending positions of the ath subinterval in the dth dimension respectively; k is the index of the partition interval; α is the parameter for adjusting the distribution of sampling points; m is the number of subintervals; n is the number of large intervals; m is the population size; Step S42: Design the movement weight; the formula used is as follows: ; Where D is the total dimension of the search space; t is the number of iterations; x maxd and x mind are the maximum and minimum values ​​of the position of the dth dimension in the population; f i is the fitness value of the i-th individual; g d (·) is the individual's historical optimal position in the dth dimension; w id (·) is the moving weight of the dth dimension of the ith individual; Step S43: Superiority determination; for each non-superior solution, calculate the distance from the optimal solution in each dimension and accumulate to obtain the solution density; non-superior solutions are those solutions that do not have any other solution that performs better on all objectives, that is, the positions of all dimensions are poor; sort the non-dominated solutions in descending order according to the solution density, and reinitialize the individual positions of 10% of the non-dominated solutions in each update; Step S44: Location update; the formula used is as follows: ; Where b d (·) is the position of the optimal individual in the population in the dth dimension; c1 is the global factor; Step S45: search judgment; a fitness threshold is set in advance. When there is an individual fitness value higher than the fitness threshold, the agricultural supply chain demand forecasting model obtained based on the individual position is used as the established agricultural supply chain demand forecasting model; otherwise, if the maximum number of iterations is reached, return to step S41, otherwise return to step S42.

5. The method for predicting agricultural supply chain demand based on digital drive according to claim 1 is characterized in that: In step S1, the historical agricultural supply chain data includes crop growth data, climate data, environmental data, soil data, time data and historical demand assessment levels; the historical demand assessment levels are used as data labels to obtain a time series data set.

6. The method for predicting agricultural supply chain demand based on digital drive according to claim 1 is characterized by: In step S5, the digitally driven agricultural supply chain demand forecasting is based on the established agricultural supply chain demand forecasting model, and real-time crop growth data, climate data, environmental data, soil data and time data are collected. After preprocessing, the data are input into the agricultural supply chain demand forecasting model, and the demand assessment level output by the model is used as the agricultural supply chain demand forecasting result.

7. The method for predicting agricultural supply chain demand based on digital drive according to claim 1 is characterized in that: In step S2, the data preprocessing is to clean the collected data, convert the data and divide the data set; the data cleaning is to process missing values, duplicate values ​​and abnormal values; The data conversion is to convert the data into a vector form and perform standardization processing; the data set division is to divide the data set into a test set and a training set.

8. A digitally driven agricultural supply chain demand forecasting system, used to implement a digitally driven agricultural supply chain demand forecasting method as described in any one of claims 1 to 7, characterized in that: It includes a data collection module, a data preprocessing module, an agricultural supply chain demand forecasting model building module, a model parameter adjustment module and an agricultural supply chain demand forecasting module; The data collection module collects historical agricultural supply chain data and sends the data to the data preprocessing module; The data preprocessing module performs data cleaning, data conversion and data set division on the collected data; and sends the data to the agricultural supply chain demand forecasting model establishment module; The agricultural supply chain demand forecasting model building module slices the original data based on a series operation of feature representation; Based on the introduction of adaptive kernel functions and loss functions with adjustable parameters, the agricultural supply chain demand forecasting model is finally established; and the data is sent to the model parameter adjustment module; The model parameter adjustment module divides the parameter interval into multiple sub-intervals to design a unique initialization strategy; designs a mobile weight adjustment mechanism; finally realizes the model parameter adjustment through superiority judgment and calculation of solution density; and sends the data to the agricultural supply chain demand forecasting module; The agricultural supply chain demand forecasting module realizes agricultural supply chain demand forecasting based on the established agricultural supply chain demand forecasting model.

Citation Information

Patent Citations

  • Photovoltaic power generation ultra-short-term power interval prediction method based on Pareto optimum

    CN115618982A

  • Enterprise supply chain risk prediction method based on artificial intelligence

    CN118485302A

  • Tea production line multi-target process parameter optimization method based on improved particle swarm

    CN118627851A

Cited By

  • Agricultural supply chain demand analysis and prediction system based on digital driving

    CN119904263A