A big data-based grain supply and demand intelligent management method and system

Abstract

The application discloses a kind of big data-based grain supply and demand intelligent management method and system, relating to grain supply and demand management technical field.The big data-based grain supply and demand intelligent management method and system obtain multi-source heterogeneous data through Internet of Things sensor and block chain technology, ensure the accuracy and integrity of data;Through the construction of grain supply and demand information network, the multi-source data is modeled using graph neural network, the correlation between data is analyzed, and the connection strength is dynamically updated;Through self-adaptive integrated learning algorithm, the correlation model constructed is identified for abnormal mode, a supply and demand fluctuation index is generated, an abnormal node of supply and demand fluctuation is detected, and key factors affecting supply and demand balance are identified;Through reinforcement learning algorithm, the supply and demand balance state is adjusted according to these key factors, and the dynamic optimization and real-time balance of supply and demand relationship are realized.Provide intelligent and efficient decision support for grain management department, help to ensure food security and stability.

Description

Technical Field

[0001] This invention relates to the field of grain supply and demand management technology, specifically to a big data-based intelligent management method and system for grain supply and demand. Background Technology

[0002] With the continued growth of the global population and the impact of climate change on agricultural production, the issue of food supply and demand balance has received increasing attention. The relationship between food supply and demand not only affects market stability but also directly relates to national food security. Therefore, governments and agricultural management departments worldwide urgently need to adopt more advanced technologies to achieve intelligent management of food production, distribution, and storage, ensuring the efficient and stable operation of the food supply chain. Against this backdrop, new technologies such as big data, artificial intelligence, and predictive models are gradually being applied to food supply and demand management. Through precise data analysis and intelligent decision-making, the management level of the food supply chain is improved, meeting the ever-growing food demands of society.

[0003] For example, a method, system, and medium for intelligent management of grain supply and demand based on big data, as disclosed in announcement number CN116894588B, includes: determining a time interval; obtaining historical grain supply and demand within the time interval from a grain information database based on big data technology; forecasting supply and demand based on a prediction model using the historical grain supply and demand; and issuing early warnings of supply and demand anomalies based on the forecast results; when an early warning of supply and demand anomalies occurs, obtaining factors influencing grain supply and demand, and performing correlation analysis based on the historical grain supply and demand to obtain highly correlated influencing factors; analyzing abnormal values ​​of the highly correlated influencing factors; determining abnormal supply and demand nodes based on the abnormal values ​​of the factors; and adjusting the supply and demand balance based on the abnormal supply and demand nodes. This invention utilizes big data to achieve intelligent analysis of supply and demand balance, emphasizes the correlation between grain supply and demand and influencing factors, provides reliable reference data for relevant departments, ensures dynamic balance of grain supply and demand, and promotes increased grain production and income.

[0004] Existing intelligent management methods for grain supply and demand based on big data rely on historical data analysis, neglecting real-time dynamic factors. This can lead to delayed forecasts that fail to reflect current market conditions and unforeseen events. Secondly, existing methods often employ simple linear models, failing to fully exploit the complex relationships between multi-source heterogeneous data, resulting in an incomplete identification of key factors affecting supply and demand balance. Furthermore, many systems lack in-depth analysis and dynamic adjustment capabilities regarding supply and demand relationships, making it difficult to effectively cope with market fluctuations and unforeseen risks in practice, thus limiting their application potential in complex agricultural environments. Therefore, a more intelligent and dynamic management method is urgently needed to improve the accuracy and flexibility of grain supply and demand management. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for intelligent management of grain supply and demand based on big data, thus solving the problems mentioned in the background.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a big data-based intelligent management method for grain supply and demand, comprising the following steps: S1. Acquiring multi-source heterogeneous data through IoT sensors and blockchain to construct a grain supply and demand information network; S2. Based on the grain supply and demand information network, modeling the multi-source data using a graph neural network, analyzing the correlation between the multi-source heterogeneous data, and dynamically updating the connection strength; S3. Using an adaptive ensemble learning algorithm to identify abnormal patterns in the constructed correlation model, generating a supply and demand fluctuation index, detecting abnormal nodes in supply and demand fluctuations, and identifying key factors affecting the supply and demand balance through node feature analysis; S4. Adjusting the grain supply and demand balance state using a reinforcement learning algorithm based on the key factors affecting the supply and demand balance, thereby achieving dynamic optimization and real-time balance of the supply and demand relationship.

[0007] Furthermore, the specific process of constructing a grain supply and demand information network by acquiring multi-source heterogeneous data through IoT sensors and blockchain is as follows: Multi-source heterogeneous data, including meteorological conditions, soil moisture, crop growth status, and market price information, is acquired through IoT sensors, and the collected data is stored in a decentralized manner through blockchain; the nodes of the grain supply and demand information network are designed, including producers, consumers, market factors, and environmental factors; based on the multi-source heterogeneous data, the connection relationships between nodes are constructed to form network edges, reflecting the interaction and dependence relationships between each node, thus constructing the grain supply and demand information network.

[0008] Furthermore, based on the grain supply and demand information network, the specific process of modeling multi-source data using graph neural networks is as follows: the grain supply and demand information network is transformed into a graph structure, and the relationship between nodes and edges is defined; features of multi-source heterogeneous data are extracted from each node to form a node feature matrix; graph convolution operations are applied to propagate and aggregate node features to update node representations; the parameters of the graph neural network are optimized through the backpropagation algorithm to train the model.

[0009] Furthermore, the specific process of analyzing the correlation between multi-source heterogeneous data and dynamically updating the connection strength is as follows: the similarity between nodes in the grain supply and demand information network is calculated by similarity measurement, and the corresponding feature vectors are extracted; based on the calculated similarity values, a correlation matrix is ​​constructed to represent the mutual dependence between nodes and is used as the input of the graph neural network; based on multi-source heterogeneous data, the correlation matrix is ​​adjusted through a weighting mechanism to update the connection strength between nodes.

[0010] Furthermore, the specific process of generating a supply and demand fluctuation index by performing abnormal pattern recognition on the constructed correlation model using an adaptive ensemble learning algorithm is as follows: Feature data for anomaly detection is extracted from the grain supply and demand information network, including historical supply and demand data, external influencing factors, and relevant node features; the historical dataset is divided into a training set and a test set, with the training set used for model learning and the test set used to evaluate model performance, marking data points in normal and abnormal states to form a supervised learning environment; the training set is used to model the data, and key features affecting supply and demand fluctuations are identified through feature importance evaluation, generating a supply and demand fluctuation index for each time point.

[0011] Furthermore, the specific process for detecting abnormal nodes in supply and demand fluctuations and identifying key factors affecting supply and demand balance through node feature analysis is as follows: Based on the supply and demand fluctuation index, an abnormal threshold is set to identify nodes with abnormal fluctuations; the supply and demand fluctuation index is monitored node by node, and nodes exceeding the abnormal threshold range are screened and recorded; feature data of the identified abnormal nodes are extracted, including historical supply and demand data, market dynamics, and meteorological conditions, and a feature matrix of abnormal nodes is established; clustering of the feature data of abnormal nodes is performed using cluster analysis methods to identify node groups with similar characteristics, determine the key influencing factors within each cluster, and assess the degree of influence of each factor on supply and demand balance.

[0012] Furthermore, based on the key factors affecting supply and demand balance, the specific process of adjusting the grain supply and demand balance using reinforcement learning algorithms is as follows: First, determine the reinforcement learning environment model, including a state space, action space, and reward mechanism. The state space represents the current state of the grain supply and demand system, the action space represents the selectable adjustment measures, and the reward mechanism is used to evaluate the impact of each action on the supply and demand balance. Second, based on the key influencing factors identified by cluster analysis, construct a state representation vector and transform it into the input of the reinforcement learning model. Third, initialize the parameters of the reinforcement learning algorithm. At each time step, select an action using the current state vector, execute the selected action, observe the changes in the system state, and record the new state vector and the obtained reward value. Fourth, update the strategy of the reinforcement learning model based on the observation results, and adjust the grain supply and demand balance through deep reinforcement learning algorithms to achieve real-time balance in the grain supply and demand relationship.

[0013] A big data-based intelligent management system for grain supply and demand includes the following modules: a data acquisition module, a graph modeling module, an anomaly detection module, and a dynamic adjustment module. The data acquisition module acquires multi-source heterogeneous data through IoT sensors and blockchain to construct a grain supply and demand information network. The graph modeling module models the multi-source data using a graph neural network based on the grain supply and demand information network, analyzes the correlations between the heterogeneous data, and dynamically updates the connection strength. The anomaly detection module uses an adaptive ensemble learning algorithm to identify abnormal patterns in the constructed correlation model, generates a supply and demand fluctuation index, detects abnormal nodes in supply and demand fluctuations, and identifies key factors affecting supply and demand balance through node feature analysis. The dynamic adjustment module adjusts the grain supply and demand balance state based on the key factors affecting supply and demand balance using a reinforcement learning algorithm, achieving dynamic optimization and real-time balance of the supply and demand relationship.

[0014] The present invention has the following beneficial effects:

[0015] (1) This intelligent management method for grain supply and demand based on big data utilizes IoT sensors to collect multi-dimensional grain supply and demand data in real time, enhancing the timeliness and accuracy of the data. Decentralized storage via blockchain technology ensures data security and immutability, improves information transparency, and provides a reliable foundation for subsequent analysis. Based on a grain supply and demand information network, graph neural networks are used to model multi-source data. Graph neural networks effectively capture and model the complex relationships between nodes, enabling in-depth analysis of data dependencies and interactions, and improving the understanding of dynamic changes in grain supply and demand. Dynamically updating connection strength helps to reflect market changes in a timely manner and enhances decision-making agility.

[0016] (2) This big data-based intelligent management system for grain supply and demand uses an adaptive ensemble learning algorithm to identify abnormal patterns in the constructed correlation model, generating a supply and demand fluctuation index. This improves the accuracy and robustness of the model, effectively identifying abnormal fluctuation patterns and generating the supply and demand fluctuation index. It provides quantitative reference data and, through in-depth analysis of abnormal nodes, helps identify key factors affecting the supply and demand balance, promoting scientific decision-making. Based on the key factors affecting the supply and demand balance, the system uses a reinforcement learning algorithm to adjust the grain supply and demand balance, achieving dynamic optimization and real-time balance of the supply and demand relationship. This process enhances the system's intelligence level, helps cope with complex and dynamic market environments, and achieves efficient resource allocation and effective risk control.

[0017] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0018] Figure 1 This is a flowchart of a big data-based intelligent management method for grain supply and demand according to the present invention.

[0019] Figure 2 This is a flowchart of a big data-based intelligent management system for grain supply and demand according to the present invention. Detailed Implementation

[0020] This application provides a big data-based intelligent management method and system for grain supply and demand, which solves the problems of insufficient response to dynamic market changes and incomplete identification of key influencing factors in traditional grain supply and demand management.

[0021] The problem addressed in this application's embodiments can be summarized as follows:

[0022] By acquiring multi-source heterogeneous data through IoT sensors and blockchain, a food supply and demand information network can be built.

[0023] Based on the grain supply and demand information network, a graph neural network is used to model multi-source data, analyze the correlation between multi-source heterogeneous data, and dynamically update the connection strength.

[0024] Anomaly pattern recognition is performed on the constructed correlation model using an adaptive ensemble learning algorithm to generate a supply and demand fluctuation index, detect abnormal nodes in supply and demand fluctuations, and identify key factors affecting the supply and demand balance through node feature analysis.

[0025] Based on the key factors affecting the supply and demand balance, reinforcement learning algorithms are used to adjust the state of grain supply and demand balance, thereby achieving dynamic optimization and real-time balance of the supply and demand relationship.

[0026] Please see Figure 1 This invention provides a technical solution: a big data-based intelligent management method for grain supply and demand, comprising the following steps: S1. Acquiring multi-source heterogeneous data through IoT sensors and blockchain to construct a grain supply and demand information network; S2. Based on the grain supply and demand information network, modeling the multi-source data using a graph neural network, analyzing the correlation between the multi-source heterogeneous data, and dynamically updating the connection strength; S3. Using an adaptive ensemble learning algorithm to identify abnormal patterns in the constructed correlation model, generating a supply and demand fluctuation index, detecting abnormal nodes in supply and demand fluctuations, and identifying key factors affecting the supply and demand balance through node feature analysis; S4. Adjusting the grain supply and demand balance state using a reinforcement learning algorithm based on the key factors affecting the supply and demand balance, thereby achieving dynamic optimization and real-time balance of the supply and demand relationship.

[0027] In this implementation plan, S1. Multi-source heterogeneous data is acquired through IoT sensors and blockchain to construct a grain supply and demand information network. In this stage, IoT sensors collect real-time data on grain production, circulation, and consumption, while blockchain technology ensures data security and reliability. By integrating this multi-source heterogeneous data (market demand, weather conditions, production capacity), a grain supply and demand information network is constructed, providing a foundation for subsequent data analysis and decision-making. S2. Based on the grain supply and demand information network, graph neural networks are used to model the multi-source data, analyze the correlations between the heterogeneous data, and dynamically update the connection strength. This step utilizes graph neural networks (GNNs) for deep learning modeling of the constructed grain supply and demand information network, aiming to reveal the relationships between different data nodes. By analyzing the correlations between these nodes and dynamically updating the connection strength between nodes, changes in the market environment and supply and demand relationships are reflected. This method enables the model to capture trends in supply and demand changes in real time. S3. Anomaly pattern recognition is performed on the constructed correlation model using an adaptive ensemble learning algorithm to generate a supply-demand fluctuation index, detect anomalous nodes in supply-demand fluctuations, and identify key factors affecting supply-demand balance through node feature analysis. In this stage, the adaptive ensemble learning algorithm is applied to train the previously constructed correlation model and perform anomaly pattern recognition. This algorithm can combine the prediction results of multiple models to improve the accuracy of anomaly detection. The generated supply-demand fluctuation index can reflect the real-time changes in market supply and demand, while identifying nodes of anomalous fluctuations. By analyzing the characteristics of these nodes, key factors affecting supply-demand balance can be identified, providing a basis for decision-making. S4. Based on the key factors affecting supply-demand balance, the grain supply-demand balance is adjusted using a reinforcement learning algorithm to achieve dynamic optimization and real-time balance of supply and demand. Based on the key factors identified in step three, a reinforcement learning algorithm is used to intelligently adjust the grain supply-demand balance. Reinforcement learning, through continuous trial and error and feedback mechanisms, can optimize the decision-making process, thereby achieving dynamic adjustment of supply and demand, ensuring effective grain supply and market stability. This step achieves real-time monitoring and optimization of grain supply and demand, enabling management decisions to respond quickly to market changes. The Internet of Things (IoT) refers to connecting various information sensing devices to the internet for data exchange and communication. IoT sensors are devices used to collect environmental data (temperature, humidity, light intensity, soil moisture) in real time. These sensors can be deployed in various locations such as farmland, warehouses, and markets to monitor various factors affecting food supply and demand in real time. Blockchain is a distributed ledger technology that securely and transparently records data transactions. It ensures the immutability and reliability of data through decentralization. In food supply and demand management, blockchain can be used to record data on food production, transportation, and sales, ensuring the authenticity and traceability of information and helping managers make more reliable decisions.The grain supply and demand information network is a network composed of various information nodes (producers, distributors, and consumers) and the relationships between them. This network aggregates various supply and demand-related data. By analyzing this data, we can understand market demand, production capacity, and dynamic changes in supply and demand, providing a basis for decision-making. Graph Neural Networks (GNNs) are deep learning algorithms that process graph-structured data, effectively capturing the relationships and features between nodes. In grain supply and demand management, GNNs are used to model the grain supply and demand information network. By learning the characteristics of nodes and their connections, they analyze the complex correlations between multi-source data, thereby achieving more accurate supply and demand forecasts. Adaptive Ensemble Learning (AEL) is a method that improves predictive performance by combining multiple learning algorithms. It leverages the advantages of different models to enhance anomaly detection. In grain supply and demand management, this algorithm is used to identify abnormal patterns in supply and demand fluctuations, ensuring the accuracy of the supply and demand fluctuation index. The supply and demand fluctuation index is a comprehensive indicator reflecting the degree and amplitude of changes in grain supply and demand. By monitoring this index, managers can promptly identify anomalies in the market and take corresponding measures to adjust supply and demand relationships, ensuring market stability. Reinforcement learning is a machine learning method that learns optimal strategies through interaction with the environment. It adjusts the decision-making process based on feedback mechanisms, enabling the algorithm to optimize decisions in a constantly changing environment. In food supply and demand management, reinforcement learning algorithms can dynamically adjust the food supply and demand balance based on changes in key factors, achieving real-time optimization.

[0028] Specifically, the process of constructing a grain supply and demand information network by acquiring multi-source heterogeneous data through IoT sensors and blockchain is as follows: Multi-source heterogeneous data, including meteorological conditions, soil moisture, crop growth status, and market price information, is acquired through IoT sensors and stored in a decentralized manner using blockchain; the nodes of the grain supply and demand information network are designed, including producers, consumers, market factors, and environmental factors; based on the multi-source heterogeneous data, connections between nodes are constructed to form network edges, reflecting the interactions and dependencies between nodes, thus building the grain supply and demand information network.

[0029] This implementation plan acquires multi-source heterogeneous data: IoT sensors: Various types of data are collected in real time through IoT sensors deployed in farmland and markets. This data includes: Meteorological conditions: Temperature, humidity, precipitation, and wind speed, which directly affect crop growth and market demand. Soil moisture: Monitoring soil moisture levels affects water supply and irrigation needs for crop growth. Crop growth status: Monitoring crop growth processes through sensors, including growth stages, health status, and pest and disease conditions. Market price information: Collecting real-time prices of grain products in the market, reflecting supply and demand and market dynamics. Decentralized data storage via blockchain: Decentralized storage: Data collected through IoT is stored on the blockchain. Blockchain technology ensures data security and reliability, with characteristics including: Data immutability: Once recorded on the blockchain, data cannot be altered, guaranteeing its authenticity and validity. Transparency: All participants can access the same data, promoting information sharing and cooperation. Decentralization: Avoiding single points of failure, improving system security and stability. Designing nodes for the grain supply and demand information network: Node definition: When constructing the grain supply and demand information network, the nodes in the network must first be designed. Key nodes include: Producers: Grain growers or production units responsible for producing and supplying grain. Consumers: End users of grain, including individuals and businesses, directly impacting market demand. Market: As a place for trading and circulation, reflecting supply and demand and price changes. Environmental factors: Including other factors affecting grain production and the market, such as climate change and policies and regulations. Building connections between nodes: Establishing connections: Based on multi-source heterogeneous data analysis, the interactions and dependencies between nodes are analyzed, forming network edges through the following methods: Data analysis: Analyzing the relationship between meteorological conditions, soil moisture, crop growth status, and market prices to identify the mutual influence of each node in grain supply and demand. Interaction: The crop growth status of producers directly affects the supply in the market, thus influencing consumer purchasing behavior. Forming a grain supply and demand information network: Network construction: Through the above steps, a complete grain supply and demand information network is finally formed. This network demonstrates the connections and interactions between different nodes, making information flow more efficient and supporting subsequent analysis and decision-making: Information flow: Information in the network can flow rapidly between nodes, improving response speed. Decision support: By analyzing the network, managers can better understand supply and demand dynamics and formulate corresponding management strategies.

[0030] Specifically, the process of modeling multi-source data using graph neural networks based on the grain supply and demand information network is as follows: the grain supply and demand information network is transformed into a graph structure, and the relationship between nodes and edges is defined; features of multi-source heterogeneous data are extracted from each node to form a node feature matrix; graph convolution operations are applied to propagate and aggregate the node features and update the node representation; the parameters of the graph neural network are optimized through the backpropagation algorithm, and the model is trained.

[0031] In this implementation plan, the grain supply and demand information network is transformed into a graph structure: Graph Structure Definition: The grain supply and demand information network is viewed as a graph containing nodes and edges. Nodes: Represent different entities in the grain supply and demand information network, including producers, consumers, markets, weather conditions, and soil moisture. Edges: Represent the relationships between nodes, the supply relationship between producers and consumers, the purchasing relationship between consumers and the market, and the influence relationship between weather conditions and producers. Defining the Relationships Between Nodes and Edges: Relationship Modeling: Clarify the relationships between each node and define the weights of the edges to reflect the strength of different relationships. Supply Relationships: The weights of the edges between producers and consumers can be set according to the ratio of supply to demand. Influence Relationships: The impact of weather conditions on crop growth can be quantified into edge weights through statistical analysis of historical data. Extracting Features from Multi-Source Heterogeneous Data: Feature Extraction: Extract multi-source heterogeneous data features related to grain supply and demand from each node to form a node feature matrix. Node Features: These may include weather conditions, soil moisture, crop growth status, market prices, etc., which affect supply and demand relationships and market dynamics. Feature Matrix: The features of all nodes are combined into a feature matrix of shape N×F (where N is the number of nodes and F is the number of features) for subsequent processing. Applying Graph Convolution Operations: Graph Convolution: Graph convolution operations are used to propagate and aggregate node features. Node Propagation: Each node passes its feature information to its neighboring nodes, which use the received information to update their own representations. Feature Aggregation: Through weighted averaging or other aggregation methods, the features of neighboring nodes are integrated into the current node to form a new node representation, updating the node's feature information. Updating Node Representations: Node Representation Update: After the graph convolution operation, the feature vector of each node is updated, reflecting its new state and information in the network. The core of graph convolution operations lies in updating node features through the information of neighboring nodes. The update formula is: Parameter explanation: H (l) : The node feature matrix of the l-th layer. The normalized adjacency matrix is ​​typically achieved by adding self-loops and then normalizing. (l) σ: The learnable weight matrix of the l-th layer. σ: The activation function (ReLU). Output: The updated feature representation H for each node. (l+1)New Feature Vector: The updated node feature vector not only contains the node's feature information but also incorporates the features of its neighboring nodes, thus more comprehensively reflecting its role and relationships within the entire supply and demand network. Optimization of Graph Neural Network Parameters via Backpropagation: Model Training: The parameters of the graph neural network are optimized using the backpropagation algorithm. Loss Function: A loss function is defined to evaluate the model's performance based on the difference between the model's predicted output and the true label. Backpropagation: Based on the value of the loss function, the weights of the parameters in each layer of the network are adjusted, and the loss function is minimized through gradient descent or other optimization algorithms, gradually improving the model's accuracy.

[0032] Specifically, the process of analyzing the correlation between multi-source heterogeneous data and dynamically updating the connection strength is as follows: The similarity between nodes in the grain supply and demand information network is calculated using a similarity metric, and corresponding feature vectors are extracted; based on the calculated similarity values, a correlation matrix is ​​constructed to represent the interdependencies between nodes and serves as the input to the graph neural network; based on the multi-source heterogeneous data, the correlation matrix is ​​adjusted through a weighting mechanism to update the connection strength between nodes.

[0033] In this implementation plan, the similarity between nodes in the grain supply and demand information network is calculated, and the similarity value between nodes is calculated based on the node feature vectors. If the feature vectors of node i and node j are f i and f j Then the formula for calculating cosine similarity is: Sim(i,j) represents the similarity between node i and node j. The calculated similarity value reflects the similarity between node i and node j. An association matrix is ​​constructed based on the calculated similarity values. The similarity values ​​are organized into matrix A, where element A[i][j] represents the similarity between node i and node j. The matrix has an n×n dimension (where n is the number of nodes), comprehensively representing the interdependencies between nodes. The association matrix is ​​used as input to the graph neural network. Application: The graph neural network uses this association matrix to propagate node information and updates node features through graph convolution operations. The connection strength of nodes is dynamically adjusted according to the similarity values ​​to ensure the model can adapt to different supply and demand states. Based on multi-source heterogeneous data, the association matrix is ​​adjusted through a weighting mechanism. Method: A weight parameter w is introduced to weight the association matrix. The updated association matrix A' can be expressed as: A'=W⊙A; where ⊙ represents element-wise multiplication, and W is the weight matrix, dynamically adjusted according to the relevance and importance of features. The weight settings can be appropriately adjusted based on external influencing factors such as external market dynamics and weather changes. The weighted correlation matrix is ​​used to update the connection strength between nodes. Application: Through convolutional operations in graph neural networks, node features are propagated and aggregated, ensuring that each node's features are influenced not only by its own features but also by the features of its neighboring nodes. This dynamically adjusts connection strength to reflect changes in current supply and demand relationships.

[0034] Specifically, the process of generating a supply and demand fluctuation index by performing anomaly pattern recognition on the constructed correlation model using an adaptive ensemble learning algorithm is as follows: Feature data for anomaly detection is extracted from the grain supply and demand information network, including historical supply and demand data, external influencing factors, and relevant node features; the historical dataset is divided into a training set and a test set, with the training set used for model learning and the test set used to evaluate model performance, marking data points in normal and abnormal states to form a supervised learning environment; the training set is used to model the data, and key features affecting supply and demand fluctuations are identified through feature importance evaluation, generating a supply and demand fluctuation index for each time point.

[0035] In this implementation plan, feature data for anomaly detection is extracted from the grain supply and demand information network, mainly including the following: Historical supply and demand data: reflecting grain production and consumption within a specific time period, and monthly grain supply and demand. External influencing factors: including climate conditions, market price fluctuations, and policy changes, which may affect grain production and circulation. Relevant node features: producer production capacity, consumer purchasing power, and market liquidity. The extracted historical dataset is divided into training and testing sets. The specific steps are as follows: Training set: accounting for 70%-80% of the data, used for model training to identify normal and abnormal patterns. Testing set: accounting for 20%-30% of the data, used to evaluate the performance of the trained model and verify the model's adaptability to new data. A supervised learning environment is formed by labeling data points in normal and abnormal states. This can be achieved through expert annotation or statistical methods based on historical records. The process of generating the supply and demand fluctuation index is as follows: The supply and demand fluctuation index integrates multiple features and introduces a weighting mechanism to reflect the importance of each feature. There are m features, including historical supply and demand data, external influencing factors, and relevant node features. The supply and demand fluctuation index is as follows: Explanation of parameters in the formula The supply and demand forecast at time point t represents the predicted supply and demand balance. i The weight of the i-th feature reflects its influence on supply and demand fluctuations. The weight can be obtained through feature importance assessment and satisfies the following condition: External influencing factors I t The relevant characteristic function describes the impact of specific factors on supply and demand fluctuations. It can be defined as: in, Let be the mean of the i-th feature, and σ(I(i)) be its standard deviation, reflecting the degree of standardization of the feature over its historical performance. Volatility(D) t Historical supply and demand data D t The volatility of supply and demand can be measured by calculating the standard deviation of supply and demand: in, The mean of the supply and demand data is given, and N is the number of data points. α and β are the weighting coefficients for the supply and demand forecasts and volatility measures, respectively, reflecting the importance of each component in the overall index.

[0036] Specifically, the process of detecting abnormal nodes in supply and demand fluctuations and identifying key factors affecting supply and demand balance through node feature analysis is as follows: Based on the supply and demand fluctuation index, an abnormal threshold is set to identify nodes with abnormal fluctuations; the supply and demand fluctuation index is monitored node by node, and nodes exceeding the abnormal threshold range are screened and recorded; feature data of the identified abnormal nodes are extracted, including historical supply and demand data, market dynamics, and meteorological conditions, and a feature matrix of abnormal nodes is established; clustering of the feature data of abnormal nodes is performed using cluster analysis methods to identify groups of nodes with similar characteristics, determine the key influencing factors within each cluster, and assess the degree of influence of each factor on supply and demand balance.

[0037] In this implementation scheme, statistical analysis is performed based on historical supply and demand fluctuation indices to calculate the mean and standard deviation. By setting a threshold (mean plus two or three times the standard deviation), normal and abnormal states can be effectively distinguished, helping to identify nodes that significantly deviate from the normal range. The supply and demand fluctuation index of each node is monitored in real time. All nodes in the network are traversed, and their fluctuation indices are calculated and compared with the set abnormal threshold. Nodes exceeding the threshold are marked as abnormal nodes, and their relevant information, node ID, and fluctuation index value are recorded. For the identified abnormal nodes, feature data related to supply and demand balance is extracted. These features include historical supply and demand data, market dynamics (market prices), and meteorological conditions (rainfall and temperature). This data is integrated to form a feature matrix for subsequent analysis. Cluster analysis is performed on the feature data of abnormal nodes. An appropriate clustering algorithm (K-means DBSCAN) is selected to group nodes with similar features. The clustering results will help identify node groups and their characteristics, thereby finding nodes with similar supply and demand fluctuation patterns. By further analyzing the nodes within each cluster, the impact of each feature on supply and demand balance is evaluated. Statistical methods are used to identify the importance of each factor and summarize the key influencing factors. The identification of these factors will provide an important basis for adjusting and optimizing the supply and demand relationship.

[0038] Specifically, the process of adjusting the grain supply and demand balance using reinforcement learning algorithms, based on key factors affecting the supply and demand balance, is as follows: First, a reinforcement learning environment model is determined, including a state space, action space, and reward mechanism. The state space represents the current state of the grain supply and demand system, the action space represents selectable adjustment measures, and the reward mechanism is used to evaluate the impact of each action on the supply and demand balance. Second, based on the key influencing factors identified through cluster analysis, a state representation vector is constructed and transformed into the input of the reinforcement learning model. Third, the parameters of the reinforcement learning algorithm are initialized. At each time step, an action is selected using the current state vector, the selected action is executed, and the changes in the system state are observed, recording the new state vector and the obtained reward value. Fourth, the strategy of the reinforcement learning model is updated based on the observation results, and the grain supply and demand balance is adjusted using a deep reinforcement learning algorithm to achieve real-time balance in the grain supply and demand relationship.

[0039] In this implementation scheme, the state space defines the current state of the grain supply and demand system, including market supply and demand, inventory levels, price fluctuations, and weather conditions. Each state reflects the characteristics of the system at a specific point in time. The action space lists possible adjustment measures, including increasing or decreasing production, adjusting storage strategies, and changing sales strategies. Each action represents a strategy the system can take to improve the supply-demand balance. A reward mechanism is designed to evaluate the impact of each action on the supply-demand balance. The reward value should reflect whether the supply-demand relationship has improved after executing a specific action; commonly used reward indicators may include the stability of inventory levels and the magnitude of market price fluctuations. Based on key influencing factors identified through cluster analysis, these factors are integrated into a state representation vector. This vector describes the current system state, including the values ​​of each influencing factor, and is transformed into the input of the reinforcement learning model. The parameters of the reinforcement learning algorithm are initialized, typically including the learning rate, discount factor, and exploration strategy. These parameters have a significant impact on the model's learning speed and strategy updates. At each time step, based on the current state vector, an action is determined by selecting a strategy (ε-greedy strategy). After the selected action is executed, the system state changes. Record the new state vector and corresponding reward value for subsequent learning and optimization. Based on the observed results (new state vector and reward value), update the model's policy using a deep reinforcement learning algorithm. This update process is typically implemented through a value function or policy gradient method, with the aim of adjusting the policy to improve the expected reward for future action choices.

[0040] Please see Figure 2A big data-based intelligent management system for grain supply and demand includes the following modules: a data acquisition module, a graph modeling module, an anomaly detection module, and a dynamic adjustment module. The data acquisition module acquires multi-source heterogeneous data through IoT sensors and blockchain to construct a grain supply and demand information network. The graph modeling module models multi-source data using graph neural networks based on the grain supply and demand information network, analyzes the correlation between heterogeneous data sources, and dynamically updates connection strength. The anomaly detection module uses an adaptive ensemble learning algorithm to identify abnormal patterns in the constructed correlation model, generates a supply and demand fluctuation index, detects abnormal nodes in supply and demand fluctuations, and identifies key factors affecting supply and demand balance through node feature analysis. The dynamic adjustment module adjusts the grain supply and demand balance state based on the key factors affecting supply and demand balance using reinforcement learning algorithms, achieving dynamic optimization and real-time balance of supply and demand.

[0041] In this implementation plan, the data acquisition module is responsible for acquiring multi-source heterogeneous data through IoT sensors and blockchain technology to construct a grain supply and demand information network. IoT sensors collect various data related to grain supply and demand in real time, including information on weather conditions, soil moisture, crop growth status, and market prices. Decentralized storage: Blockchain technology is used to store the collected data in a decentralized manner, ensuring data security, reliability, and traceability, and preventing data tampering and loss. Information integration: Multi-source heterogeneous data is integrated to form a comprehensive grain supply and demand information network, providing a foundation for subsequent data analysis and modeling. Graph modeling module: Based on the grain supply and demand information network, graph neural networks are used to model the multi-source data, analyze the correlation between data, and dynamically update the connection strength. Graph structure transformation: The grain supply and demand information network is transformed into a graph structure, defining nodes. Feature extraction: Features of the multi-source heterogeneous data are extracted from each node to form a node feature matrix for graph convolution operations. Correlation analysis: The correlation between nodes is analyzed through graph neural networks, and the connection strength is dynamically updated, enabling the model to reflect real-time data changes and improving the accuracy of prediction and analysis. Anomaly Detection Module: Utilizes an adaptive ensemble learning algorithm to identify anomaly patterns in the constructed correlation model, generating a supply-demand fluctuation index and detecting anomalous nodes in supply-demand fluctuations. Feature Extraction: Extracts historical supply-demand data, external influencing factors, and node features from the grain supply-demand information network to form a feature dataset. Anomaly Pattern Recognition: Models the dataset using an adaptive ensemble learning algorithm to identify anomalous patterns in supply-demand fluctuations, generating a supply-demand fluctuation index to reflect market dynamics. Key Factor Identification: Analyzes the feature data of anomalous nodes to identify key factors affecting supply-demand balance, helping decision-makers understand changes and trends in supply-demand relationships. Dynamic Adjustment Module: Based on the identified key influencing factors, adjusts the grain supply-demand balance using a reinforcement learning algorithm to achieve dynamic optimization and real-time balance. Environment Model Establishment: Determines the environment model for reinforcement learning, including state space, action space, and reward mechanism, to facilitate intelligent adjustment of the supply-demand system. Strategy Optimization: Selects appropriate adjustment measures for the current supply-demand state using a reinforcement learning algorithm, evaluates their effects in real time, and updates the strategy to optimize the supply-demand balance. Real-time feedback: Based on market feedback and changes in system status, the supply and demand relationship is continuously adjusted to achieve dynamic optimization and ensure the real-time nature and effectiveness of the grain supply and demand balance.

[0042] In summary, this application has at least the following effects:

[0043] By acquiring multi-source heterogeneous data through IoT sensors and blockchain technology, the system ensures the accuracy and completeness of grain supply and demand information, thus providing a reliable foundation for subsequent analysis. Utilizing graph neural networks and adaptive ensemble learning algorithms, the system can dynamically analyze and identify supply and demand fluctuations, promptly detect anomalies, and promote real-time optimization and adjustment of supply and demand relationships. Through the application of deep learning and reinforcement learning algorithms, the system can provide precise decision support based on real-time data and market changes, helping management departments make scientific decisions regarding grain supply and demand management. The system can provide personalized supply and demand matching for different market conditions and user needs, enhancing the utilization efficiency of grain resources and improving the satisfaction of relevant participants. Through intelligent management and real-time monitoring, the system effectively reduces the risks caused by fluctuations in grain supply and demand, ensuring food security and promoting sustainable development.

[0044] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0045] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0046] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0047] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0048] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0049] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for intelligent management of grain supply and demand based on big data, characterized in that, Includes the following steps: S1. Obtain multi-source heterogeneous data through IoT sensors and blockchain to construct a grain supply and demand information network. The multi-source heterogeneous data includes meteorological conditions, soil moisture, crop growth status and market price information. The nodes of the grain supply and demand information network include producers, consumers, markets and environmental factors. S2. Based on the aforementioned grain supply and demand information network, a graph neural network is used to model multi-source data, analyze the correlation between heterogeneous multi-source data, and dynamically update the connection strength. Specifically, the similarity between nodes in the grain supply and demand information network is calculated using a similarity metric, corresponding feature vectors are extracted, and a correlation matrix is ​​constructed based on the calculated similarity values. This correlation matrix represents the interdependencies between nodes and serves as the input to the graph neural network. Based on the heterogeneous multi-source data, the correlation matrix is ​​adjusted using a weighting mechanism to update the connection strength between nodes. S3. Anomaly pattern recognition is performed on the constructed correlation model using an adaptive ensemble learning algorithm to generate a supply and demand fluctuation index, detect abnormal nodes in supply and demand fluctuations, and identify key factors affecting supply and demand balance through node feature analysis; wherein, feature data for anomaly detection is extracted from the grain supply and demand information network, the feature data includes historical supply and demand data, external influencing factors, and relevant node features, the historical dataset is divided into a training set and a test set, the training set is used for model learning, the test set is used to evaluate the model performance, data points in normal and abnormal states are marked to form a supervised learning environment, modeling is performed on the training set using the adaptive ensemble learning algorithm, key features affecting supply and demand fluctuations are identified through feature importance evaluation, and a supply and demand fluctuation index is generated for each time point; Based on the supply and demand fluctuation index, an abnormal threshold is set, nodes with abnormal fluctuations are identified, the supply and demand fluctuation index is monitored node by node, nodes exceeding the abnormal threshold range are screened and recorded, feature data of the identified abnormal nodes are extracted, the feature data of the identified abnormal nodes include historical supply and demand data, market dynamics, and meteorological conditions, a feature matrix of abnormal nodes is established, the feature data of the abnormal nodes are clustered by cluster analysis method, nodes with similar characteristics are identified, key influencing factors within each cluster are determined, and the degree of influence of each factor on the supply and demand balance is evaluated; S4. Based on the key factors affecting supply and demand balance, a reinforcement learning algorithm is used to adjust the state of grain supply and demand balance, achieving dynamic optimization and real-time balance of supply and demand. Specifically, a reinforcement learning environment model is determined, comprising a state space, an action space, and a reward mechanism. The state space represents the current state of the grain supply and demand system, the action space represents selectable adjustment measures, and the reward mechanism is used to evaluate the impact of each action on supply and demand balance. A state representation vector is constructed based on key influencing factors identified through cluster analysis. This state representation vector is then transformed into the input of the reinforcement learning model, initializing the parameters of the reinforcement learning algorithm. At each time step, an action is selected using the current state vector, executed, and the changes in the system state are observed. The new state vector and the obtained reward value are recorded. The strategy of the reinforcement learning model is updated based on the observation results. The grain supply and demand balance is adjusted through a deep reinforcement learning algorithm, ensuring real-time balance of grain supply and demand.

2. The intelligent grain supply and demand management method based on big data according to claim 1, characterized in that: The specific process of constructing a grain supply and demand information network by acquiring multi-source heterogeneous data through IoT sensors and blockchain is as follows: Multi-source heterogeneous data is acquired through IoT sensors, and the collected data is stored in a decentralized manner through blockchain. The nodes of the food supply and demand information network are designed, including producers, consumers, markets, and environmental factors. Based on multi-source heterogeneous data, connections between nodes are constructed to form network edges, reflecting the interactions and dependencies between nodes, thus building a grain supply and demand information network.

3. The intelligent grain supply and demand management method based on big data according to claim 1, characterized in that: The specific process of modeling multi-source data using graph neural networks based on the grain supply and demand information network is as follows: Transform the grain supply and demand information network into a graph structure and define the relationships between nodes and edges; Features of multi-source heterogeneous data are extracted from each node to form a node feature matrix. Graph convolution operation is applied to propagate and aggregate the node features and update the node representation. The parameters of the graph neural network are optimized using the backpropagation algorithm, and the model is trained.

4. The intelligent grain supply and demand management method based on big data according to claim 1, characterized in that: The similarity metric includes cosine similarity. If the feature vectors of node i and node j are fi and fj, respectively, then the similarity between node i and node j is calculated using cosine similarity. Based on the calculated similarity values, an association matrix A is constructed, where the element Aij represents the similarity between node i and node j. The correlation matrix A is weighted by the weight matrix W to obtain the updated correlation matrix A', and the connection strength between nodes is updated by the updated correlation matrix A'.

5. The intelligent grain supply and demand management method based on big data according to claim 1, characterized in that: When generating the supply and demand volatility index, the volatility measures of supply and demand status prediction results, each feature weight item and historical supply and demand data are combined. The feature weights in each feature weight item are obtained through feature importance assessment, and each feature weight meets the normalization requirements. The volatility measures of historical supply and demand data are obtained by calculating the standard deviation of supply and demand to reflect the degree of volatility of historical supply and demand data within the corresponding time range.

6. The intelligent grain supply and demand management method based on big data according to claim 1, characterized in that: When setting the abnormal threshold, statistical analysis is performed based on the historical supply and demand fluctuation index to calculate the mean and standard deviation, and the mean plus two or three times the standard deviation is used as the abnormal threshold. The supply and demand fluctuation index of each node is monitored in real time. All nodes in the network are traversed, and their supply and demand fluctuation index is calculated and compared with the set abnormal threshold. Nodes that exceed the threshold are marked as abnormal nodes, and the node ID and fluctuation index value are recorded. The clustering analysis method includes K-means or DBSCAN.

7. The intelligent grain supply and demand management method based on big data according to claim 1, characterized in that: The state space includes market supply and demand, inventory levels, price fluctuations, and weather conditions; the adjustment measures in the action space include increasing or decreasing production, adjusting storage strategies, and changing sales strategies; the reward value in the reward mechanism reflects whether the supply and demand relationship is improved after performing a specific action, and the reward indicators include the stability of inventory levels and the volatility of market prices; at each time step, based on the current state vector, the action to be taken is determined through an ε-greedy strategy.

8. A big data-based intelligent management system for grain supply and demand, employing the big data-based intelligent management method for grain supply and demand as described in any one of claims 1-7, characterized in that, It includes the following modules: data acquisition module, graph modeling module, anomaly detection module, and dynamic adjustment module; The data acquisition module is used to acquire multi-source heterogeneous data through IoT sensors and blockchain to build a grain supply and demand information network. The graph modeling module is used to model multi-source data based on the grain supply and demand information network using a graph neural network, analyze the correlation between multi-source heterogeneous data, and dynamically update the connection strength. The graph modeling module is also used to calculate the similarity between nodes in the grain supply and demand information network using a similarity metric, construct a correlation matrix, and adjust the correlation matrix based on the multi-source heterogeneous data using a weighting mechanism to update the connection strength between nodes. The anomaly detection module is used to identify abnormal patterns in the constructed correlation model using an adaptive ensemble learning algorithm, generate a supply and demand fluctuation index, detect abnormal nodes of supply and demand fluctuations, and identify key factors affecting the supply and demand balance through node feature analysis. The anomaly detection module is also used to monitor the supply and demand fluctuation index node by node, filter nodes exceeding the abnormal threshold range, extract historical supply and demand data, market dynamics, and meteorological conditions of the identified abnormal nodes, establish a feature matrix of the abnormal nodes, and identify groups of nodes with similar characteristics and key influencing factors through cluster analysis. The dynamic adjustment module is used to adjust the grain supply and demand balance state through reinforcement learning algorithms based on key factors affecting the supply and demand balance, so as to achieve dynamic optimization and real-time balance of supply and demand relationship. The dynamic adjustment module is also used to construct state representation vectors based on key influencing factors identified by cluster analysis, convert the state representation vectors into inputs of reinforcement learning models, and update the strategy of reinforcement learning models according to the new state vectors and reward values ​​after the action is executed.

Images