Agricultural product big data service system based on trusted data space

Abstract

The invention relates to the technical field of agricultural product big data processing and application, and discloses an agricultural product big data service system based on a trusted data space. The system comprises a multi-source data integration module for collecting agricultural product full-chain multi-dimensional data; the data cleaning and arranging module is used for cleaning and dynamically arranging the data; the data association graph construction module is used for generating a cross-domain data association graph; the multi-dimensional analysis engine module is used for extracting a data evolution mode and generating classification decision features and risk early warning indexes; and the service interface generation module is used for constructing a configurable service interface set. The multi-dimensional data covers production bases, processes, product attributes, market conditions and cold-chain logistics data. The system performs normalization, smooth correction and other processing on data, realizes data association analysis, feature fusion and other functions through multiple algorithms, provides adaptive data service schemes for different scenes, assists digital development of the agricultural product industry, and realizes accurate decision making and risk early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural product big data processing and application, and specifically to an agricultural product big data service system based on a trusted data space. Background Art

[0002] In the agricultural product industry, with the development of information technology, the amount of data has increased explosively. However, there are many problems with this data, which seriously restricts the digital development and refined management of the agricultural product industry. The data sources are extensive but scattered. The agricultural product industrial chain covers multiple links such as production, processing, circulation, and sales, and each link involves different data types. In the production link, production base data includes base certification information, scale and output indicators, etc., and production process data involves environmental monitoring parameters and operation records; in the circulation link, cold chain logistics data records information such as temperature and humidity during transportation; at the sales end, market condition data reflects price fluctuations, supply and demand relationships, etc. These data are stored in systems of different entities respectively, with inconsistent formats and a lack of effective integration mechanisms, resulting in difficulty in sharing and comprehensively utilizing the data, forming individual "data islands". For example, it is difficult for agricultural product production enterprises to obtain market condition data to accurately plan production, and logistics enterprises cannot reasonably arrange transportation resources based on the actual output of production bases.

[0003] The data quality is uneven. Due to differences in collection equipment, personnel operations, etc., there are a large number of noises, errors, and missing values in the data. Taking the environmental monitoring data in the production process as an example, sensor failures may lead to abnormal values in the monitoring data, and human errors may occur when manually recording operation records. If these incorrect data are not processed, it will seriously affect the accuracy of subsequent data analysis. If agricultural production measures are adjusted based on incorrect environmental monitoring data, it may affect the growth of crops and cause economic losses.

[0004] Traditional data analysis methods are difficult to meet the requirements. The agricultural product industry faces a complex and ever-changing market environment and production conditions, and it is necessary to deeply explore the laws behind the data for accurate market prediction, risk warning, and decision-making support. However, existing analysis methods are mostly based on single data types or simple statistical models, unable to fully integrate the characteristics of multi-dimensional data, and difficult to discover the deep spatio-temporal correlations and semantic dependencies between data. When analyzing market condition data, only price fluctuations are considered, while ignoring the correlations with production processes, product attributes, and logistics factors, making it impossible to accurately predict price trends and market supply and demand changes.

[0005] The service interface lacks flexibility and configurability. Different user groups, such as agricultural product producers, distributors, consumers, etc., have different requirements for data services. However, the current interfaces of agricultural product data service systems are often fixed and cannot be customized and configured according to the specific needs of users, resulting in poor pertinence of data services and the inability to fully utilize the value of data. Agricultural product producers need to understand risk warning information during the production process and suggestions for optimizing production decisions, while the existing service interfaces may not be able to provide these personalized data services.

[0006] To solve the above problems, the present invention proposes an agricultural product big data service system based on a trusted data space, aiming to effectively integrate, clean, and analyze multi-dimensional data of the entire agricultural product chain, construct a data association graph, generate accurate decision-making features and warning indicators, and provide flexible and configurable data service interfaces to promote the digital transformation and sustainable development of the agricultural product industry. Summary of the Invention

[0007] The purpose of the present invention is to provide an agricultural product big data service system based on a trusted data space to solve the problems raised in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solutions: An agricultural product big data service system based on a trusted data space, the system includes: Multi-source data integration module: used to obtain a multi-dimensional data set of the entire agricultural product chain; Data cleaning and orchestration module: cleans the multi-dimensional data set to generate a standardized data sequence, and dynamically orchestrates the sequence through data association rules; Data association graph construction module: based on the cleaned standardized data sequence, generates a cross-domain data association graph through a distributed graph calculation algorithm, and the graph includes the spatio-temporal association weights and semantic dependency relationships between entity nodes; Multi-dimensional analysis engine module: inputs the data association graph into a hybrid analysis model, extracts data evolution patterns through multi-modal feature fusion, and generates classification decision-making features and risk warning indicators; Service interface generation module: based on the classification decision-making features and risk warning indicators, constructs a configurable service interface set through a policy mapping algorithm, and outputs a data service solution adapted to different scenarios.

[0009] Preferably, the multi-dimensional data includes production base data, production process data, product attribute data, market condition data, and cold chain logistics data; among them, the production base data includes base certification information, scale, and output indicators, the production process data includes environmental monitoring parameters and operation records, and the product attribute data includes test results and nutritional components.

[0010] Preferably, the data cleaning and orchestration module includes: Perform entity normalization on the text description fields in the production base data to eliminate naming ambiguities; Adopt a sliding window mechanism to smooth and correct the time-series outliers in the production process data; Build quality confidence labels based on the detection results of product attribute data and associate them with the standardized data sequence; Dynamically adjust the data association weights through a rule engine to generate a hierarchical orchestration structure.

[0011] Preferably, the implementation of the distributed graph computing algorithm includes: Perform spatial grid coding on the geographical coordinates of the production base data and cold chain logistics data; Extract the equipment identifiers and operation timestamps in the production process data to construct an event chain; Calculate the dynamic association strength between entity nodes through an incremental graph embedding algorithm; Fuse the spatio-temporal coding, event chain, and dynamic association strength to generate a multi-dimensional data association map.

[0012] Preferably, the hybrid analysis model includes: Perform periodic decomposition on the price fluctuation sequence in the market condition data to extract trend features and residual components; Matricially splice the product detection data and nutritional component data to generate a quality feature vector; Fuse the trend features, residual components, and quality feature vector through a hierarchical attention mechanism to construct a multi-dimensional decision space; Divide the risk level threshold in the decision space to generate classification decision features and warning trigger conditions.

[0013] Preferably, the implementation of the hierarchical attention mechanism includes: aligning the trend features in the time dimension to generate a first attention weight matrix; extracting high-frequency mutation features in the residual components to construct a second attention weight matrix; mapping the quality feature vector into a spatial distribution heat map to generate a third attention weight matrix; outputting multi-modal decision features by weighted superposition and fusion of the three types of weight matrices.

[0014] Preferably, the construction method of the policy mapping algorithm includes: Define the interface function template and parameter constraint conditions according to the data service scenario; Semantically match the classification decision features with the function template to generate an initial interface configuration scheme; Evaluate the compatibility between the scheme and the risk warning indicators through a cost function and iteratively optimize the parameter combination; Generate a service interface set that supports dynamic expansion and bind the data access permission policy.

[0015] Preferably, the specific method for entity normalization processing includes: constructing a standard thesaurus and an alias mapping table for the names of agricultural product bases, using a fuzzy matching algorithm to identify synonymous entities in the text description, filtering out low-probability matching results through a confidence threshold, merging duplicate entity records, and attaching a unique identifier and a data traceability label to the normalized entity.

[0016] Preferably, the optimization method for the incremental graph embedding algorithm includes: Dividing the static and dynamic partitions of the graph structure according to the data update frequency; Generating benchmark vectors using a batch embedding learning algorithm in the static partition; Performing local gradient propagation optimization on the newly added nodes in the dynamic partition to update the association strength; Fusing the features of the static and dynamic partitions through residual connection.

[0017] Preferably, the construction method for the cost function includes: Defining a balance coefficient for the interface response delay and the data calculation complexity, and initializing the compatibility score of the parameter combination as the benchmark value; Calculating the influence gradient of parameter adjustment on the warning index through the backpropagation algorithm, and generating an optimal parameter configuration sequence according to the gradient direction and the balance coefficient.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: At the data processing level, the system obtains a multi-dimensional data set of the entire agricultural product chain through the multi-source data integration module, breaking the data island phenomenon. Data such as production base data, production process data, product attribute data, market condition data, and cold chain logistics data are gathered, providing a rich data basis for comprehensively analyzing the agricultural product industry situation. The data cleaning and arrangement module performs refined processing on these data, conducts entity normalization processing on the text description fields in the production base data to eliminate naming ambiguities, makes the data more standardized and unified in expression, and improves the usability of the data; adopts a sliding window mechanism to smooth and correct the time-series outliers in the production process data to ensure the accuracy of the data and provide a reliable basis for subsequent analysis; constructs a quality confidence label based on the detection results of the product attribute data and associates it with the standardized data sequence, making the logical relationship between the data closer and laying a foundation for deeply mining the data value.

[0019] In terms of data association analysis, the data association graph construction module generates a cross-domain data association graph based on the cleaned and standardized data sequences using a distributed graph computing algorithm. It performs spatial grid coding on the geographical coordinates of production base data and cold chain logistics data, extracts the device identifiers and operation timestamps in the production process data to construct an event chain, calculates the dynamic association strength between entity nodes through an incremental graph embedding algorithm, and then fuses them to generate a multi-dimensional data association graph. This graph contains the spatio-temporal association weights and semantic dependency relationships between entity nodes, can clearly present the complex connections between the data of each link in the entire agricultural product chain, provides a new perspective for in-depth understanding of the production, circulation, and sales processes of agricultural products, and assists enterprises in making more scientific decisions.

[0020] From the perspective of data analysis and early warning, the multi-dimensional analysis engine module inputs the data association graph into a hybrid analysis model, extracts data evolution patterns through multi-modal feature fusion, and generates classification decision features and risk early warning indicators. It performs periodic decomposition on the price fluctuation sequence in the market condition data, extracts trend features and residual components, combines product detection data and nutritional component data to generate a quality feature vector, and then constructs a multi-dimensional decision space by fusing various features through a hierarchical attention mechanism and divides the risk level threshold. This enables the system to accurately analyze market dynamics, timely detect potential risks, such as predicting large fluctuations in agricultural product prices and product quality problems in advance, helping enterprises take measures in advance to reduce business risks.

[0021] In terms of data service provision, the service interface generation module constructs a configurable service interface set based on the classification decision features and risk early warning indicators through a policy mapping algorithm, and outputs data service solutions adapted to different scenarios. It defines interface function templates and parameter constraint conditions according to the data service scenario, performs semantic matching between the classification decision features and the function templates to generate an initial interface configuration plan, evaluates the compatibility between the plan and the risk early warning indicators through a cost function, iteratively optimizes the parameter combination, generates a service interface set supporting dynamic expansion, and binds data access permission policies at the same time. This meets the diverse needs of different user groups. Agricultural product producers can obtain production optimization suggestions and risk early warnings, distributors can get market trend analysis and inventory management suggestions, and consumers can also obtain product quality traceability information, comprehensively improving the quality and value of data services, and promoting the digital development and management level improvement of the agricultural product industry. Brief Description of the Drawings

[0022] Figure 1 It is the working principle diagram of the agricultural product big data service system based on the trusted data space described in the present invention; Figure 2 It is the flow chart implemented by the distributed graph computing algorithm; Figure 3 It is the working flow chart of the hybrid analysis model; Figure 4 Flowchart implemented by the hierarchical attention mechanism. Specific implementation manners

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] Please refer to Figures 1-4 , the present invention provides an agricultural product big data service system based on a trusted data space, aiming to integrate multi-dimensional data of the entire agricultural product chain by using advanced data processing technologies and algorithms, and provide accurate and efficient data services for relevant fields. The following are the specific implementation manners of the present invention.

[0025] The multi-source data integration module is responsible for obtaining the multi-dimensional data set of the entire agricultural product chain. These data cover all aspects such as agricultural product production, processing, circulation, and sales, and are the basis for the subsequent analysis and services of the system. By establishing connections with various data sources, such as sensor devices in production bases, database of market trading platforms, information systems of logistics enterprises, etc., automatic collection and aggregation of multi-dimensional data are realized.

[0026] The data cleaning and orchestration module performs cleaning operations on the obtained multi-dimensional data set to generate a standardized data sequence. Since the original data may have problems such as noise, error values, duplicate data, etc., it needs to be processed through a series of data cleaning technologies. At the same time, this module also dynamically orchestrates the standardized data sequence according to data association rules for more efficient data analysis and utilization in the future.

[0027] The data association graph construction module generates a cross-domain data association graph based on the cleaned standardized data sequence by using a distributed graph computing algorithm. This graph not only includes the spatio-temporal association weights between entity nodes, but also reflects the semantic dependency relationship, providing a powerful tool for deeply exploring the potential connections between data.

[0028] The multi-dimensional analysis engine module inputs the data association graph into a hybrid analysis model, and extracts data evolution patterns through multi-modal feature fusion technology. Based on these patterns, classification decision features and risk warning indicators are generated to provide valuable decision support information for users.

[0029] The service interface generation module constructs a configurable service interface set based on the classification decision features and risk warning indicators by using a policy mapping algorithm. These interfaces can output data service solutions adapted to different scenarios to meet the diverse needs of users.

[0030] Next, the specific implementation details of the present invention will be described in detail through 6 embodiments.

[0031] Embodiment 1:

[0032] During the operation of the agricultural product big data service system, the specific working process of the multi-source data integration module is as follows. For the production base data, the system obtains data such as base certification information, scale, and output indicators by docking with the information management systems of each production base. For example, by establishing a data transmission channel with the internal system of a large vegetable production base, data such as the organic certification information, planting area, and annual output of different vegetable varieties of the base can be obtained in real time. For the production process data, various sensors installed in the production base, such as temperature and humidity sensors, soil pH sensors, etc., are used to collect environmental monitoring parameters; at the same time, the operation records of the operators, including sowing time, types and dosages of fertilizers, irrigation time, etc., are recorded with the help of production management software. In terms of obtaining product attribute data, cooperation with professional agricultural product testing institutions is carried out to obtain the test results of the products, such as pesticide residue test data, heavy metal content test data, etc.; and the nutritional component data of the agricultural products is obtained from the agricultural product nutritional component research database. For the market condition data, web crawler technology is used to collect price information, trading volume data, etc. of agricultural products from major agricultural product trading platforms and industry information websites. For the cold chain logistics data, docking with the logistics management system of the logistics enterprise is carried out to obtain information such as temperature, humidity data, transportation trajectory, and distribution time of agricultural products during transportation and storage.

[0033] The data cleaning and orchestration module cleans and orchestrates the acquired multi-dimensional data. Taking the production base data as an example, in the processing of the text description field, a standard thesaurus and an alias mapping table for the names of agricultural product bases are constructed, and a fuzzy matching algorithm is used to identify synonymous entities in the text description. For example, when there are two names, "XX Organic Farm" and "XX Ecological Agriculture Base", through the standard thesaurus, alias mapping table and fuzzy matching algorithm, it is identified that they may refer to the same entity. By setting a confidence threshold to filter out low-probability matching results, duplicate entity records are merged, and a unique identifier and a data traceability label are attached to the normalized entity to facilitate subsequent data query and management. For production process data, a sliding window mechanism is used to smooth and correct time series outliers. Suppose there are abnormal fluctuations in the temperature monitoring data during the production process. Using the sliding window mechanism, the data within a certain time range is selected for mean calculation, and the calculated mean is used to replace the outlier to make the data more stable and reliable. A quality confidence label is constructed based on the detection results of product attribute data and associated with the standardized data sequence. If the pesticide residue detection result of a certain agricultural product is lower than the national standard, a higher quality confidence label can be assigned; otherwise, a lower label is assigned and this label is associated with other relevant data. The data association weight is dynamically adjusted through a rule engine to generate a hierarchical orchestration structure. According to the association tightness between different data and business requirements, corresponding rules are set to adjust the data association weight so that important data associations are in a more prominent position in the hierarchical structure.

[0034] Example 2:

[0035] When the data association graph construction module generates a cross-domain data association graph, it performs spatial grid coding on the geographical coordinates of the production base data and cold chain logistics data. For example, the longitude and latitude coordinates of the production base are divided according to a certain grid accuracy, and each grid is assigned a unique code. Suppose a grid unit is 100 meters × 100 meters, and the coordinates of a production base are located within a specific grid, and the code of this grid is "010101", then the geographical coordinates of this production base are encoded as this grid number. In this way, the geographical coordinates are converted into digital codes that are convenient for processing, providing convenience for subsequent analysis.

[0036] Extract the equipment identifier and operation timestamp in the production process data to construct an event chain. When a certain irrigation equipment starts to work, record its equipment identifier and the operation start timestamp; when the equipment stops working, record the stop timestamp. Connect these equipment identifiers and timestamps in the order of operation to form an event chain. For example, equipment A starts to work at time t1 and stops working at time t2, and equipment B starts to work at time t3... In this way, an event chain reflecting the equipment operation sequence and time relationship during the production process is constructed.

[0037] Calculate the dynamic association strength between entity nodes through an incremental graph embedding algorithm. Divide the static and dynamic partitions of the graph structure according to the data update frequency. For parts with a low data update frequency, such as the basic information of production bases (land area, geographical location, etc.), they are divided into static partitions; for parts with a high data update frequency, such as real-time monitoring data during the production process, they are divided into dynamic partitions. In the static partition, use a batch embedding learning algorithm to generate benchmark vectors. By learning and analyzing a large amount of data in the static partition, generate benchmark vectors representing the characteristics of each entity node. Optimize the local gradient propagation for the new nodes in the dynamic partition to update the association strength. When new production process data or logistics data is added, for these new nodes, use the local gradient propagation algorithm to optimize and adjust their association strength with other nodes. Fuse the features of the static and dynamic partitions through residual connection, and fuse the benchmark vectors of the static partition and the optimized association strength of the dynamic partition to obtain a more accurate dynamic association strength between entity nodes.

[0038] Fuse spatio-temporal coding, event chains, and dynamic association strength to generate a multi-dimensional data association map. Integrate the geospatial coordinate information after spatial grid coding, the constructed event chains, and the calculated dynamic association strength to construct a multi-dimensional data association map containing the spatio-temporal association weights and semantic dependency relationships between entity nodes. For example, a node in the map represents a certain production base. Through spatio-temporal coding, its geographical location information can be reflected. The event chain reflects the production operation process of the base, and the dynamic association strength shows the degree of closeness of the association between the base and other entities (such as logistics nodes, market nodes, etc.).

[0039] Example 3:

[0040] The hybrid analysis model of the multi-dimensional analysis engine module performs periodic decomposition on the price fluctuation sequence in the market condition data. Assume that the price fluctuation sequence in the market condition data is , and adopt a periodic decomposition algorithm to decompose it into a trend feature and a residual component , that is, . Among them, t represents time, represents the agricultural product price at time t, represents the trend part of the price, reflecting the long-term change trend of the price, represents the remaining fluctuation part after removing the trend, including information such as short-term random fluctuations and outliers. Through the analysis of historical price data, adopt a suitable periodic decomposition algorithm, such as seasonal decomposition method (STL decomposition), to extract the trend feature and the residual component.

[0041] Matrixize and splice the product detection data and the nutritional component data to generate a quality feature vector. Arrange the various detection data of the product (such as pesticide residue amount, heavy metal content, etc.) and the nutritional component data (such as protein content, vitamin content, etc.) in a matrix form according to a certain order. Assume that there are m indicators for the detection data and n indicators for the nutritional component data, then construct a matrix, and this matrix is the quality feature vector. For example, for a certain kind of fruit, there are 3 items of its pesticide residue detection data and 5 items of its nutritional component data. Arrange these 8 items of data in sequence to form a vector , which serves as the quality feature vector of this fruit.

[0042] Fuse the trend feature, the residual component, and the quality feature vector through a hierarchical attention mechanism to construct a multi-dimensional decision space. Align the trend feature in the time dimension to generate the first attention weight matrix . Since the importance of the trend feature may be different at different time points, through time dimension alignment, according to the characteristics of the data at each time point, calculate the corresponding weight for each time point and form the first attention weight matrix . Extract the high-frequency mutation features in the residual component to construct the second attention weight matrix . The high-frequency mutation features in the residual component often reflect the abnormal fluctuations in the market or the sudden changes in product quality. Extract these features through a specific algorithm and construct the second attention weight matrix according to their importance . Map the quality feature vector into a spatial distribution heat map to generate the third attention weight matrix . According to the numerical values of the elements in the quality feature vector, perform a visual mapping in space to form a heat map, and then generate the third attention weight matrix according to the heat distribution in different regions of the heat map . Output the multi-modal decision features through weighted superposition and fusion of the three types of weight matrices, that is , where M represents the multi-modal decision feature and Q represents the quality feature vector.

[0043] Divide the risk level thresholds in the decision space to generate classification decision features and early warning trigger conditions. Set different risk level thresholds according to the historical data of the agricultural product market and industry experience. For example, divide the price fluctuation risk into three levels: low, medium, and high. When the price fluctuation exceeds a certain threshold, trigger the corresponding early warning. At the same time, combine the quality feature vector and other relevant data to generate classification decision features, such as judging whether a batch of agricultural products is suitable to enter a certain market, or whether special marketing strategies need to be adopted, etc.

[0044] Example 4:

[0045] When constructing the policy mapping algorithm, the service interface generation module defines the interface function template and parameter constraint conditions according to the data service scenario. For example, for the decision-making support scenario of agricultural product production enterprises, an interface function template is defined, which provides functions such as agricultural product yield prediction and market demand analysis. At the same time, parameter constraint conditions are set, such as the time range parameter for yield prediction (predicting the yield in the next 1 month, 3 months, or 6 months), the regional parameter for market demand analysis (specifying a certain region or the national scope), etc.

[0046] Semantically match the classification decision features with the function template to generate an initial interface configuration plan. When the system generates classification decision features, such as predicting that the market demand for a certain agricultural product in a certain region will increase significantly in the next 3 months, semantically match this decision feature with the previously defined interface function template. According to the matching result, select the appropriate function module and set the corresponding parameters to generate an initial interface configuration plan. Suppose there is a "market demand prediction" function module in the interface function template, and set the prediction time and regional parameters to values that match the decision features to obtain the initial interface configuration.

[0047] Evaluate the compatibility between the solution and the risk warning indicators through a cost function, and iteratively optimize the parameter combination. Define the balance coefficient between the interface response delay and the data calculation complexity, and initialize the compatibility score of the parameter combination to a reference value. Suppose the balance coefficient is , the interface response delay is D, the data calculation complexity is C, and the initial value of the compatibility score is . Calculate the influence gradient of parameter adjustment on the warning indicators through the backpropagation algorithm, and generate the optimal parameter configuration sequence according to the gradient direction and the balance coefficient. When adjusting a certain parameter (such as the time range parameter for prediction), use the backpropagation algorithm to calculate the impact degree of this parameter adjustment on the risk warning indicators (such as the accuracy and timeliness of prediction), and obtain the influence gradient. According to this gradient direction, combined with the balance coefficient , adjust the parameters to continuously optimize the compatibility score, and finally obtain the optimal parameter configuration sequence.

[0048] Generate a service interface set that supports dynamic expansion and bind the data access permission policy. According to the optimized parameter configuration sequence, generate specific service interfaces. These interfaces can be dynamically expanded according to the user's needs. For example, when the user proposes new functional requirements, new function modules can be easily added to the interface. At the same time, in order to ensure data security and privacy, bind the data access permission policy. For different user roles (such as production enterprise users, dealer users, regulatory department users, etc.), set different data access permissions, and only users with corresponding permissions can access specific data and use the corresponding interface functions.

[0049] Example 5:

[0050] In the data cleaning and orchestration module, the specific process of entity normalization is further elaborated. Building a standard thesaurus and alias mapping table for agricultural product base names is the basis of entity normalization. By collecting a large number of agricultural product base names, a database containing standard names and common aliases is established. For example, "XX Green Vegetable Base" is the standard name, and "XX Green Farm", "XX Vegetable Plantation", etc. are its aliases. When actually processing text description fields, a fuzzy matching algorithm is used to identify synonymous entities in the text description. Suppose the input text contains "XX Ecological Farm", the fuzzy matching algorithm will search for similar names in the standard thesaurus and alias mapping table. By calculating the text similarity (such as using the edit distance algorithm), when the similarity exceeds the set threshold, it is considered that a synonymous entity has been found.

[0051] Filter low-probability matching results through a confidence threshold. Since fuzzy matching may produce some inaccurate matching results, a confidence threshold is set. For example, when the confidence corresponding to the calculated result of the matching similarity is lower than 0.8, it is considered a low-probability matching result and is filtered out without subsequent processing. Merge duplicate entity records. When multiple text descriptions are identified as pointing to the same entity, these records are merged into one record. For example, if it is found that "XX Organic Farm 1" and "XX Organic Farm 2" are actually different records of the same farm, they are merged into one record and the relevant data information is updated. Attach a unique identifier and data traceability label to the normalized entity. Assign a globally unique identifier to each normalized entity for accurate identification and management in the system. At the same time, add a data traceability label to record the source of the entity data, such as information on the data collection time, the system where the data was collected, etc., for subsequent tracing of the accuracy and reliability of the data.

[0052] Example 6:

[0053] In the data association graph construction module, the optimization process of the incremental graph embedding algorithm is introduced in depth. Dividing the static and dynamic partitions of the graph structure according to the data update frequency is the key step in the optimization. By statistically analyzing the update frequencies of various types of data in the system, the division criteria are determined. For example, for the basic information of production bases (such as the geographical location and land area of the bases), this data usually does not change for a long time and is divided into the static partition; while for the real-time monitoring data during the production process (such as temperature and humidity data, equipment operation status data, etc.) and the real-time price data of the market conditions, their update frequencies are high and they are divided into the dynamic partition.

[0054] Generate benchmark vectors using the batch embedding learning algorithm in the static partition. The batch embedding learning algorithm extracts data features by centrally learning and analyzing a large amount of data in the static partition, and generates benchmark vectors representing the features of each entity node. For example, for a production base node, by learning data such as its related land area, planted varieties, and historical yields, a benchmark vector representing the characteristics of the production base is generated. This benchmark vector reflects the basic attributes and characteristics of the production base, providing a basis for subsequent calculation of dynamic association strength.

[0055] Optimize the local gradient propagation for newly added nodes in the dynamic partition to update the association strength. When new data is added to the dynamic partition, such as newly added production process monitoring data at a certain time point, for these newly added nodes, the local gradient propagation algorithm is used for optimization. The local gradient propagation algorithm calculates the influence gradient on the association strength according to the association relationship between the newly added nodes and the surrounding nodes, and adjusts the association strength according to this gradient. For example, when a new temperature and humidity monitoring data is added, calculate the influence of this data on the association strength of surrounding production base nodes, equipment nodes, etc. through the local gradient propagation algorithm, and then update the values of these association strengths.

[0056] Fuse the features of the static and dynamic partitions through residual connections. Residual connection is an effective fusion method that fuses the benchmark vectors generated by the static partition and the optimized association strength in the dynamic partition. Through residual connection, the basic features of the static partition are retained, and at the same time, the latest change information in the dynamic partition can be incorporated in a timely manner. For example, perform a residual connection operation on the benchmark vector of the production base node in the static partition and the association strength updated according to the latest monitoring data in the dynamic partition to obtain a more accurate node feature representation that can better reflect the actual situation, thereby improving the accuracy and real-time performance of the data association map.

[0057] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device.

[0058] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An agricultural product big data service system based on a trusted data space, characterized in that, Including: Multi-source data integration module: used to obtain the multi-dimensional data set of the entire agricultural product chain; Data cleaning and orchestration module: clean the multi-dimensional data set to generate a standardized data sequence, and dynamically orchestrate the sequence through data association rules; Data association graph construction module: based on the cleaned standardized data sequence, generate a cross-domain data association graph through a distributed graph computing algorithm, and the graph includes the spatio-temporal association weights and semantic dependency relationships between entity nodes; Multi-dimensional analysis engine module: input the data association graph into a hybrid analysis model, extract the data evolution pattern through multi-modal feature fusion, and generate classification decision features and risk warning indicators; Service interface generation module: based on the classification decision features and risk warning indicators, construct a configurable service interface set through a policy mapping algorithm, and output a data service solution adapted to different scenarios.

2. The agricultural product big data service system based on a trusted data space according to claim 1, wherein The multi-dimensional data includes production base data, production process data, product attribute data, market condition data, and cold chain logistics data; among them, the production base data includes base certification information, scale, and output indicators, the production process data includes environmental monitoring parameters and operation records, and the product attribute data includes test results and nutritional components.

3. An agricultural product big data service system based on a trusted data space according to claim 1, characterized in that, The data cleaning and orchestration module includes: Perform entity normalization processing on the text description fields in the production base data to eliminate naming ambiguities; Adopt a sliding window mechanism to smooth and correct the time series outliers in the production process data; Construct a quality confidence label based on the test results of the product attribute data and associate it with the standardized data sequence; Dynamically adjust the data association weights through a rule engine to generate a hierarchical orchestration structure.

4. The agricultural product big data service system based on a trusted data space according to claim 1, characterized in that, The implementation of the distributed graph computing algorithm includes: Perform spatial grid coding on the geographical coordinates of the production base data and cold chain logistics data; Extract the equipment identifier and operation timestamp in the production process data to construct an event chain; Calculate the dynamic association strength between entity nodes through an incremental graph embedding algorithm; Fuse spatio-temporal coding, event chain, and dynamic association strength to generate a multi-dimensional data association graph.

5. The agricultural product big data service system based on a trusted data space according to claim 1, characterized in that, The hybrid analysis model includes: Perform periodic decomposition on the price fluctuation sequence in the market condition data to extract trend features and residual components; Matrix splice the product test data and nutritional component data to generate a quality feature vector; Fuse trend features, residual components, and quality feature vectors through a hierarchical attention mechanism to construct a multi-dimensional decision space; Divide the risk level threshold in the decision space to generate classification decision features and warning trigger conditions.

6. The agricultural product big data service system based on a trusted data space according to claim 5, characterized in that, The implementation of the hierarchical attention mechanism includes: align the trend features in the time dimension to generate a first attention weight matrix; extract high-frequency mutation features in the residual components to construct a second attention weight matrix; map the quality feature vector into a spatial distribution heat map to generate a third attention weight matrix; output multi-modal decision features by weighted superposition and fusion of the three types of weight matrices.

7. An agricultural product big data service system based on a trusted data space according to claim 1, characterized in that, The construction method of the policy mapping algorithm includes: Define the interface function template and parameter constraint conditions according to the data service scenario; Semantically match the classification decision features with the function template to generate an initial interface configuration plan; Evaluate the compatibility between the cost function evaluation scheme and the risk warning indicators, and iteratively optimize the parameter combination; Generate a service interface set that supports dynamic expansion and bind data access permission policies.

8. The agricultural product big data service system based on a trusted data space according to claim 3, characterized in that, The specific method of the entity normalization process includes: constructing a standard thesaurus and an alias mapping table for the names of agricultural product bases, using a fuzzy matching algorithm to identify synonymous entities in the text description, filtering low-probability matching results through a confidence threshold, merging duplicate entity records, and attaching a unique identifier and a data traceability label to the normalized entity.

9. The agricultural product big data service system based on a trusted data space according to claim 4, wherein The optimization method of the incremental graph embedding algorithm includes: Dividing the static and dynamic partitions of the graph structure according to the data update frequency; Using a batch embedding learning algorithm to generate benchmark vectors in the static partition; Performing local gradient propagation optimization on the newly added nodes in the dynamic partition to update the association strength; Fusing the static and dynamic partition features through residual connection.

10. The agricultural product big data service system based on a trusted data space according to claim 6, wherein, The construction method of the cost function includes: Defining a balance coefficient between the interface response delay and the data calculation complexity, and initializing the compatibility score of the parameter combination to a benchmark value; Calculating the influence gradient of the parameter adjustment on the warning indicators through the backpropagation algorithm, and generating an optimal parameter configuration sequence according to the gradient direction and the balance coefficient.

Images