Grain emergency guarantee system optimization method, system and device based on deep learning

Abstract

The invention relates to the technical field of industrial big data management, and particularly discloses a grain emergency guarantee system optimization method, system and device based on deep learning, and the method comprises the steps: carrying out the data preprocessing of grain guarantee data through a grain emergency optimization device, and obtaining a grain guarantee data processing evaluation index; judging whether data preprocessing is carried out again or not to ensure that the data preprocessing achieves an expected effect; the grain emergency optimization device carries out feature extraction on the grain guarantee data after data preprocessing, judges whether feature extraction is carried out on the grain guarantee data again or not and records the feature extraction as grain guarantee feature parameters, a good data basis is provided for model training, and a grain guarantee system prediction model is generated through grain guarantee feature parameter training. According to the method, the prediction model is optimized, simulation prediction is performed, whether the prediction model is optimized or not is judged, the prediction model is stored in the grain emergency guarantee system, optimization of the prediction model in the grain emergency guarantee system is completed, and the effectiveness and feasibility of the grain emergency guarantee system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial big data management, and specifically to an optimization method, system and device for the grain emergency guarantee system based on deep learning. Background Art

[0002] With the full promotion of the marketization of grain purchase and sale, profound changes have taken place in the field of grain circulation; the trend that grain production, circulation and reserves are subject to dual constraints of nature and market has become more obvious; some major natural disasters, major accident disasters, major public health events and social security events still occur from time to time, and these may all become uncertain factors triggering fluctuations in the grain market. Therefore, improving the grain emergency guarantee system and enhancing the speed and efficiency of grain emergency response are necessary to adapt to the development of grain circulation work and ensure grain security, and the optimization method for the grain emergency guarantee system has emerged as the times require.

[0003] For example, the invention patent with the publication number CN117078102B discloses a quantitative evaluation method for the regional grain security guarantee ability based on spatial matching degree, which is based on the collection and induction of basic data of multi-sector grain security; processes the grain security data set, and performs format recognition, image recognition, classification extraction and other processing operations on multi-source and multi-format data to establish a grain security data resource pool; designs a spatial data generator, extracts data from the grain security data resource pool for spatial processing, constructs a spatial database and attaches attribute information to the spatial data; at the same time, constructs an index evaluation system and conducts calculations. In terms of the spatial aggregation degree index, it considers the guaranteed population, grain storage scale and grain processing capacity, and in terms of the spatial matching degree index, it considers multiple dimensions of population, storage, processing, emergency and transportation.

[0004] For example, the invention patent with the publication number CN117933885A discloses an intelligent dynamic early warning method and system for grain inventory assets. The method includes obtaining price data in a grain price database and purchase and sale data of a grain storage ledger; training a grain inventory asset early warning model based on the price data and the purchase and sale data to optimize dynamic warning parameters; performing periodic early warning calculations based on the optimized grain inventory asset early warning model to generate an early warning state, and sending a warning notice in response to the early warning state reaching the warning state.

[0005] Combined with the above technical solutions, it is found that in the current process of optimizing the grain emergency guarantee system, it is usually defaulted that each step of the system construction is accurate. This situation will cause a large error between the constructed grain emergency guarantee system and the required grain emergency guarantee system, reduce the accuracy of the grain emergency guarantee system, and it is difficult to cope with complex and changeable grain emergency situations, thereby increasing the grain security risk and ultimately resulting in the inability to guarantee the feasibility and effectiveness of the grain emergency guarantee system. Summary of the Invention

[0006] Aiming at the deficiencies of the prior art, the present invention provides an optimization method, system and device for the food emergency guarantee system based on deep learning, which can effectively solve the problems involved in the above-mentioned background technology.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: In the first aspect of the present invention, an optimization method for the food emergency guarantee system based on deep learning is provided, including: S1. Obtain food guarantee data through the food emergency optimization device, perform data preprocessing on the food guarantee data according to a preset method to obtain data processing information during the data preprocessing process, analyze to obtain a food guarantee data processing evaluation index, and compare it with the food guarantee data processing evaluation threshold preset in the food optimization library to determine whether to perform data preprocessing on the food guarantee data again; S2. Use the deep learning algorithm preset in the food emergency optimization device to extract features from the food guarantee data after data preprocessing to obtain feature extraction-related data, thereby analyzing to obtain a food guarantee data feature value, and compare it with the food guarantee data feature threshold preset in the food optimization library to determine whether to re-extract features from the food guarantee data after data preprocessing; S3. Record the food guarantee data after feature extraction by the food emergency optimization device as food guarantee feature parameters, input the food guarantee feature parameters into a preset prediction model for training to generate a food guarantee system prediction model, perform simulation prediction through the food guarantee system prediction model to obtain simulation test parameters, thereby analyzing to obtain a simulation test evaluation index, and compare it with the simulation test evaluation threshold preset in the food optimization library to determine whether to optimize the food guarantee system prediction model, and store the food guarantee system prediction model in the food emergency guarantee system to complete the optimization of the prediction model in the food emergency guarantee system.

[0008] As a further method, the specific analysis process of the food guarantee data processing evaluation index is as follows: The food guarantee data processing evaluation index is comprehensively analyzed through the data signal smoothness of the food guarantee data during the data preprocessing period, the data signal volatility of the food guarantee data during the data preprocessing period, the data error rate of the food guarantee data during the data preprocessing period, the data consistency check passing rate of the food guarantee data during the data preprocessing period, and the data filling rate of the food guarantee data during the data preprocessing period. The food guarantee data processing evaluation index is used to comprehensively and quantitatively evaluate the data preprocessing effect of the food guarantee data. The food guarantee data processing evaluation index represents the degree of influence of the data signal smoothness, data signal volatility, data error rate, data consistency check passing rate, and data filling rate on the data preprocessing effect of the food guarantee data to quantify the data.

[0009] As a further method, the process of determining whether to perform data preprocessing on food security data again is as follows: If the food security data processing evaluation index is greater than or equal to the food security data processing evaluation threshold preset in the food optimization library, there is no need to perform data preprocessing on the food security data again.

[0010] If the food security data processing evaluation index is less than the food security data processing evaluation threshold preset in the food optimization library, it is necessary to perform data preprocessing on the food security data again.

[0011] The process of performing data preprocessing on the food security data again specifically refers to first standardizing the food security data into the standard normal distribution to which the data processing belongs through the Z-score formula, and then performing data preprocessing on the food security data through a preset method.

[0012] As a further method, the process of analyzing the eigenvalue of the food security data is as follows: The data related to feature extraction specifically includes the total duration of the feature extraction period, the actual feature extraction duration of the food security data in the feature extraction period, the number of correctly extracted features of the food security data in the feature extraction period, the total number of features of the food security data in the feature extraction period, and the Pearson correlation coefficient of each food security data in the feature extraction period.

[0013] The ratio of the actual feature extraction duration of the food security data in the feature extraction period to the total duration of the feature extraction period is processed to obtain the proportion of the feature extraction duration of the food security data in the feature extraction period.

[0014] The ratio of the number of correctly extracted features of the food security data in the feature extraction period to the total number of features of the food security data in the feature extraction period is processed to obtain the feature extraction accuracy of the food security data in the feature extraction period.

[0015] The Pearson correlation coefficient of each food security data in the feature extraction period is compared with the Pearson correlation coefficient threshold preset in the food optimization library, and the number of several food security data with a Pearson correlation coefficient greater than the Pearson correlation coefficient threshold is counted, which is recorded as the redundant data volume of the food security data in the feature extraction period. The ratio of the redundant data volume of the food security data in the feature extraction period to the total data volume of the food security data is processed to obtain the feature redundancy of the food security data in the feature extraction period.

[0016] The eigenvalue of food security data is obtained through comprehensive analysis of the proportion of the feature extraction duration of food security data in the feature extraction period, the feature extraction accuracy of food security data in the feature extraction period, the feature redundancy of food security data in the feature extraction period, and the food security data processing evaluation index. The eigenvalue of food security data is used to comprehensively and quantitatively evaluate the feature extraction effect of food security data. The eigenvalue of food security data represents the degree of influence of the proportion of feature extraction duration, feature extraction accuracy, feature redundancy, and food security data processing evaluation index on the feature extraction effect of food security data, which is a quantitative data.

[0017] As a further method, the process of determining whether to re-extract features from the preprocessed food security data is as follows: If the eigenvalue of food security data is greater than or equal to the food security data feature threshold preset in the food optimization library, there is no need to re-extract features from the preprocessed food security data.

[0018] If the eigenvalue of food security data is less than the food security data feature threshold preset in the food optimization library, it is necessary to re-extract features from the preprocessed food security data.

[0019] The process of re-extracting features from the preprocessed food security data is as follows: specifically, the difference between the eigenvalue of food security data and the food security data feature threshold is calculated to obtain the food security data feature deviation value. Then, the food security data feature deviation value is matched with the feature extraction algorithm optimization set corresponding to each food security data feature deviation value interval preset in the food optimization library, so as to obtain the feature extraction algorithm optimization set corresponding to the food security data feature deviation value. According to the feature extraction algorithm optimization set, the feature extraction algorithm is optimized to obtain the second feature extraction algorithm. Finally, the preprocessed food security data is re-extracted using the second feature extraction algorithm.

[0020] As a further method, the specific analysis process of the simulation test evaluation index is as follows: The simulation test parameters specifically include the total data volume of food security feature parameters, the accurate prediction volume of food security feature parameters by the food security system prediction model in the simulation test period, the simulation test response duration of the food security system prediction model in the simulation test period, and the deviation degree of food demand prediction by the food security system prediction model in the simulation test period.

[0021] The ratio of the accurate prediction volume of food security feature parameters by the food security system prediction model in the simulation test period to the total data volume of food security feature parameters is calculated to obtain the prediction coverage rate of the food security system prediction model in the simulation test period.

[0022] Through the prediction coverage rate of the food security system prediction model during the simulation test period, the simulation test response duration of the food security system prediction model during the simulation test period, the deviation of food demand prediction of the food security system prediction model during the simulation test period, the food security data processing evaluation index, and the food security data characteristic value, the simulation test evaluation index is comprehensively analyzed. The simulation test evaluation index is used to comprehensively and quantitatively evaluate the simulation test effect of the food security system prediction model. The simulation test evaluation index represents the quantitative data of the combined influence degree of the prediction coverage rate, the simulation test response duration, the deviation of food demand prediction, the food security data processing evaluation index, and the food security data characteristic value on the simulation test effect of the food security system prediction model.

[0023] As a further method, the process of determining whether to optimize the food security system prediction model is as follows: If the simulation test evaluation index is greater than or equal to the preset simulation test evaluation threshold in the food optimization library, there is no need to optimize the food security system prediction model.

[0024] If the simulation test evaluation index is less than the preset simulation test evaluation threshold in the food optimization library, the food security system prediction model needs to be optimized.

[0025] The process of optimizing the food security system prediction model is as follows: specifically, the simulation test evaluation index is matched with the prediction model optimization set corresponding to each simulation test evaluation index interval preset in the food optimization library, so as to obtain the prediction model optimization set corresponding to the simulation test evaluation index, and the food security system prediction model is optimized according to the prediction model optimization set.

[0026] The second aspect of the present invention provides a food emergency security system optimization system based on deep learning, including: a food security data preprocessing module, which is used to obtain food security data through a food emergency optimization device, perform data preprocessing on the food security data according to a preset method to obtain data processing information during the data preprocessing process, analyze to obtain a food security data processing evaluation index, and compare it with the preset food security data processing evaluation threshold in the food optimization library to determine whether to perform data preprocessing on the food security data again.

[0027] A food security data feature extraction module, which is used to extract features from the food security data after data preprocessing through a deep learning algorithm preset by the food emergency optimization device, obtain feature extraction-related data, and thus analyze to obtain a food security data characteristic value, and compare it with the preset food security data characteristic threshold in the food optimization library to determine whether to re-extract features from the food security data after data preprocessing.

[0028] The grain emergency guarantee system optimization module is used to record the grain guarantee data after feature extraction as grain guarantee feature parameters through the grain emergency optimization device, input the grain guarantee feature parameters into a preset prediction model for training to generate a grain guarantee system prediction model, conduct simulation prediction through the grain guarantee system prediction model to obtain simulation test parameters, thereby analyze and obtain simulation test evaluation indicators, compare with the simulation test evaluation threshold preset in the grain optimization library to determine whether to optimize the grain guarantee system prediction model, and store the grain guarantee system prediction model in the grain emergency guarantee system to complete the optimization of the prediction model in the grain emergency guarantee system.

[0029] The third aspect of the present invention provides a device for the optimization method of the grain emergency guarantee system based on deep learning, characterized in that the device includes a memory for storing computer program instructions and a processor for executing the program instructions. Wherein, when the computer program instructions are executed by the processor, the device is triggered to execute the above-mentioned method.

[0030] Compared with the prior art, the embodiments of the present invention at least have the following advantages or beneficial effects: (1) The present invention provides an optimization method, system and device for the grain emergency guarantee system based on deep learning. The grain emergency optimization device performs data preprocessing on the grain guarantee data and obtains a grain guarantee data processing evaluation index to determine whether to perform data preprocessing again to ensure that the data preprocessing achieves the expected effect and provides more standardized data for feature extraction. The grain emergency optimization device extracts features from the grain guarantee data after data preprocessing, determines whether to re-extract its features, and records it as grain guarantee feature parameters, providing a good data basis for model training. A grain guarantee system prediction model is generated through training with the grain guarantee feature parameters and simulation prediction is carried out to determine whether to optimize it, ensuring the accuracy of model construction, guaranteeing the reliability of its simulation effect, and storing it in the grain emergency guarantee system to complete the optimization of the prediction model in the grain emergency guarantee system, improving the effectiveness and feasibility of the grain emergency guarantee system.

[0031] (2) The present invention obtains grain guarantee data through the grain emergency optimization device, performs data preprocessing on the grain guarantee data according to a preset method to obtain data processing information during the data preprocessing process, can comprehensively understand the data preprocessing effect of the grain guarantee data, analyze and obtain a grain guarantee data processing evaluation index, can quantitatively evaluate the data preprocessing effect of the grain guarantee data, and compare with the grain guarantee data processing evaluation threshold preset in the grain optimization library to determine whether to perform data preprocessing on the grain guarantee data again, improving the accuracy of data preprocessing to ensure that the data quality meets the high-standard requirements of subsequent feature extraction.

[0032] (3) Through the preset deep learning algorithm, the food emergency optimization device of the present invention extracts features from the food security data after data preprocessing to obtain data related to feature extraction, enabling a comprehensive understanding of the feature extraction effect of the food security data. From this, the eigenvalue of the food security data can be analyzed, the feature extraction effect of the food security data can be quantitatively evaluated, and compared with the preset food security data feature threshold in the food optimization library to determine whether to re-extract features from the food security data after data preprocessing, so that the feature extraction effect reaches the expected level, providing reliable data support for the construction of the subsequent food security system prediction model.

[0033] (4) The food emergency optimization device of the present invention records the food security data after feature extraction as food security feature parameters, and inputs the food security feature parameters into a preset prediction model for training to generate a food security system prediction model. Through the simulation prediction of the food security system prediction model, the simulation prediction results of the food security system prediction model can be scientifically evaluated to obtain simulation test parameters. From this, the simulation test evaluation index can be analyzed, enabling a comprehensive quantitative evaluation of the simulation test effect of the food security system prediction model, and compared with the preset simulation test evaluation threshold in the food optimization library to determine whether to optimize the food security system prediction model, improving the accuracy and practicality of the food security system prediction model, and storing the food security system prediction model in the food emergency security system to complete the optimization of the prediction model in the food emergency security system, providing a scientific basis for the food security guarantee decision of the food emergency security system, and improving the accuracy and reliability of the food emergency security system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to the following drawings.

[0035] Figure 1 It is a schematic flowchart of the method steps of the present invention.

[0036] Figure 2 It is a schematic diagram of the connection of system modules of the present invention.

[0037] Figure 3 It is a curve of the amplitude change of the data signal involved in the present invention.

[0038] Reference numerals in the drawings: 1, highest position point; 2, lowest position point; 3, curve change position point. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0040] Referring to Figure 1 As shown, the first aspect of the present invention provides an optimization method for the food emergency guarantee system based on deep learning, including: S1. Obtain food guarantee data through a food emergency optimization device, perform data preprocessing on the food guarantee data according to a preset method to obtain data processing information during the data preprocessing, analyze to obtain a food guarantee data processing evaluation index, and compare it with a food guarantee data processing evaluation threshold preset in the food optimization library to determine whether to perform data preprocessing on the food guarantee data again.

[0041] It should be noted that the above-mentioned food guarantee data specifically includes food reserve data, such as the quantity, variety and distribution of reserve grains, the update cycle and quality inspection situation of reserve grains; food production data, such as the planting area, yield and yield fluctuation of food crops; food circulation data, such as the transfer situation of food purchase, sale, transportation and storage, and the supply and demand changes in the food market; food consumption data, such as the impact of population quantity and structure on food demand; food emergency response data, such as the quantity and distribution of emergency processing, storage and transportation, distribution and supply enterprises; food quality and safety data, such as food quality inspection standards and results, and the monitoring and treatment of food pesticide residues; food market information data, such as the quantity, variety and price of food imports and exports.

[0042] In a specific embodiment, the above-mentioned food emergency optimization device specifically refers to a high-performance computer integrating data processing, feature extraction, model training, and simulation prediction functions, which is used for data processing and prediction model optimization in the food emergency support system to improve the efficiency and accuracy of food emergency response. It mainly includes a data processing module, a feature extraction module, and a model training and simulation prediction module. Among them, the data processing module performs preprocessing operations such as cleaning and conversion on food security data through the Pandas library in Python. The feature extraction module uses the deep learning algorithm recurrent neural network (RNN) in the machine learning framework PyTorch to perform feature extraction operations on food security data. The model training and simulation prediction module uses the time series model LSTM (long short-term memory network) based on deep learning in the machine learning framework PyTorch to implement model training, generate a food security system prediction model, and complete simulation prediction through the food security system prediction model. The above-mentioned data preprocessing of food security data according to a preset method specifically refers to using K-nearest neighbor imputation in the Pandas library in Python to process missing values in food security data. The specific process is to collect complete food security data, mark the samples containing missing values, set the K value to 5, and for the samples containing missing values, use the Euclidean distance method to calculate the distance between it and other samples in the food security data. According to the calculated distance, select the K samples closest to the sample containing missing values as the nearest neighbors. For the feature containing missing values, use the median of this feature in the nearest neighbor samples to fill the missing values, and complete the processing of missing values in food security data. Use the isolation forest algorithm to process outliers in food security data. The specific process is to set the number of trees to 100, the subsample size to 256, and the outlier proportion estimate to 0.05. Use the isolation forest algorithm to predict outliers in food security data and judge whether each food security data is an outlier. Specifically, for each food security data, calculate its average path length in all isolation trees. According to the average path length of all data points, the algorithm will calculate a threshold based on the outlier proportion estimate. If the average path length of a certain food security data is less than or equal to this threshold, then this food security data is considered an outlier; otherwise, it is considered a normal point. The prediction result will return an array, where -1 represents an outlier and 1 represents a normal point. Correct the data of the outliers to complete the processing of outliers in food security data, and finally complete the data preprocessing.

[0043] Specifically, the data processing information specifically includes the total data volume of food security data, the error values of each food security data during the data preprocessing period, the amount of data passing the consistency check of food security data during the data preprocessing period, the total number of missing values of food security data during the data preprocessing period, and the number of filled values to which the missing values of food security data belong during the data preprocessing period.

[0044] It should be elaborated that the above-mentioned data preprocessing period specifically refers to the time period from when the food emergency optimization device starts to perform data preprocessing on food security data to the end of the processing. The specific duration is jointly determined by factors such as the difficulty of the data preprocessing algorithm and the quantity of food security data, and can be obtained through the time module of the timing function in Python.

[0045] It should be elaborated that the total data volume of the above-mentioned food security data, the error values of each food security data during the data preprocessing period, the amount of data passing the consistency check of food security data during the data preprocessing period, the total number of missing values of food security data during the data preprocessing period, and the number of filled values to which the missing values of food security data belong during the data preprocessing period can all be obtained through the pandas library in Python; among them, the total data volume of food security data specifically refers to the total number of food security data; the error value of each food security data during the data preprocessing period specifically refers to the difference between the actual value of the food security data and the value after preprocessing due to various reasons such as data cleaning errors and data conversion errors during the data preprocessing process; the amount of data passing the consistency check of food security data during the data preprocessing period specifically refers to the number of data records that are confirmed to meet the data consistency requirements and are recognized as correct after passing the consistency check during the preprocessing of food security data; the total number of missing values of food security data during the data preprocessing period specifically refers to the total number of missing data items in food security data due to various reasons such as incomplete data collection, record loss, and data entry errors during the data preprocessing process; the number of filled values to which the missing values of food security data belong during the data preprocessing period specifically refers to the quantity of data filled by the filling method adopted for the missing values existing in food security data during the data preprocessing process.

[0046] Obtain the data signal amplitude change curve of food security data during the data preprocessing period, extract the signal amplitudes corresponding to the curve change position points from the data signal amplitude change curve of food security data during the data preprocessing period, perform difference processing on the signal amplitudes corresponding to adjacent curve change position points, and perform sum-of-squares processing on the difference processing results to obtain the data signal smoothness of food security data during the data preprocessing period.

[0047] It should be elaborated that the curve of the data signal amplitude change of the above-mentioned food security data during the data preprocessing period can be obtained through a Savitzky-Golay filter; the smoothness of the data signal of the above-mentioned food security data during the data preprocessing period specifically refers to the smoothness of the curve of the data signal amplitude change of the food security data during the data preprocessing period.

[0048] According to Figure 3 shown, it is the curve of the data signal amplitude change of the food security data during the data preprocessing period. It can be known from the figure that the abscissa is the data preprocessing time, with the unit of millisecond, and the ordinate is the signal amplitude, with the unit of milliampere; among them, the change position points of each curve are several curve change position points 3 in the curve of the data signal amplitude change. By obtaining the signal amplitudes corresponding to each curve change position point from the curve of the data signal amplitude change, performing difference processing on the signal amplitudes corresponding to adjacent curve change position points, and performing sum-of-squares processing on the difference processing results, the smoothness of the data signal of the food security data during the data preprocessing period is obtained; among them, in this embodiment, it is not limited to Figure 3 the limited number of curve change position points 3 clearly marked in the shown figure. These marked points are actually only used as exemplary representatives. In fact, a large number of points are selected for calculation and analysis. When selecting research points, randomly select the position points where the curve changes for research. These points can better reflect the data preprocessing effect performance of the signal amplitude under different conditions.

[0049] Extract the signal amplitudes corresponding to the highest position point and the lowest position point from the curve of the data signal amplitude change of the food security data during the data preprocessing period, and perform difference processing and absolute value processing in sequence to obtain the data signal volatility of the food security data during the data preprocessing period.

[0050] According to Figure 3 shown, for the signal amplitudes corresponding to the highest position point 1 and the lowest position point 2 of the curve of the data signal amplitude in the figure, perform difference processing and absolute value processing in sequence to obtain the data signal volatility of the food security data during the data preprocessing period.

[0051] It should be elaborated that the data signal volatility of the above-mentioned food security data during the data preprocessing period specifically refers to the change amplitude of the food security data from the highest position point to the lowest position point during the data preprocessing period, which can be obtained by calculating the difference between the signal amplitudes corresponding to the highest position point and the lowest position point and taking the absolute value.

[0052] Square the error values of each food security data during the data preprocessing period, and perform mean processing on the squared processing results to obtain the data error rate of the food security data during the data preprocessing period.

[0053] The ratio of the amount of data passing the consistency check during the data preprocessing period of the food security data to the total amount of the food security data is processed to obtain the data consistency check passing rate of the food security data during the data preprocessing period.

[0054] The ratio of the number of filled missing values of the food security data during the data preprocessing period to the total number of missing values of the food security data during the data preprocessing period is processed to obtain the data filling rate of the food security data during the data preprocessing period.

[0055] It should be elaborated that the data error rate of the above-mentioned food security data during the data preprocessing period specifically refers to a quantitative representation of the degree of difference between the preprocessed data and the original data during the preprocessing of the food security data; the data consistency check passing rate of the above-mentioned food security data during the data preprocessing period specifically refers to the proportion of data records that meet the consistency requirements in the total amount of the food security data when performing a consistency check on the food security data during the data preprocessing stage; the data filling rate of the above-mentioned food security data during the data preprocessing period specifically refers to the proportion of the filled data volume to the original missing data volume after filling the existing missing or incomplete data during the data preprocessing process.

[0056] Furthermore, the specific analysis process of the food security data processing evaluation index is as follows: The food security data processing evaluation index is comprehensively analyzed through the data signal smoothness of the food security data during the data preprocessing period, the data signal volatility of the food security data during the data preprocessing period, the data error rate of the food security data during the data preprocessing period, the data consistency check passing rate of the food security data during the data preprocessing period, and the data filling rate of the food security data during the data preprocessing period. The food security data processing evaluation index is used to comprehensively and quantitatively evaluate the data preprocessing effect of the food security data. The food security data processing evaluation index represents the degree of influence of the data signal smoothness, data signal volatility, data error rate, data consistency check passing rate, and data filling rate on the data preprocessing effect of the food security data. The specific acquisition method is as follows: In the formula, represents the data signal smoothness of the food security data during the data preprocessing period, represents the data signal volatility of the food security data during the data preprocessing period, represents the data error rate of the food security data during the data preprocessing period, represents the data consistency check passing rate of the food security data during the data preprocessing period, Indicates the data filling rate of food security data during the data preprocessing period, where e is the natural constant, Indicates the data processing evaluation index of food security data. The data signal smoothness measures the smoothness of data changes in the time series. When the data signal smoothness is high, the data signal volatility is usually low. When the data signal smoothness is high and the data signal volatility is low, the data stability is high, and the error in the preprocessing process is correspondingly reduced, resulting in a small data error rate and an increase in the food security data processing evaluation index. The data consistency check passing rate measures the degree of data consistency maintained during data preprocessing. When the data filling rate is high, it indicates that missing data has been effectively filled, which helps to maintain data consistency. The higher the data consistency check passing rate, the greater the increase in the food security data processing evaluation index.

[0057] Indicates the data processing influence coefficient of the data signal smoothness preset in the food optimization library, Indicates the data processing influence coefficient of the data signal volatility preset in the food optimization library, Indicates the data processing influence coefficient of the data error rate preset in the food optimization library, Indicates the data processing influence coefficient of the data consistency check passing rate preset in the food optimization library, Indicates the data processing influence coefficient of the data filling rate preset in the food optimization library.

[0058] It should be noted that the data processing influence coefficients corresponding to the data signal smoothness, data signal volatility, data error rate, data consistency check passing rate, and data filling rate are respectively used to indicate the importance of the data signal smoothness, data signal volatility, data error rate, data consistency check passing rate, and data filling rate in analyzing the food security data processing evaluation index. For example, there is a pre-set mapping relationship between the real-time data processing information and the corresponding data processing influence coefficient in the food optimization library. Through the pre-set mapping relationship, the data processing influence coefficient corresponding to the real-time data processing information can be obtained. The data signal smoothness, data signal volatility, data error rate, data consistency check passing rate, and data filling rate are respectively matched with the pre-set mapping relationship to obtain the data processing influence coefficient corresponding to the data signal smoothness, the data processing influence coefficient corresponding to the data signal volatility, the data processing influence coefficient corresponding to the data error rate, the data processing influence coefficient corresponding to the data consistency check passing rate, and the data processing influence coefficient corresponding to the data filling rate. In this embodiment, the value range is (0, 1).

[0059] Specifically, the process of determining whether to perform data preprocessing on food security data again is as follows: If the food security data processing evaluation index is greater than or equal to the food security data processing evaluation threshold preset in the food optimization library, there is no need to perform data preprocessing on the food security data again; the food security data processing evaluation threshold specifically refers to the critical value used to judge whether the data preprocessing effect of the food security data meets the qualified standard during the evaluation of the data preprocessing effect of the food security data, and is used to measure whether the data preprocessing effect of the food security data meets the qualified standard; the above food security data processing evaluation index being greater than or equal to the food security data processing evaluation threshold preset in the food optimization library indicates that the data preprocessing effect of the food security data is qualified, and there is no need to perform data preprocessing on the food security data again.

[0060] If the food security data processing evaluation index is less than the food security data processing evaluation threshold preset in the food optimization library, it indicates that the data preprocessing effect of the food security data is unqualified, and it is necessary to perform data preprocessing on the food security data again.

[0061] The above-mentioned data preprocessing on the food security data again specifically means first standardizing the food security data into the standard normal distribution to which the data processing belongs through the Z-score formula, and then performing data preprocessing on the food security data through a preset method.

[0062] It should be elaborated that the above-mentioned data processing belongs to the standard normal distribution, specifically the standard normal distribution with a mean of 0 and a variance of 1. The specific process of standardizing the food security data into the standard normal distribution to which the data processing belongs by the above Z-score formula is as follows: First, calculate the mean μ and standard deviation σ of the food security data, then calculate the difference between each food security data X and the mean μ (X - μ), and divide this difference by the standard deviation σ to obtain the Z-score of each food security data X, completing the data standardization of the food security data. Among them, by using the Z-score formula to standardize the food security data into the standard normal distribution to which the data processing belongs, after standardization, the data distribution is more uniform, and the algorithm can better capture the patterns in the data, improving the processing effect of data preprocessing. Many preprocessing methods, such as outlier detection, rely on the distribution characteristics of the original data. If standardized at the beginning, it will mask the true characteristics of the data, such as extreme values and skewed distributions, resulting in the loss of key information. For example, if there are outliers caused by dimensional differences in the original data, such as the unit of food production in one region is "tons" and in another region is "kilograms", directly standardizing will "average out" these outliers, causing subsequent outlier detection to fail. If it is evaluated that the preprocessing effect does not meet the standard, it means that the original data distribution does not conform to the assumptions of the subsequent data preprocessing algorithm, that is, linear regression requires the data to be normally distributed. At this time, through Z-score standardization, the data distribution can be forcibly adjusted, the dimensional difference can be eliminated, the algorithm performance can be improved, and after eliminating the dimensional difference, the data preprocessing will be more accurate. To sum up, the initial preprocessing is to give priority to solving data quality problems, such as noise and missing values, and retain the original distribution characteristics. The food security data processing evaluation index is used to judge whether the data distribution needs to be adjusted. If the evaluation is qualified after the first preprocessing, it means that the data quality problem has been solved. If the evaluation is unqualified, then the distribution problem is solved through standardization, avoiding blind operations, introducing standardization only when necessary, reducing unnecessary computational costs, and the secondary standardization preprocessing can accurately optimize the distribution problem and improve the algorithm compatibility. The phased processing is to balance the dual goals of "data quality repair" and "data distribution optimization", solve the basic problems first, and then optimize them specifically, avoiding information loss or computational waste caused by premature standardization, being more adaptable, and can also dynamically adjust the process according to actual needs, taking into account flexibility and efficiency.

[0063] S2. Use the preset deep learning algorithm of the food emergency optimization device to extract features from the preprocessed food security data, obtain the data related to feature extraction, and thus analyze and obtain the food security data feature values, and compare them with the preset food security data feature thresholds in the food optimization library to judge whether to re-extract features from the preprocessed food security data.

[0064] In a specific embodiment, the above-mentioned preset deep learning algorithm specifically refers to a recurrent neural network (RNN); the specific analysis process of feature extraction from the preprocessed food security data is to convert the preprocessed food security data into a format suitable for input into a deep learning model. For example, time series data is converted into sequence segments of a fixed length, and multi-dimensional data such as weather conditions is encoded as a numerical vector or matrix. Subsequently, the food security data after format conversion is input into the deep learning model. The input layer of the model receives the food security data after format conversion, and the food security data after format conversion is passed layer by layer through the hidden layer in the RNN model. The neurons in the hidden layer will learn the internal laws and features of the data, and these features include trends, seasonal variations, periodic fluctuations, etc.; In each convolutional layer of the model, the weights and bias parameters of the convolutional kernel perform a convolutional operation on the food security data; in each layer, the data is subjected to a non-linear transformation through the ReLU activation function. After the forward propagation of the model, the extracted features are represented as points or vectors in a high-dimensional space. These feature vectors contain useful information extracted from the original food security data, such as trend features, seasonal features, policy impact features, etc. According to the error between the prediction result of the model and the true label, the gradient is calculated through the backpropagation algorithm, and the weights and bias parameters of the model are updated. This process is carried out iteratively until the performance of the model reaches a satisfactory level, and the output feature vector is denoted as the food security feature parameter.

[0065] Specifically, the specific analysis process of the food security data eigenvalue is as follows: The data related to feature extraction specifically includes the total duration of the feature extraction period, the actual feature extraction duration of the food security data during the feature extraction period, the number of correctly extracted features of the food security data during the feature extraction period, the total number of features of the food security data during the feature extraction period, and the Pearson correlation coefficient of each food security data during the feature extraction period.

[0066] It should be elaborated that the above-mentioned feature extraction period specifically refers to the period from when the food emergency optimization device starts to extract features from the food security data to the end of the extraction. The specific duration is jointly determined by factors such as the difficulty of the feature extraction algorithm and the preprocessing effect of the food security data, and can be obtained through the time module of the timing function in Python.

[0067] It should be elaborated that the total duration of the above-mentioned feature extraction period specifically refers to the total length of the time period of the feature extraction period, which can be obtained through the time module of the timing function in Python; the actual feature extraction duration of the above-mentioned food security data during the feature extraction period specifically refers to the actual duration experienced from the input of the food security data into the preset deep learning algorithm to the completion of feature extraction, which can be obtained through the time module of the timing function in Python; the number of correctly extracted features of the above-mentioned food security data during the feature extraction period specifically refers to the number of effective features successfully extracted from the food security data through the preset deep learning algorithm, which can be directly obtained from the output of the preset deep learning algorithm; the total number of features of the above-mentioned food security data during the feature extraction period specifically refers to the sum of all features extracted from the food security data during the feature extraction process. These features cover all aspects of food production, inventory, and transportation, such as yield, planting area, unit yield, inventory, consumption, price fluctuations, policy impacts, etc., which can be directly obtained from the output of the preset deep learning algorithm; the Pearson correlation coefficient of each of the above-mentioned food security data during the feature extraction period specifically refers to a statistic used to measure the strength and direction of the linear relationship between the features of two food security data. This coefficient reflects the degree of linear correlation between the two feature variables, and its value range is between -1 and 1, which can be obtained through pearsonr in the scipy.stats library of Python.

[0068] The ratio of the actual feature extraction duration of the food security data during the feature extraction period to the total duration of the feature extraction period is processed to obtain the proportion of the feature extraction duration of the food security data during the feature extraction period.

[0069] The ratio of the number of correctly extracted features of the food security data during the feature extraction period to the total number of features of the food security data during the feature extraction period is processed to obtain the feature extraction accuracy of the food security data during the feature extraction period.

[0070] The Pearson correlation coefficient of each of the food security data during the feature extraction period is compared with the preset Pearson correlation coefficient threshold in the food optimization library, and the number of several food security data with a Pearson correlation coefficient greater than the Pearson correlation coefficient threshold is counted, which is recorded as the redundant data volume of the food security data during the feature extraction period. The ratio of the redundant data volume of the food security data during the feature extraction period to the total data volume of the food security data is processed to obtain the feature redundancy of the food security data during the feature extraction period.

[0071] It should be elaborated that the proportion of the feature extraction duration of the above food security data in the feature extraction period specifically refers to the proportion of the actual feature extraction duration to the total duration of the feature extraction period; the feature extraction accuracy of the above food security data in the feature extraction period specifically refers to the proportion of the number of correctly extracted features to the total number of features, which is an important indicator to measure the accuracy and effectiveness of the feature extraction process; the feature redundancy of the above food security data in the feature extraction period specifically refers to the proportion of the number of food security data with a Pearson correlation coefficient greater than the preset threshold to the total data volume, which is an important indicator to measure the amount of redundant features in the feature extraction result; the above Pearson correlation coefficient threshold specifically refers to a preset standard used to judge the strength of the linear correlation between two feature variables during feature extraction and correlation analysis. In this embodiment, the Pearson correlation coefficient threshold is specifically 0.5.

[0072] Through the proportion of the feature extraction duration of the food security data in the feature extraction period, the feature extraction accuracy of the food security data in the feature extraction period, the feature redundancy of the food security data in the feature extraction period, and the food security data processing evaluation index, the food security data eigenvalue is comprehensively analyzed. The food security data eigenvalue is used to comprehensively and quantitatively evaluate the feature extraction effect of the food security data. The food security data eigenvalue represents the degree of influence of the proportion of the feature extraction duration, the feature extraction accuracy, the feature redundancy, and the food security data processing evaluation index on the feature extraction effect of the food security data. The specific acquisition method is as follows: In the formula, represents the proportion of the feature extraction duration of the food security data in the feature extraction period, represents the reference proportion of the feature extraction duration preset in the food optimization library, represents the feature extraction accuracy of the food security data in the feature extraction period, represents the feature redundancy of the food security data in the feature extraction period, represents the food security data processing evaluation index, Indicates the eigenvalue of food security data. When the proportion of the feature extraction duration is too small, it means that the analysis and processing time invested in food security data is insufficient. Within a limited time, some important data features will be ignored, resulting in incomplete extracted features and difficulty in effectively distinguishing noise and real features in the data, thereby increasing the error of feature extraction and leading to a decrease in the accuracy of feature extraction. In a time-pressured situation, in order to extract as many features as possible, some unnecessary or duplicate information will be generated, increasing the feature redundancy and thus reducing the eigenvalue of food security data; when the food security data processing evaluation index is low, it indicates that the data quality is poor and the feature extraction is difficult. When the proportion of the feature extraction duration increases, the excessive extraction time will result in the extraction of too many features, some of which are irrelevant or redundant, which will instead reduce the accuracy of feature extraction and increase the feature redundancy, ultimately leading to a decrease in the eigenvalue of food security data.

[0073] It should be elaborated that the specific reference proportion of the above-mentioned feature extraction duration refers to the optimal reference value of the feature extraction duration proportion preset in the food optimization library.

[0074] Indicates the data feature correction weight of the feature extraction duration proportion preset in the food optimization library. Indicates the data feature influence weight of the feature extraction accuracy preset in the food optimization library. Indicates the data feature influence weight of the feature redundancy preset in the food optimization library. Indicates the data feature influence weight of the food security data processing evaluation index preset in the food optimization library.

[0075] It should be noted that the data feature impact weights corresponding to the feature extraction duration ratio, feature extraction accuracy, feature redundancy, and food security data processing evaluation index are used to indicate the importance of the feature extraction duration ratio, feature extraction accuracy, feature redundancy, and food security data processing evaluation index in analyzing the food security data feature values. For example, there is a pre-set mapping relationship between the real-time feature extraction related data and the corresponding data feature impact weights in the food optimization library. Through the pre-set mapping relationship, the data feature impact weights corresponding to the real-time feature extraction related data can be matched. The feature extraction duration ratio, feature extraction accuracy, and feature redundancy are respectively matched with the pre-set mapping relationship to obtain the data feature correction weights corresponding to the feature extraction duration ratio, the data feature impact weights corresponding to the feature extraction accuracy, and the data feature impact weights corresponding to the feature redundancy. For example, there is a pre-set mapping relationship between the real-time food security data processing evaluation index and the corresponding data feature impact weights in the food optimization library. Through the pre-set mapping relationship, the data feature impact weights corresponding to the real-time food security data processing evaluation index can be matched. The food security data processing evaluation index is matched with the pre-set mapping relationship to obtain the data feature impact weights corresponding to the food security data processing evaluation index. In this embodiment, the value ranges are all (0, 1).

[0076] Furthermore, the specific judgment process for determining whether to re-extract features from the food security data after data preprocessing is as follows: If the food security data feature value is greater than or equal to the food security data feature threshold preset in the food optimization library, there is no need to re-extract features from the food security data after data preprocessing. The food security data feature threshold specifically refers to the critical value used to judge whether the feature extraction effect of the food security data reaches the qualified standard during the evaluation of the feature extraction effect of the food security data, and is used to measure whether the feature extraction effect of the food security data reaches the qualified standard. The above-mentioned food security data feature value being greater than or equal to the food security data feature threshold preset in the food optimization library indicates that the feature extraction effect of the food security data is good, so there is no need to re-extract features from the food security data after data preprocessing.

[0077] If the food security data feature value is less than the food security data feature threshold preset in the food optimization library, it indicates that the feature extraction effect of the food security data is poor, and it is necessary to re-extract features from the food security data after data preprocessing.

[0078] The feature extraction of the food security data after re-preprocessing the data is as follows. Specifically, the difference between the feature values of the food security data and the feature threshold values of the food security data is calculated to obtain the feature deviation values of the food security data. The feature deviation values of the food security data are matched with the optimized set of feature extraction algorithms corresponding to each interval of the feature deviation values of the food security data preset in the food optimization library, so as to obtain the optimized set of feature extraction algorithms corresponding to the feature deviation values of the food security data. Then, the feature extraction algorithm is optimized according to the optimized set of feature extraction algorithms, and the second feature extraction algorithm is obtained. Finally, the food security data after re-preprocessing the data is re-extracted using the second feature extraction algorithm.

[0079] It should be noted that the matching of the feature deviation values of the food security data with the optimized set of feature extraction algorithms corresponding to each interval of the feature deviation values of the food security data preset in the food optimization library is as follows: The optimized set of feature extraction algorithms corresponding to each interval of the feature deviation values of the food security data is formulated by authoritative feature extraction monitoring agencies or standardization organizations after data collection, collation, and encryption processing. Each feature deviation value of the food security data is carefully divided, and for each interval range of the feature deviation values of the food security data, an optimized set of feature extraction algorithms is designed based on feature extraction standards, threat and risk analysis, and application requirements and compliance requirements. For example, if the feature deviation value of the food security data in this example is 2, it belongs to the interval [1, 3] corresponding to the feature deviation values of the food security data, and the specific content of the optimized set of feature extraction algorithms corresponding to the interval [1, 3] of the feature deviation values of the food security data preset in the food optimization library is "add L1 regularization in the RNN to improve the generalization ability; at the same time, introduce residual connections to reduce the problem of gradient disappearance".

[0080] It should be noted that the above-mentioned second feature extraction algorithm specifically refers to the algorithm obtained after optimizing the feature extraction algorithm according to the optimized set of feature extraction algorithms, denoted as the second feature extraction algorithm, which specifically includes L1 regularization, residual connections, and the recurrent neural network (RNN).

[0081] S3. The food security data after feature extraction is recorded as food security feature parameters by the food emergency optimization device, and the food security feature parameters are input into a preset prediction model for training to generate a food security system prediction model. The simulation prediction is carried out through the food security system prediction model to obtain simulation test parameters, and then the simulation test evaluation index is analyzed and compared with the simulation test evaluation threshold preset in the food optimization library to determine whether to optimize the food security system prediction model, and the food security system prediction model is stored in the food emergency security system to complete the optimization of the prediction model in the food emergency security system.

[0082] In a specific embodiment, the above-mentioned preset prediction model specifically refers to the time series model LSTM (Long Short-Term Memory Network) based on deep learning; the specific process of generating the food security system prediction model is to randomly shuffle the food security characteristic parameters to ensure the randomness of the samples, and divide them into a training set, a validation set, and a test set according to a set ratio (such as 70% training set, 15% validation set, 15% test set) for model training, validation, and testing; use the training set as the input variables of the LSTM model, and use the recursive feature elimination algorithm to select the most appropriate feature combination; construct the LSTM network structure, including an input layer, an LSTM layer, and an output layer, set two LSTM layers, each layer contains 128 units, inside the LSTM layer, the forget gate, the input gate, and the output gate use the sigmoid function as the activation function, the candidate memory unit uses the tanh function as the activation function, the optimizer is the Adam optimizer, and the learning rate is set to 0.001; input the training set data into the LSTM model for training. During the training process, update the weight and bias parameters of the model through the backpropagation algorithm, use the validation set data to validate the model, evaluate the prediction performance of the model, and adjust the parameters and structure of the model according to the validation results to finally generate the food security system prediction model; the specific process of the above-mentioned simulation prediction is to transfer the test set data to the food security system prediction model for forward propagation calculation. In this process, the model will gradually calculate the state of the hidden layer and the prediction results of the output layer according to the characteristics of the test set data and the previously learned rules, and obtain the simulation test parameters from the output layer of the food security system prediction model.

[0083] Specifically, the process of specifically analyzing the simulation test evaluation index is as follows: The above-mentioned simulation test parameters specifically include the total data volume of the food security characteristic parameters, the accurate prediction volume of the food security characteristic parameters by the food security system prediction model during the simulation test period, the simulation test response duration of the food security system prediction model during the simulation test period, and the deviation degree of the food demand prediction by the food security system prediction model during the simulation test period.

[0084] It should be elaborated that the above-mentioned feature extraction period specifically refers to the period from when the food emergency optimization device starts to conduct simulation tests on the food security characteristic parameters to the end of the test. The specific duration is jointly determined by factors such as the difficulty of the simulation test algorithm and the data volume of the food security characteristic parameter test set, and can be obtained through the time module of the timing function in Python.

[0085] It should be noted that the total data volume of the above food security characteristic parameters specifically refers to the total number of data of the food security characteristic parameters, which can be directly obtained from the output of the RNN model; the accurate prediction volume of the food security characteristic parameters of the above food security system prediction model during the simulation test period specifically refers to the data volume within the qualified range of the error between the predicted value and the actual value given by the model for the food security characteristic parameters during the simulation test process of the food security system prediction model, which can be directly obtained from the output of the food security system prediction model; the simulation test response duration of the above food security system prediction model during the simulation test period specifically refers to the duration from when the food security system prediction model receives the food security characteristic parameters to when the model starts to truly calculate and conduct the simulation test, which can be obtained through the time module of the timing function in Python; the deviation degree of the food demand prediction of the above food security system prediction model during the simulation test period specifically refers to the degree of difference between the predicted food demand and the actual food demand of the food security system prediction model during the simulation test period, which can be obtained by calculating the difference between the prediction result of the food security system prediction model and the actual food demand.

[0086] The ratio of the accurate prediction volume of the food security characteristic parameters of the food security system prediction model during the simulation test period to the total data volume of the food security characteristic parameters is processed to obtain the prediction coverage rate of the food security system prediction model during the simulation test period.

[0087] It should be noted that the prediction coverage rate of the above food security system prediction model during the simulation test period specifically refers to the ratio between the accurate prediction volume made for the food security characteristic parameters and the total data volume of this characteristic parameter by the food security system prediction model during the simulation test process. This ratio reflects the prediction ability of the model for the food security characteristic parameters during the simulation test stage.

[0088] Through the prediction coverage rate of the food security system prediction model during the simulation test period, the simulation test response duration of the food security system prediction model during the simulation test period, the deviation degree of the food demand prediction of the food security system prediction model during the simulation test period, the food security data processing evaluation index, and the food security data characteristic value, the simulation test evaluation index is comprehensively analyzed. The simulation test evaluation index is used to comprehensively and quantitatively evaluate the simulation test effect of the food security system prediction model. The simulation test evaluation index represents the quantitative data of the influence degree of the prediction coverage rate, the simulation test response duration, the deviation degree of the food demand prediction, the food security data processing evaluation index, and the food security data characteristic value on the simulation test effect of the food security system prediction model. The specific acquisition method is as follows: In the formula, Indicates the prediction coverage rate of the food security system prediction model during the simulation test period. Indicates the simulation test response duration of the food security system prediction model during the simulation test period. Indicates the deviation degree of food demand prediction of the food security system prediction model during the simulation test period. Indicates the food security data processing evaluation index. Indicates the food security data eigenvalue. Indicates the simulation test evaluation index. When the food security data processing evaluation index is low, it means that the data preprocessing effect is not good, the difficulty of data feature extraction increases, which will lead to a decrease in the food security data eigenvalue, indicating that the quality of the food security feature parameters is poor. This will increase the simulation test response duration and also cause deviations in the model during the prediction process, increasing the deviation degree of food demand prediction, meaning that some important features are missed in the prediction process of the model, resulting in a decrease in the prediction coverage rate, and ultimately reducing the simulation test evaluation index.

[0089] Indicates the simulation test weight factor of the preset prediction coverage rate of the food optimization library. Indicates the simulation test weight factor of the preset simulation test response duration of the food optimization library. Indicates the simulation test weight factor of the preset deviation degree of food demand prediction of the food optimization library. Indicates the simulation test weight factor of the preset food security data processing evaluation index of the food optimization library. Indicates the simulation test weight factor of the preset food security data eigenvalue of the food optimization library. p is the first simulation test during the simulation test period.

[0090] It should be noted that the simulation test weight factors corresponding to the prediction coverage rate, the simulation test response duration, the deviation degree of food demand prediction, the evaluation index of food security data processing, and the eigenvalue of food security data are respectively used to indicate the importance of the prediction coverage rate, the simulation test response duration, the deviation degree of food demand prediction, the evaluation index of food security data processing, and the eigenvalue of food security data in the analysis to obtain the simulation test evaluation index. For example, there is a pre-set mapping relationship between the real-time simulation test parameters and the corresponding simulation test weight factors in the food optimization library. Through the pre-set mapping relationship, the simulation test weight factor corresponding to the real-time simulation test parameters can be matched. The prediction coverage rate, the simulation test response duration, and the deviation degree of food demand prediction are respectively matched with the pre-set mapping relationship to obtain the simulation test weight factor corresponding to the prediction coverage rate, the simulation test weight factor corresponding to the simulation test response duration, and the simulation test weight factor corresponding to the deviation degree of food demand prediction. For example, there is a pre-set mapping relationship between the real-time evaluation index of food security data processing and the corresponding simulation test weight factor in the food optimization library. Through the pre-set mapping relationship, the simulation test weight factor corresponding to the real-time evaluation index of food security data processing can be matched. The evaluation index of food security data processing is matched with the pre-set mapping relationship to obtain the simulation test weight factor corresponding to the evaluation index of food security data processing. For example, there is a pre-set mapping relationship between the real-time eigenvalue of food security data and the corresponding simulation test weight factor in the food optimization library. Through the pre-set mapping relationship, the simulation test weight factor corresponding to the real-time eigenvalue of food security data can be matched. The eigenvalue of food security data is matched with the pre-set mapping relationship to obtain the simulation test weight factor corresponding to the eigenvalue of food security data. In this embodiment, the value range is (0, 1).

[0091] Furthermore, the specific process of determining whether to optimize the food security system prediction model is as follows: If the simulation test evaluation index is greater than or equal to the simulation test evaluation threshold preset in the food optimization library, there is no need to optimize the food security system prediction model. The simulation test evaluation threshold specifically refers to the critical value used to determine whether the simulation test evaluation effect of the food security system prediction model reaches the prediction standard during the evaluation of the simulation test evaluation effect of the food security system prediction model, and is used to measure whether the simulation test evaluation effect of the food security system prediction model reaches the prediction standard. The above simulation test evaluation index being greater than or equal to the simulation test evaluation threshold preset in the food optimization library indicates that the simulation test evaluation effect of the food security system prediction model is good, and there is no need to optimize the food security system prediction model.

[0092] If the simulation test evaluation index is less than the simulation test evaluation threshold preset in the food optimization library, it indicates that the simulation test effect of the food security system prediction model is not good, and then the food security system prediction model needs to be optimized.

[0093] The optimization of the food security system prediction model is specifically carried out by matching the simulation test evaluation index with the prediction model optimization set corresponding to each simulation test evaluation index interval preset in the food optimization library, so as to obtain the prediction model optimization set corresponding to the simulation test evaluation index, and optimize the food security system prediction model according to the prediction model optimization set.

[0094] It should be elaborated that the above matching of the simulation test evaluation index with the prediction model optimization set corresponding to each simulation test evaluation index interval preset in the food optimization library is specifically carried out as follows: For each of the above simulation test evaluation index intervals, the corresponding prediction model optimization set is formulated by an authoritative model optimization monitoring agency or standardization organization through data collection, collation, and encryption processing. Each simulation test evaluation index is carefully divided, and for each simulation test evaluation index interval range, based on model optimization criteria, threat and risk analysis, as well as application requirements and compliance requirements, the corresponding prediction model optimization set is designed; for example, if the simulation test evaluation index in this example is 6, it belongs to the [3, 6] corresponding to the simulation test evaluation index interval, and the specific content of the prediction model optimization set corresponding to the [3, 6] simulation test evaluation index interval preset in the food optimization library is "increase the number of LSTM layers of the food security system prediction model to 5 layers, and add the sum of squares of weights as a penalty term in the loss function".

[0095] Refer to Figure 2 As shown, the second aspect of the present invention provides a food emergency security system optimization system based on deep learning, including: a food security data preprocessing module, a food security data feature extraction module, a food emergency security system optimization module, and a food optimization library.

[0096] The food security data preprocessing module is connected to the food security data feature extraction module, the food security data feature extraction module is connected to the food emergency security system optimization module, and the food security data preprocessing module, the food security data feature extraction module, and the food emergency security system optimization module are all connected to the food optimization library.

[0097] The food security data preprocessing module is used to obtain food security data through the food emergency optimization device, preprocess the food security data according to a preset method to obtain data processing information during the data preprocessing, analyze and obtain a food security data processing evaluation index, and compare it with the food security data processing evaluation threshold preset in the food optimization library to determine whether to preprocess the food security data again.

[0098] The food security data feature extraction module is used to extract features from the preprocessed food security data through the deep learning algorithm preset in the food emergency optimization device, obtain feature extraction-related data, thereby analyzing and obtaining food security data feature values, and comparing them with the food security data feature thresholds preset in the food optimization library to determine whether to re-extract features from the preprocessed food security data.

[0099] The food emergency security system optimization module is used to record the food security data after feature extraction as food security feature parameters through the food emergency optimization device, input the food security feature parameters into a preset prediction model for training to generate a food security system prediction model, perform simulation prediction through the food security system prediction model to obtain simulation test parameters, thereby analyzing and obtaining simulation test evaluation indicators, and comparing them with the simulation test evaluation thresholds preset in the food optimization library to determine whether to optimize the food security system prediction model, and store the food security system prediction model in the food emergency security system to complete the optimization of the prediction model in the food emergency security system.

[0100] The food optimization library is used to store food security data processing evaluation thresholds, food security data feature thresholds, simulation test evaluation thresholds, Pearson correlation coefficient thresholds, reference proportion of feature extraction duration, data processing influence coefficients of data signal smoothness, data processing influence coefficients of data signal volatility, data processing influence coefficients of data error rate, data processing influence coefficients of data consistency verification pass rate, data processing influence coefficients of data filling rate, data feature correction weights of feature extraction duration proportion, data feature influence weights of feature extraction accuracy, data feature influence weights of feature redundancy, data processing influence coefficients of food security data processing evaluation index, simulation test weight factors of prediction coverage rate, simulation test weight factors of simulation test response duration, simulation test weight factors of food demand prediction deviation degree, simulation test weight factors of food security data processing evaluation index, simulation test weight factors of food security data feature values, optimization sets of feature extraction algorithms corresponding to intervals of each food security data feature deviation value, and optimization sets of prediction models corresponding to intervals of each simulation test evaluation index.

[0101] The third aspect of the present invention provides an apparatus for optimizing a food emergency security system based on deep learning, characterized in that the apparatus includes a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the apparatus is triggered to execute the method described above.

[0102] The above content is only an example and explanation of the structure of the present invention. Those skilled in the art of the present technology can make various modifications or supplements or use similar methods to replace the specific embodiments described, as long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should fall within the protection scope of the present invention.

Claims

1. A method for optimizing the food emergency guarantee system based on deep learning, characterized in that: include: S1. Obtain food security data through the food emergency optimization device, perform data preprocessing on the food security data according to a preset method to obtain data processing information in the data preprocessing process, analyze and obtain the food security data processing evaluation index, and compare it with the food security data processing evaluation threshold preset in the food optimization library to determine whether to perform data preprocessing on the food security data again; S2. Extract features from the pre-processed food security data using the deep learning algorithm preset by the food emergency optimization device, obtain feature extraction related data, analyze and obtain feature values ​​of the food security data, and compare them with the feature thresholds of the food security data preset by the food optimization library to determine whether to re-extract features from the pre-processed food security data; S3. The food security data after feature extraction is recorded as food security feature parameters through the food emergency optimization device, and the food security feature parameters are input into the preset prediction model for training to generate a food security system prediction model. The food security system prediction model is used for simulation prediction to obtain simulation test parameters, and the simulation test evaluation indicators are analyzed and compared with the simulation test evaluation threshold preset in the food optimization library to determine whether the food security system prediction model is optimized, and the food security system prediction model is stored in the food emergency security system to complete the optimization of the prediction model in the food emergency security system.

2. The method for optimizing the food emergency guarantee system based on deep learning according to claim 1 is characterized in that: The data processing information specifically includes the total amount of food security data, the error value of each food security data during the data preprocessing period, the amount of data that has passed the consistency check of the food security data during the data preprocessing period, the total number of missing values ​​of the food security data during the data preprocessing period, and the number of missing values ​​filled in the food security data during the data preprocessing period; Obtaining a data signal amplitude change curve of the food security data during the data preprocessing period, extracting the signal amplitude corresponding to each curve change position point from the data signal amplitude change curve of the food security data during the data preprocessing period, performing difference processing on the signal amplitudes corresponding to adjacent curve change position points, and performing square sum processing on the difference processing results to obtain the data signal smoothness of the food security data during the data preprocessing period; The signal amplitudes corresponding to the highest position point and the lowest position point are extracted from the data signal amplitude change curve of the food security data during the data preprocessing period, and difference processing and absolute value processing are performed in sequence to obtain the data signal volatility of the food security data during the data preprocessing period; The error values ​​of each food security data in the data preprocessing period are squared, and the squared results are averaged to obtain the data error rate of the food security data in the data preprocessing period; The data consistency check pass rate of the food security data in the data preprocessing period is obtained by performing ratio processing on the amount of data that has passed the consistency check of the food security data in the data preprocessing period and the total amount of data of the food security data; The data filling rate of the food security data during the data preprocessing period is obtained by ratio processing the number of missing values ​​filled in the food security data during the data preprocessing period with the total number of missing values ​​of the food security data during the data preprocessing period.

3. The method for optimizing the food emergency guarantee system based on deep learning according to claim 2 is characterized in that: The specific analysis process of the food security data processing assessment index is as follows: Through comprehensive analysis of the data signal smoothness of the food security data in the data preprocessing period, the data signal volatility of the food security data in the data preprocessing period, the data error rate of the food security data in the data preprocessing period, the data consistency check pass rate of the food security data in the data preprocessing period and the data filling rate of the food security data in the data preprocessing period, the food security data processing evaluation index is obtained. The food security data processing evaluation index is used for comprehensive and quantitative evaluation of the data preprocessing effect of the food security data. The food security data processing evaluation index represents quantitative data on the degree of influence of the data signal smoothness, data signal volatility, data error rate, data consistency check pass rate and data filling rate on the data preprocessing effect of the food security data.

4. The method for optimizing the food emergency guarantee system based on deep learning according to claim 1 is characterized in that: The specific judgment process of whether to perform data preprocessing on the food security data again is as follows: If the food security data processing assessment index is greater than or equal to the food security data processing assessment threshold preset by the food optimization library, there is no need to pre-process the food security data again; If the food security data processing evaluation index is less than the food security data processing evaluation threshold preset by the food optimization library, the food security data needs to be preprocessed again; The aforementioned data preprocessing of the food security data again specifically refers to first standardizing the food security data into a standard normal distribution for data processing through a Z-score formula, and then preprocessing the food security data through a preset method.

5. The method for optimizing the food emergency guarantee system based on deep learning according to claim 1 is characterized in that: The specific analysis process of the food security data characteristic value is as follows: The feature extraction related data specifically include the total duration of the feature extraction period, the actual feature extraction duration of the food security data in the feature extraction period, the number of correctly extracted features of the food security data in the feature extraction period, the total number of features of the food security data in the feature extraction period, and the Pearson correlation coefficient of each food security data in the feature extraction period; The actual feature extraction time of the food security data in the feature extraction period is processed by ratio with the total time of the feature extraction period to obtain the feature extraction time ratio of the food security data in the feature extraction period; The number of correctly extracted features of the food security data in the feature extraction period is processed by ratio with the total number of features of the food security data in the feature extraction period to obtain the feature extraction accuracy of the food security data in the feature extraction period; The Pearson correlation coefficient of each food security data in the feature extraction period is compared with the Pearson correlation coefficient threshold preset by the food optimization library, and the number of food security data with a Pearson correlation coefficient greater than the Pearson correlation coefficient threshold is counted, which is recorded as the redundant data volume of the food security data in the feature extraction period, and the redundant data volume of the food security data in the feature extraction period is processed by ratio processing with the total data volume of the food security data to obtain the feature redundancy of the food security data in the feature extraction period; The characteristic value of food security data is obtained through comprehensive analysis of the proportion of feature extraction time of food security data in the feature extraction period, the feature extraction accuracy of food security data in the feature extraction period, the feature redundancy of food security data in the feature extraction period and the food security data processing evaluation index. The characteristic value of food security data is used for comprehensive quantitative evaluation of feature extraction effect of food security data. The characteristic value of food security data represents quantitative data of the degree of influence of the proportion of feature extraction time, feature extraction accuracy, feature redundancy and food security data processing evaluation index on the feature extraction effect of food security data.

6. The method for optimizing the food emergency guarantee system based on deep learning according to claim 1 is characterized in that: The specific judgment process of whether to re-extract features from the food security data after data preprocessing is as follows: If the characteristic value of the food security data is greater than or equal to the characteristic threshold of the food security data preset by the food optimization library, there is no need to re-extract the characteristics of the food security data after data preprocessing; If the characteristic value of the food security data is less than the characteristic threshold of the food security data preset by the food optimization library, it is necessary to re-extract the characteristics of the food security data after data preprocessing; The re-feature extraction of the food security data after data preprocessing is specifically carried out in a process of performing difference processing on the food security data feature value and the food security data feature threshold to obtain the food security data feature deviation value, matching the food security data feature deviation value with the feature extraction algorithm optimization set corresponding to each food security data feature deviation value interval preset in the food optimization library, thereby obtaining the feature extraction algorithm optimization set corresponding to the food security data feature deviation value, and optimizing the feature extraction algorithm according to the feature extraction algorithm optimization set to obtain the second feature extraction algorithm, and finally re-extracting the feature of the food security data after data preprocessing by the second feature extraction algorithm.

7. The method for optimizing the food emergency guarantee system based on deep learning according to claim 1 is characterized in that: The simulation test evaluation index, the specific analysis process is as follows: The simulation test parameters specifically include the total data volume of food security characteristic parameters, the accurate prediction volume of food security characteristic parameters of the food security system prediction model during the simulation test period, the simulation test response time of the food security system prediction model during the simulation test period, and the food demand prediction deviation of the food security system prediction model during the simulation test period; The accurate prediction amount of food security characteristic parameters of the food security system prediction model in the simulation test period is processed with the total data amount of food security characteristic parameters to obtain the prediction coverage rate of the food security system prediction model in the simulation test period; Through comprehensive analysis of the prediction coverage of the food security system prediction model in the simulation test period, the simulation test response time of the food security system prediction model in the simulation test period, the food demand prediction deviation of the food security system prediction model in the simulation test period, the food security data processing evaluation index and the food security data characteristic value, the simulation test evaluation index is obtained. The simulation test evaluation index is used to comprehensively and quantitatively evaluate the simulation test effect of the food security system prediction model. The simulation test evaluation index represents quantitative data on the degree of influence of the prediction coverage, simulation test response time, food demand prediction deviation, food security data processing evaluation index and food security data characteristic value on the simulation test effect of the food security system prediction model.

8. The method for optimizing the food emergency guarantee system based on deep learning according to claim 1 is characterized in that: The specific process of judging whether to optimize the food security system prediction model is as follows: If the simulation test evaluation index is greater than or equal to the simulation test evaluation threshold preset by the grain optimization library, there is no need to optimize the grain security system prediction model; If the simulation test evaluation index is less than the simulation test evaluation threshold preset by the grain optimization library, the grain security system prediction model needs to be optimized; The optimization of the food security system prediction model is carried out, and the specific optimization process is to match the simulation test evaluation indicators with the prediction model optimization sets corresponding to the simulation test evaluation indicator intervals preset in the food optimization library, thereby obtaining the prediction model optimization sets corresponding to the simulation test evaluation indicators, and optimizing the food security system prediction model according to the prediction model optimization sets.

9. A system using the deep learning-based food emergency security system optimization method according to any one of claims 1 to 8, characterized in that: include: A food security data preprocessing module is used to obtain food security data through a food emergency optimization device, perform data preprocessing on the food security data according to a preset method to obtain data processing information in the data preprocessing process, analyze and obtain a food security data processing evaluation index, and compare it with a food security data processing evaluation threshold preset in a food optimization library to determine whether to perform data preprocessing on the food security data again; The food security data feature extraction module is used to extract features from the food security data after data preprocessing through the deep learning algorithm preset by the food emergency optimization device, obtain feature extraction related data, thereby analyzing and obtaining the feature value of the food security data, and comparing it with the food security data feature threshold preset by the food optimization library to determine whether to re-extract features from the food security data after data preprocessing; The food emergency security system optimization module is used to record the food security data after feature extraction as food security feature parameters through the food emergency optimization device, and input the food security feature parameters into the preset prediction model for training, generate a food security system prediction model, perform simulation prediction through the food security system prediction model, obtain simulation test parameters, analyze and obtain simulation test evaluation indicators, and compare them with the simulation test evaluation threshold preset in the food optimization library to determine whether to optimize the food security system prediction model, and store the food security system prediction model in the food emergency security system to complete the optimization of the prediction model in the food emergency security system.

10. A device for optimizing a food emergency guarantee system based on deep learning, characterized in that: The device comprises a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the method according to any one of claims 1 to 8.

Images