AI-based agricultural product futures transaction prediction method and system

Abstract

The invention provides an agricultural product futures transaction prediction method and system based on AI, and relates to the field, and the method comprises the following steps: collecting the historical data of agricultural product futures transaction, determining the uncertain factors of the agricultural product price through a correlation analysis method, and constructing an agricultural product price uncertainty probability model; performing binning processing on the historical data, and fitting a probability distribution function and a cumulative distribution function corresponding to each binning to form a binning distribution model; multi-source heterogeneous data are collected in real time through edge nodes, and agricultural product supply chain digital twinborn bodies are constructed; and constructing and driving an artificial intelligence prediction model, generating an agricultural product price scene through inverse transformation sampling, and outputting a probability distribution-based transaction prediction result. According to the invention, scientificity, real-time performance and anti-risk capability of agricultural product futures transaction decision making are improved, and an intelligent solution is provided for agricultural supply chain finance and market risk management.

Description

Technical Field

[0001] The present invention relates to the field of futures price forecasting, and more specifically, to an AI-based agricultural product futures trading forecasting method and system. Background Art

[0002] In today's agricultural product markets, the two dominant trading methods—spot trading and guaranteed origin trading—each face unique challenges. These challenges not only impact the sales and quality of agricultural products but also have profound implications for the economic well-being of farmers and consumers. For spot trading, a core issue lies in the highly concentrated ripening and market cycles of agricultural products. This often results in the market being flooded with a large number of similar agricultural products at the same time.

[0003] Currently, traditional agricultural futures forecasting methods rely on single data sources or simple statistical models, making it difficult to fully consider the dynamic impact of multiple factors on agricultural product prices, such as market transactions, weather, and logistics. Furthermore, due to a lack of real-time perception and precise analysis of complex supply chain conditions, it is difficult to accurately assess the combined impact of various uncertainties on agricultural product prices, resulting in significant risks for both parties involved in pricing and trading decisions.

[0004] Currently, no effective solutions have been proposed for the problems in related technologies. Summary of the Invention

[0005] In view of this, the present invention provides an AI-based agricultural product futures trading prediction method and system to solve the above-mentioned problems.

[0006] In order to solve the above problems, the specific technical solutions adopted by the present invention are as follows:

[0007] According to one aspect of the present invention, there is provided an AI-based agricultural product futures trading prediction method, comprising the following steps:

[0008] S1. Collect historical data on agricultural product futures trading and use correlation analysis to determine the uncertainty factors of agricultural product prices. Based on these uncertainty factors, construct an uncertainty probability model for agricultural product prices.

[0009] S2. Bin the historical data according to the value range of the uncertainty factor, and combine it with the uncertainty probability model of agricultural product prices to fit the probability distribution function and cumulative distribution function corresponding to each bin to form a bin distribution model;

[0010] S3: Build a cloud-edge collaborative agricultural product data perception network, collect multi-source heterogeneous data in real time through edge nodes, and build a digital twin of the agricultural product supply chain;

[0011] S4. Based on the digital twin of the agricultural product supply chain and the bin distribution model, build and drive an artificial intelligence prediction model, generate agricultural product price scenarios through inverse transformation sampling, and output transaction prediction results based on probability distribution.

[0012] Preferably, the process of collecting historical data of agricultural product futures transactions and determining the uncertainty factors of agricultural product prices through correlation analysis, and constructing an agricultural product price uncertainty probability model based on the uncertainty factors comprises the following steps:

[0013] S11. Collect and pre-process historical data on agricultural futures trading, including market trading data, weather data, logistics data, and agricultural product price data;

[0014] S12. Use time-lagged cross-analysis and causal forest models to analyze the dynamic correlation between market transaction data, meteorological data, logistics data, and agricultural product price fluctuations, and screen out uncertain factors in agricultural product prices based on the analysis results;

[0015] S13. Based on the screened uncertainty factors, a Gaussian mixture model is used to construct an uncertainty probability model for agricultural product prices.

[0016] Preferably, the time-lag cross analysis and causal forest model are used to analyze the dynamic correlation between market transaction data, meteorological data, logistics data and agricultural product price fluctuations, and the uncertain factors of agricultural product prices are screened out based on the analysis results, including:

[0017] S121. Based on the pre-processed historical data, determine the factors affecting the price fluctuations of agricultural products using market transaction data, meteorological data, and logistics data, and construct candidate factors.

[0018] S122. For each candidate factor, calculate the lagged correlation coefficient between it and the agricultural product price within the preset time lag window;

[0019] S123. Based on the causal forest model, the causal effect strength of each candidate factor is evaluated, and combined with the lagged correlation coefficient, the uncertain factors of agricultural product prices are screened out.

[0020] Preferably, the step of evaluating the causal effect strength of each candidate factor based on the causal forest model and screening out the uncertain factors of agricultural product prices in combination with the lagged correlation coefficient includes the following steps:

[0021] S1231. Use candidate factors to construct a time-lagged feature matrix and calculate the conditional average treatment effect, factor importance, and effect direction stability of each candidate factor based on the causal forest model.

[0022] S1232. Combine the lagged correlation coefficients of candidate factors to construct a comprehensive screening matrix that includes lagged correlation coefficients, causal effect strength, factor importance, and directional stability;

[0023] S1233. Based on the preset thresholds, candidate factors that meet the requirements of time-lag correlation coefficient, factor importance and directional stability are screened to obtain the uncertainty factors of agricultural product prices.

[0024] Preferably, the method of constructing an agricultural product price uncertainty probability model based on the screened uncertainty factors using a mixed Gaussian model comprises the following steps:

[0025] S131. Standardize the uncertain factors after screening and collect agricultural product price fluctuation data as the target variable;

[0026] S132. Based on the standardized uncertainty factors and target variables, train a Gaussian mixture model, determine the optimal number of Gaussian components using the Bayesian Information Criterion, and estimate the parameters of each component;

[0027] S133. Use the parameters of the mixed Gaussian model to construct a probability model of agricultural product price uncertainty.

[0028] Preferably, the binning process of the historical data according to the value range of the uncertainty factor, and combining the uncertainty probability model of agricultural product prices, respectively fitting the probability distribution function and cumulative distribution function corresponding to each bin, and forming the bin distribution model includes the following steps:

[0029] S21. Determine the value range of each uncertainty factor, divide the value range into several intervals, bin the historical data according to the divided intervals, and mark the bin category to which each data belongs;

[0030] S22. For the data within each bin category, use the corresponding agricultural product price fluctuation data in combination with the mixed Gaussian model to fit the price fluctuation probability density function and cumulative distribution function under each bin;

[0031] S23. Store the probability distribution function and cumulative distribution function parameters of all bins to form a bin distribution model.

[0032] Preferably, the construction of a cloud-edge collaborative agricultural product data perception network, collecting multi-source heterogeneous data in real time through edge nodes, and building a digital twin of the agricultural product supply chain includes the following steps:

[0033] S31. Determine edge deployment nodes based on the agricultural product supply chain and use them to build a cloud-edge collaborative agricultural product data perception network.

[0034] S32. Collect multi-source heterogeneous data from various links in the agricultural product supply chain in real time through edge nodes, and transmit the multi-source heterogeneous data to the cloud through a cloud-edge collaborative agricultural product data perception network;

[0035] S33. Build a digital twin of the agricultural product supply chain based on multi-source heterogeneous data.

[0036] Preferably, the method of constructing and driving an artificial intelligence prediction model based on the digital twin of the agricultural product supply chain and the bin distribution model, generating agricultural product price scenarios through inverse transformation sampling, and outputting transaction prediction results based on probability distribution includes the following steps:

[0037] S41. Obtain the status of agricultural product supply chain based on the digital twin of agricultural product supply chain;

[0038] S42. Match the binning categories of the corresponding uncertain factors based on the agricultural product supply chain status output by the digital twin, and extract the corresponding price fluctuation probability distribution model;

[0039] S43. Based on the probability distribution model under binning, the inverse transformation sampling method is used to generate multiple agricultural product price scenario samples;

[0040] S44. Input the generated price scenario into the preset artificial intelligence prediction model for analysis, and output the agricultural product transaction prediction results based on probability distribution.

[0041] Preferably, the method of generating a plurality of agricultural product price scenario samples using an inverse transformation sampling method based on a probability distribution model under binning includes the following steps:

[0042] S431. Based on the probability distribution model under binning, determine the target probability distribution type of agricultural product price fluctuations and obtain its cumulative distribution function;

[0043] S432. Generate multiple uniformly distributed random numbers using a random number generator, and convert each random number into a price scenario sample based on the inverse function of the cumulative distribution function corresponding to the current bin;

[0044] S433. Repeat steps S431-S432 until a preset number of price scenario samples are generated.

[0045] According to another aspect of the present invention, there is provided an AI-based agricultural product futures trading prediction system, the system comprising:

[0046] The uncertainty probability model construction module is used to collect historical data on agricultural product futures transactions, determine the uncertainty factors of agricultural product prices through correlation analysis, and construct an uncertainty probability model for agricultural product prices based on the uncertainty factors;

[0047] The bin distribution model formation module is used to bin the historical data according to the value range of the uncertainty factor, and to fit the probability distribution function and cumulative distribution function corresponding to each bin in combination with the uncertainty probability model of agricultural product prices to form a bin distribution model;

[0048] A digital twin construction module is used to build a cloud-edge collaborative agricultural product data perception network, collect multi-source heterogeneous data in real time through edge nodes, and build a digital twin of the agricultural product supply chain;

[0049] The futures trading prediction module is used to build and drive an artificial intelligence prediction model based on the digital twin of the agricultural product supply chain and the bin distribution model. It generates agricultural product price scenarios through inverse transformation sampling and outputs trading prediction results based on probability distribution.

[0050] The beneficial effects of the present invention are:

[0051] The present invention achieves end-to-end accurate prediction from complex supply chain status to price fluctuation probability distribution, captures the multimodal characteristics of price fluctuations through mixed Gaussian models and bin distribution models, combines inverse transform sampling to generate diversified price scenario samples, and finally outputs trading strategy recommendations with probabilistic quantification support, significantly improving the scientific nature, real-time performance and risk resistance of agricultural futures trading decisions, and providing an intelligent solution for agricultural supply chain finance and market risk management. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0053] Figure 1 is a flow chart of an AI-based agricultural product futures trading prediction method according to an embodiment of the present invention;

[0054] Figure 2 This is a principle block diagram of an AI-based agricultural product futures trading prediction system according to an embodiment of the present invention.

[0055] In the picture:

[0056] 1. Uncertainty probability model construction module; 2. Bin distribution model formation module; 3. Digital twin construction module; 4. Futures trading prediction module. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0058] According to an embodiment of the present invention, an AI-based agricultural product futures trading prediction method and system are provided.

[0059] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, an AI-based agricultural product futures trading prediction method is provided, comprising the following steps:

[0060] S1. Collect historical data on agricultural product futures trading and use correlation analysis to determine the uncertainty factors of agricultural product prices. Based on these uncertainty factors, construct an uncertainty probability model for agricultural product prices.

[0061] As a preferred embodiment, the process of collecting historical data of agricultural product futures transactions and determining the uncertainty factors of agricultural product prices through correlation analysis, and constructing an uncertainty probability model of agricultural product prices based on the uncertainty factors includes the following steps:

[0062] S11. Collect and pre-process historical data on agricultural futures trading, including market trading data, weather data, logistics data, and agricultural product price data;

[0063] It should be noted that preprocessing includes missing value processing, outlier detection, time alignment, etc.

[0064] S12. Use time-lagged cross-analysis and causal forest models to analyze the dynamic correlation between market transaction data, meteorological data, logistics data, and agricultural product price fluctuations, and screen out uncertain factors in agricultural product prices based on the analysis results;

[0065] As a preferred embodiment, the time-lag cross analysis and causal forest model are used to analyze the dynamic correlation between market transaction data, meteorological data, logistics data and agricultural product price fluctuations, and based on the analysis results, the uncertain factors of agricultural product prices are screened out, including:

[0066] S121. Based on the pre-processed historical data, determine the factors affecting the price fluctuations of agricultural products using market transaction data, meteorological data, and logistics data, and construct candidate factors.

[0067] S122. For each candidate factor, calculate the lagged correlation coefficient between it and the agricultural product price within the preset time lag window;

[0068] It should be noted that when determining the factors affecting the price fluctuations of agricultural products from market transaction data, meteorological data and logistics data, it is necessary to extract factors with potential explanatory power (such as transaction volume and position size in market transactions, cumulative rainfall and temperature deviation in meteorology, transportation delay days and cold chain failure rate in logistics, etc.) and build a candidate factor library.

[0069] Then, a window is set according to the growth cycle of agricultural products and logistics timeliness (for example, it takes about 120 days from sowing to harvesting of corn, and the time lag window can be set to 0-120 days), and the lag correlation coefficient is calculated using the Pearson correlation coefficient calculation method.

[0070] S123. Based on the causal forest model, the causal effect strength of each candidate factor is evaluated, and combined with the lagged correlation coefficient, the uncertain factors of agricultural product prices are screened out.

[0071] As a preferred embodiment, the method of evaluating the causal effect strength of each candidate factor based on the causal forest model and screening out the uncertain factors of agricultural product prices in combination with the lagged correlation coefficient includes the following steps:

[0072] S1231. Use candidate factors to construct a time-lagged feature matrix and calculate the conditional average treatment effect, factor importance, and effect direction stability of each candidate factor based on the causal forest model.

[0073] It should be noted that Causal Forest, as an extension of Random Forest, implements traditional Random Forest modeling for prediction tasks by recursively splitting data, and can quantify the heterogeneous causal effects (HTE) of each sample. Specifically, after constructing a time-lag feature matrix (containing the multi-dimensional features of candidate factors at different time lags), Causal Forest implements causal inference through the following mechanisms:

[0074] In the splitting process of each tree node, the model selects the splitting variable based on the criterion of maximizing the difference of causal effects (rather than the traditional prediction error), thereby dividing the data into sub-sample groups with similar causal effects;

[0075] By aggregating the estimation results of all tree structures, the model generates a conditional average treatment effect (CATE) for each sample, that is, the specific causal impact value of the candidate factor on the agricultural product price in the sample (for example, a 1-unit increase in rainfall at a certain time lag leads to a price fluctuation of 0.02 yuan);

[0076] The importance score is calculated based on the frequency of use of the factor in the split node, reflecting its explanatory power on price fluctuations; at the same time, the stability of the effect direction is evaluated by statistically analyzing the consistency of the causal effect signs in different samples (for example, 85% of the samples show a positive effect, indicating that the influence direction of the factor is reliable).

[0077] The above causal inference process enables the causal forest to simultaneously capture the dynamic temporal impact of candidate factors (through time lag characteristics), causal effect strength (CATE), explanatory power weight (factor importance) and action direction robustness (directional stability), providing a multi-dimensional quantitative basis for screening uncertain factors in agricultural product prices.

[0078] S1232. Combine the lagged correlation coefficients of candidate factors to construct a comprehensive screening matrix that includes lagged correlation coefficients, causal effect strength, factor importance, and directional stability;

[0079] S1233. Based on the preset thresholds, candidate factors that meet the requirements of time-lag correlation coefficient, factor importance and directional stability are screened to obtain the uncertainty factors of agricultural product prices.

[0080] S13. Based on the screened uncertainty factors, a Gaussian mixture model is used to construct an uncertainty probability model for agricultural product prices.

[0081] As a preferred embodiment, the method of constructing an agricultural product price uncertainty probability model based on the screened uncertainty factors using a mixed Gaussian model includes the following steps:

[0082] S131. Standardize the uncertain factors after screening and collect agricultural product price fluctuation data as the target variable;

[0083] It should be noted that the standardization process is to eliminate the dimensional differences of different factors (such as rainfall in millimeters and temperature deviation in degrees Celsius) so that the mixed Gaussian model can fairly compare the contributions of each factor.

[0084] S132. Based on the standardized uncertainty factors and target variables, train a Gaussian mixture model, determine the optimal number of Gaussian components using the Bayesian Information Criterion, and estimate the parameters of each component;

[0085] S133. Use the parameters of the mixed Gaussian model to construct a probability model of agricultural product price uncertainty.

[0086] S2. Bin the historical data according to the value range of the uncertainty factor, and combine it with the uncertainty probability model of agricultural product prices to fit the probability distribution function and cumulative distribution function corresponding to each bin to form a bin distribution model;

[0087] As a preferred embodiment, the historical data is binned according to the value range of the uncertainty factor, and combined with the agricultural product price uncertainty probability model, the probability distribution function and cumulative distribution function corresponding to each bin are fitted respectively to form a bin distribution model, which includes the following steps:

[0088] S21. Determine the value range of each uncertainty factor, divide the value range into several intervals, bin the historical data according to the divided intervals, and mark the bin category to which each data belongs;

[0089] It should be noted that the binning process is to traverse historical samples and label each sample with a unique binning label according to the interval combination into which the values ​​of each factor fall. Each binning category represents a specific combination of uncertain factors.

[0090] S22. For the data within each bin category, use the corresponding agricultural product price fluctuation data in combination with the mixed Gaussian model to fit the price fluctuation probability density function and cumulative distribution function under each bin;

[0091] S23. Store the probability distribution function and cumulative distribution function parameters of all bins to form a bin distribution model.

[0092] S3: Build a cloud-edge collaborative agricultural product data perception network, collect multi-source heterogeneous data in real time through edge nodes, and build a digital twin of the agricultural product supply chain;

[0093] As a preferred embodiment, the construction of a cloud-edge collaborative agricultural product data perception network, real-time collection of multi-source heterogeneous data through edge nodes, and construction of a digital twin of the agricultural product supply chain includes the following steps:

[0094] S31. Determine edge deployment nodes based on the agricultural product supply chain and use them to build a cloud-edge collaborative agricultural product data perception network.

[0095] Specifically, according to the physical process of the agricultural product supply chain (production → processing → warehousing → transportation → sales), edge computing equipment (such as field sensors, cold chain vehicle-mounted terminals, warehouse smart cameras, etc.) are deployed at key nodes.

[0096] S32. Collect multi-source heterogeneous data from various links in the agricultural product supply chain in real time through edge nodes, and transmit the multi-source heterogeneous data to the cloud through a cloud-edge collaborative agricultural product data perception network;

[0097] S33. Build a digital twin of the agricultural product supply chain based on multi-source heterogeneous data.

[0098] It should be noted that the construction of digital twins includes: establishing digital counterparts for physical entities (such as fields, warehouses, and cold chain vehicles); each twin has state variables (such as temperature, inventory, delivery cycle, etc.) and event-driven mechanisms (such as chain breaks, lags, and changes in supply and demand); and then through data fusion algorithms (such as spatiotemporal synchronization, knowledge graphs, and state fusion networks), various heterogeneous data are unified as the input of the twin, realizing functions such as real-time status refresh, predictive maintenance, and simulation testing.

[0099] S4. Based on the digital twin of the agricultural product supply chain and the bin distribution model, build and drive an artificial intelligence prediction model, generate agricultural product price scenarios through inverse transformation sampling, and output transaction prediction results based on probability distribution.

[0100] As a preferred embodiment, the method of constructing and driving an artificial intelligence prediction model based on the digital twin of the agricultural product supply chain and the bin distribution model, generating agricultural product price scenarios through inverse transformation sampling, and outputting transaction prediction results based on probability distribution includes the following steps:

[0101] S41. Obtain the status of agricultural product supply chain based on the digital twin of agricultural product supply chain;

[0102] It should be noted that by driving the digital twin of the agricultural product supply chain with real-time data (such as storage temperature, transportation time, market demand, etc.), a full-dimensional snapshot of the current supply chain can be generated, including the status of the production end, the logistics end, and the market end.

[0103] S42. Match the binning categories of the corresponding uncertain factors based on the agricultural product supply chain status output by the digital twin, and extract the corresponding price fluctuation probability distribution model;

[0104] S43. Based on the probability distribution model under binning, the inverse transformation sampling method is used to generate multiple agricultural product price scenario samples;

[0105] As a preferred embodiment, the method of generating multiple agricultural product price scenario samples using an inverse transformation sampling method based on a probability distribution model under binning includes the following steps:

[0106] S431. Based on the probability distribution model under binning, determine the target probability distribution type of agricultural product price fluctuations and obtain its cumulative distribution function;

[0107] S432. Generate multiple uniformly distributed random numbers using a random number generator, and convert each random number into a price scenario sample based on the inverse function of the cumulative distribution function corresponding to the current bin;

[0108] S433. Repeat steps S431-S432 until a preset number of price scenario samples are generated.

[0109] S44. Input the generated price scenario into the preset artificial intelligence prediction model for analysis, and output the agricultural product transaction prediction results based on probability distribution.

[0110] It should be noted that the artificial intelligence prediction model uses any one of the regression network, Bayesian prediction model, reinforcement learning strategy network, etc., and then inputs the price scenario samples into the artificial intelligence prediction model one by one, outputting the profit distribution, optimal trading strategy recommendations or risk indicators; the final output is the prediction result based on probability distribution.

[0111] like Figure 2 According to another embodiment of the present invention, there is provided an AI-based agricultural product futures trading prediction system, the system comprising:

[0112] Uncertainty probability model construction module 1 is used to collect historical data of agricultural product futures transactions, and determine the uncertainty factors of agricultural product prices through correlation analysis, and construct an uncertainty probability model of agricultural product prices based on the uncertainty factors;

[0113] The bin distribution model forming module 2 is used to bin the historical data according to the value range of the uncertainty factor, and to fit the probability distribution function and cumulative distribution function corresponding to each bin in combination with the uncertainty probability model of agricultural product prices to form a bin distribution model;

[0114] Digital twin construction module 3 is used to build a cloud-edge collaborative agricultural product data perception network, collect multi-source heterogeneous data in real time through edge nodes, and build a digital twin of the agricultural product supply chain;

[0115] Futures trading prediction module 4 is used to build and drive an artificial intelligence prediction model based on the digital twin of the agricultural product supply chain and the bin distribution model, generate agricultural product price scenarios through inverse transformation sampling, and output trading prediction results based on probability distribution.

[0116] To sum up, with the help of the above-mentioned technical scheme of the present invention, the present invention realizes end-to-end accurate prediction from complex supply chain status to price fluctuation probability distribution, captures the multimodal characteristics of price fluctuations through mixed Gaussian models and bin distribution models, combines inverse transformation sampling to generate diversified price scenario samples, and finally outputs trading strategy recommendations with probabilistic quantification support, which significantly improves the scientificity, real-timeness and risk resistance of agricultural futures trading decisions, and provides an intelligent solution for agricultural supply chain finance and market risk management.

[0117] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, optical storage, etc.) containing computer-usable program code.

[0118] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An AI-based agricultural product futures trading prediction method, characterized in that: The following steps are involved: S1. Collect historical data on agricultural product futures trading and use correlation analysis to determine the uncertainty factors of agricultural product prices. Based on these uncertainty factors, construct an uncertainty probability model for agricultural product prices. S2. Bin the historical data according to the value range of the uncertainty factor, and combine it with the uncertainty probability model of agricultural product prices to fit the probability distribution function and cumulative distribution function corresponding to each bin to form a bin distribution model; S3: Build a cloud-edge collaborative agricultural product data perception network, collect multi-source heterogeneous data in real time through edge nodes, and build a digital twin of the agricultural product supply chain; S4. Based on the digital twin of the agricultural product supply chain and the bin distribution model, build and drive an artificial intelligence prediction model, generate agricultural product price scenarios through inverse transformation sampling, and output transaction prediction results based on probability distribution.

2. The AI-based agricultural product futures trading prediction method according to claim 1, characterized in that: The process of collecting historical data of agricultural product futures transactions and determining the uncertainty factors of agricultural product prices through correlation analysis, and constructing an agricultural product price uncertainty probability model based on the uncertainty factors, comprises the following steps: S11. Collect and pre-process historical data on agricultural futures trading, including market trading data, weather data, logistics data, and agricultural product price data; S12. Use time-lagged cross-analysis and causal forest models to analyze the dynamic correlation between market transaction data, meteorological data, logistics data, and agricultural product price fluctuations, and screen out uncertain factors in agricultural product prices based on the analysis results; S13. Based on the screened uncertainty factors, a Gaussian mixture model is used to construct an uncertainty probability model for agricultural product prices.

3. The AI-based agricultural product futures trading prediction method according to claim 2, characterized in that: The time-lag cross analysis and causal forest model are used to analyze the dynamic correlation between market transaction data, meteorological data, logistics data and agricultural product price fluctuations. Based on the analysis results, the uncertain factors of agricultural product prices are screened out, including: S121. Based on the pre-processed historical data, determine the factors affecting the price fluctuations of agricultural products using market transaction data, meteorological data, and logistics data, and construct candidate factors. S122. For each candidate factor, calculate the lagged correlation coefficient between it and the agricultural product price within the preset time lag window; S123. Based on the causal forest model, the causal effect strength of each candidate factor is evaluated, and combined with the lagged correlation coefficient, the uncertain factors of agricultural product prices are screened out.

4. The AI-based agricultural product futures trading prediction method according to claim 3, characterized in that: The method of evaluating the causal effect strength of each candidate factor based on the causal forest model and screening out the uncertain factors of agricultural product prices by combining the lagged correlation coefficient includes the following steps: S1231. Use candidate factors to construct a time-lagged feature matrix and calculate the conditional average treatment effect, factor importance, and effect direction stability of each candidate factor based on the causal forest model. S1232. Combine the lagged correlation coefficients of candidate factors to construct a comprehensive screening matrix that includes lagged correlation coefficients, causal effect strength, factor importance, and directional stability; S1233. Based on the preset thresholds, candidate factors that meet the requirements of time-lag correlation coefficient, factor importance and directional stability are screened to obtain the uncertainty factors of agricultural product prices.

5. The AI-based agricultural product futures trading prediction method according to claim 2, characterized in that: The method of constructing an agricultural product price uncertainty probability model based on the screened uncertainty factors using a mixed Gaussian model includes the following steps: S131. Standardize the uncertain factors after screening and collect agricultural product price fluctuation data as the target variable; S132. Based on the standardized uncertainty factors and target variables, train a Gaussian mixture model, determine the optimal number of Gaussian components using the Bayesian Information Criterion, and estimate the parameters of each component; S133. Use the parameters of the mixed Gaussian model to construct a probability model of agricultural product price uncertainty.

6. The AI-based agricultural product futures trading prediction method according to claim 1, characterized in that: The process of binning historical data according to the value range of the uncertainty factor and fitting the probability distribution function and cumulative distribution function corresponding to each bin in combination with the uncertainty probability model of agricultural product prices to form a bin distribution model includes the following steps: S21. Determine the value range of each uncertainty factor, divide the value range into several intervals, bin the historical data according to the divided intervals, and mark the bin category to which each data belongs; S22. For the data within each bin category, use the corresponding agricultural product price fluctuation data in combination with the mixed Gaussian model to fit the price fluctuation probability density function and cumulative distribution function under each bin; S23. Store the probability distribution function and cumulative distribution function parameters of all bins to form a bin distribution model.

7. The AI-based agricultural product futures trading prediction method according to claim 1, characterized in that: The construction of a cloud-edge collaborative agricultural product data perception network, collecting multi-source heterogeneous data in real time through edge nodes, and building a digital twin of the agricultural product supply chain includes the following steps: S31. Determine edge deployment nodes based on the agricultural product supply chain and use them to build a cloud-edge collaborative agricultural product data perception network. S32. Collect multi-source heterogeneous data from various links in the agricultural product supply chain in real time through edge nodes, and transmit the multi-source heterogeneous data to the cloud through a cloud-edge collaborative agricultural product data perception network; S33. Build a digital twin of the agricultural product supply chain based on multi-source heterogeneous data.

8. The AI-based agricultural product futures trading prediction method according to claim 1, characterized in that: The method of constructing and driving an artificial intelligence prediction model based on the digital twin of the agricultural product supply chain and the bin distribution model, generating agricultural product price scenarios through inverse transformation sampling, and outputting transaction prediction results based on probability distribution includes the following steps: S41. Obtain the status of agricultural product supply chain based on the digital twin of agricultural product supply chain; S42. Match the binning categories of the corresponding uncertain factors based on the agricultural product supply chain status output by the digital twin, and extract the corresponding price fluctuation probability distribution model; S43. Based on the probability distribution model under binning, the inverse transformation sampling method is used to generate multiple agricultural product price scenario samples; S44. Input the generated price scenario into the preset artificial intelligence prediction model for analysis, and output the agricultural product transaction prediction results based on probability distribution.

9. The AI-based agricultural product futures trading prediction method according to claim 8, characterized in that: The method of generating multiple agricultural product price scenario samples using an inverse transformation sampling method based on a probability distribution model under binning includes the following steps: S431. Based on the probability distribution model under binning, determine the target probability distribution type of agricultural product price fluctuations and obtain its cumulative distribution function; S432. Generate multiple uniformly distributed random numbers using a random number generator, and convert each random number into a price scenario sample based on the inverse function of the cumulative distribution function corresponding to the current bin; S433. Repeat steps S431-S432 until a preset number of price scenario samples are generated.

10. An AI-based agricultural product futures trading prediction system, used to implement the AI-based agricultural product futures trading prediction method according to any one of claims 1 to 9, characterized in that: The system includes: The uncertainty probability model construction module is used to collect historical data on agricultural product futures transactions, determine the uncertainty factors of agricultural product prices through correlation analysis, and construct an uncertainty probability model for agricultural product prices based on the uncertainty factors; The bin distribution model formation module is used to bin the historical data according to the value range of the uncertainty factor, and to fit the probability distribution function and cumulative distribution function corresponding to each bin in combination with the uncertainty probability model of agricultural product prices to form a bin distribution model; A digital twin construction module is used to build a cloud-edge collaborative agricultural product data perception network, collect multi-source heterogeneous data in real time through edge nodes, and build a digital twin of the agricultural product supply chain; The futures trading prediction module is used to build and drive an artificial intelligence prediction model based on the digital twin of the agricultural product supply chain and the bin distribution model. It generates agricultural product price scenarios through inverse transformation sampling and outputs trading prediction results based on probability distribution.

Images