Rural e-commerce service method and system based on artificial intelligence

Abstract

The invention discloses a rural e-commerce service method and system based on artificial intelligence, and the method comprises the steps: converting agricultural production data, market transaction data and consumer behavior data into a unified format through a standardized protocol, and obtaining an integrated agricultural data set through the storage of a distributed database; according to the integrated agricultural data set, adopting a federal learning algorithm to train a regional supply and demand prediction model, and if a data source of the integrated agricultural data set contains farmer production data, performing encryption processing on the data through a differential privacy technology to obtain a privacy protection prediction model; and obtaining a supply and demand prediction result from the privacy protection prediction model, and calculating the price trend and logistics demand of the agricultural products by adopting a time sequence analysis algorithm in combination with real-time market transaction data to obtain a real-time supply and demand matching scheme. The agricultural product supply and demand matching precision and the logistics distribution efficiency are improved, the data safety and the transaction credibility are guaranteed, and intelligent decision support is provided for each link of the agricultural industry chain.

Description

Technical Field

[0001] The present invention relates to the technical field of rural e-commerce services, and in particular discloses an artificial intelligence-based rural e-commerce service method and system. Background Art

[0002] Rural e-commerce, a key area in promoting agricultural modernization, plays a vital role in improving the efficiency of agricultural product distribution and increasing farmers' incomes. With the widespread adoption of digital technology, rural e-commerce is gradually transforming from traditional trading platforms to intelligent service systems. However, existing solutions still suffer from significant shortcomings in practical application. Many platforms remain limited to information dissemination and simple transactions, lacking deep integration with agricultural production, supply chain management, and market demand. This leads to inefficient supply and demand matching, high logistics losses, and difficulties in ensuring farmers' income. Furthermore, data silos are widespread, and information asymmetry between production and consumption further exacerbates resource waste and market volatility.

[0003] Against this backdrop, the intelligent development of rural e-commerce faces multiple technical challenges. Primarily, the integration of heterogeneous data from multiple sources is a major challenge. Agricultural production involves real-time environmental data, market transaction data, and consumer behavior data collected by IoT sensors. These data come in various formats and from dispersed sources, making unified analysis difficult for accurate decision-making. This lack of data integration directly leads to another core issue: the limited real-time and accuracy of intelligent decision-making. For example, the development of planting plans requires integrating historical meteorological data and market price trends. However, existing systems often struggle to provide timely and reliable recommendations due to data processing delays or insufficient model prediction accuracy. Furthermore, the conflict between data privacy and data sharing also hinders system effectiveness. Farmers, fearing data leaks, are reluctant to share production information. This, coupled with a lack of sufficient data support, limits model training effectiveness, making it difficult to establish a regional smart agriculture ecosystem.

[0004] Therefore, how to efficiently integrate multi-source heterogeneous data and achieve low-latency, high-precision intelligent decision-making while protecting farmers' data privacy has become a key issue in the intelligentization of rural e-commerce service systems. Summary of the Invention

[0005] The present invention provides an artificial intelligence-based rural e-commerce service method and system, aiming to solve at least one of the above-mentioned defects of the prior art.

[0006] One aspect of the present invention relates to an artificial intelligence-based rural e-commerce service method, comprising the following steps: Acquire multi-source heterogeneous data from IoT sensors, trading platforms, and consumer feedback systems, convert agricultural production data, market transaction data, and consumer behavior data into a unified format using standardized protocols, and store the integrated agricultural data set in a distributed database. Based on the integrated agricultural data set, a federated learning algorithm is used to train a regional supply and demand forecasting model. If the data source of the integrated agricultural data set includes farmer production data, the data is encrypted using differential privacy technology to obtain a privacy-protected forecasting model. Obtain supply and demand forecast results from a privacy-preserving forecasting model, combine them with real-time market transaction data, and use time series analysis algorithms to calculate agricultural product price trends and logistics demand to obtain a real-time supply and demand matching solution; Based on the real-time supply and demand matching solution, a graph neural network algorithm is used to optimize the supply chain management path. If the logistics loss rate exceeds the preset threshold, the transportation node weight is adjusted to obtain an optimized logistics scheduling solution; Obtain scheduling instructions from the optimized logistics scheduling plan, record transaction data between the production and consumption ends on the blockchain platform through smart contract technology, judge the integrity of transaction data, and obtain traceable supply and demand transaction records.

[0007] Furthermore, the steps of obtaining multi-source heterogeneous data from IoT sensors, trading platforms, and consumer feedback systems, converting agricultural production data, market transaction data, and consumer behavior data into a unified format using standardized protocols, and storing the integrated agricultural data set in a distributed database include: Use standardized protocols to convert the agricultural production data, market transaction data, and consumer behavior data obtained from IoT sensors, trading platforms, and consumer feedback systems to obtain a unified data set. The data sets in a unified format are stored in a distributed database, and the IoT data, transaction data and consumer data are partitioned and managed using distributed storage technology to obtain an integrated agricultural data set.

[0008] Furthermore, based on the integrated agricultural data set, a federated learning algorithm is used to train a regional supply and demand forecasting model. If the data source of the integrated agricultural data set includes farmer production data, the data is encrypted using differential privacy technology to obtain a privacy-protected forecasting model. The steps include: Obtain a unified format data set from agricultural data integration, use distributed computing technology to partition multi-source data, and divide it into farmer production data partition, market transaction data partition, and consumer behavior data partition according to the data source, to obtain a partitioned agricultural data set; If the partitioned agricultural data set contains farmer production data, the farmer production data partition is encrypted using differential privacy technology, and noise is added using the preset differential privacy parameters to obtain the encrypted farmer production data partition; Based on the encrypted farmer production data partitions and other data partitions, a federated learning algorithm is used to train in a distributed computing environment to generate model parameters for the supply and demand forecast requirements of each region, and obtain an initial regionalized supply and demand forecast model; The initial regionalized supply and demand forecasting model is used to perform forecast calculations on the partitioned agricultural data set, obtain the supply and demand forecast results for each region, and adjust the model parameters to obtain a privacy-preserving forecasting model.

[0009] Furthermore, the steps of obtaining supply and demand forecast results from the privacy-preserving forecast model, combining them with real-time market transaction data, and using a time series analysis algorithm to calculate agricultural product price trends and logistics demand to obtain a real-time supply and demand matching solution include: Obtain supply and demand forecast results from a privacy-preserving forecast model, combine it with real-time market transaction data, and use distributed computing technology to partition multi-source data. Generate partitioned data sets based on regional divisions to obtain regionalized supply and demand forecast data. Based on regional supply and demand forecast data, time series analysis algorithms are used to process real-time market transaction data, calculate price trends for agricultural products in each region, and obtain regional price trend data; If the regionalized price trend data meets the preset threshold, the regionalized supply and demand forecast data is combined with a linear regression algorithm to calculate the logistics demand of each region to obtain regionalized logistics demand data; By integrating regionalized logistics demand data and regionalized price trend data using distributed computing technology, we can generate real-time supply and demand matching solutions for each region.

[0010] Furthermore, based on the real-time supply and demand matching solution, a graph neural network algorithm is used to optimize the supply chain management path. If the logistics loss rate exceeds a preset threshold, the transportation node weight is adjusted. The steps to obtain the optimized logistics scheduling solution include: Based on the real-time supply and demand matching solution, the graph neural network algorithm is used to calculate the logistics path in supply chain management, generate an initial path set for the transportation nodes in each region, and obtain regionalized logistics path data; Using regionalized logistics path data, statistical analysis tools are used to calculate the logistics loss rate of each path. If the logistics loss rate exceeds a preset threshold, the connection strength of the transportation node is redistributed to obtain the adjusted node weight data; Based on the adjusted node weight data, graph calculation tools are used to re-plan logistics routes, generate optimized route sets based on the distribution needs of each region, and determine regional optimized route data; By optimizing regionalized route data, combining inventory allocation and regional division, and using scheduling algorithms to integrate the impact of distribution efficiency and price fluctuations, logistics scheduling plans for each region are generated.

[0011] Furthermore, the steps of obtaining scheduling instructions from the optimized logistics scheduling plan, recording transaction data between the production and consumption ends on the blockchain platform through smart contract technology, determining the integrity of the transaction data, and obtaining traceable supply and demand transaction records include: Obtain dispatch instructions from the optimized logistics dispatch plan, use data analysis tools to extract the delivery routes and time windows for each region, combine the supply and demand matching data from the production and consumption ends, generate an instruction set containing the route, time, and quantity, and obtain structured dispatch instruction data; Based on structured dispatch instruction data, a smart contract generation tool is used to generate a transaction contract on the blockchain platform. Based on the volume and price fluctuation data of the production and consumption ends, a contract record containing transaction terms is generated to determine the executable transaction contract data. Using executable transaction contract data, a data verification tool is used to verify the timestamp, quantity, and price fields of the transaction record. If the timestamp, quantity, and price fields of the transaction record are consistent with the dispatch instruction data, the transaction data integrity is confirmed and the verified transaction record data is obtained; Based on the verified transaction record data, blockchain query tools are used to generate traceable records including the production end, consumption end and distribution path. For the supply and demand matching data of each region, transaction records containing timestamps and contract numbers are generated to determine the traceable supply and demand transaction data set.

[0012] Another aspect of the present invention relates to an artificial intelligence-based rural e-commerce service system, which is used to implement the above-mentioned artificial intelligence-based rural e-commerce service method. The artificial intelligence-based rural e-commerce service system includes: The first acquisition module is used to obtain multi-source heterogeneous data from IoT sensors, trading platforms, and consumer feedback systems. It uses standardized protocols to convert agricultural production data, market transaction data, and consumer behavior data into a unified format and stores the integrated agricultural data set in a distributed database. The second acquisition module is used to train a regional supply and demand forecasting model based on the integrated agricultural data set using a federated learning algorithm. If the data source of the integrated agricultural data set includes farmer production data, the data is encrypted using differential privacy technology to obtain a privacy-preserving forecasting model. The third acquisition module is used to obtain supply and demand forecast results from the privacy-preserving prediction model. Combined with real-time market transaction data, it uses a time series analysis algorithm to calculate agricultural product price trends and logistics demand to obtain a real-time supply and demand matching solution. The fourth acquisition module is used to optimize the supply chain management path based on the real-time supply and demand matching plan using a graph neural network algorithm. If the logistics loss rate exceeds a preset threshold, the transportation node weight is adjusted to obtain an optimized logistics scheduling plan; The fifth acquisition module is used to obtain scheduling instructions from the optimized logistics scheduling plan, record the transaction data between the production and consumption ends on the blockchain platform through smart contract technology, judge the integrity of the transaction data, and obtain traceable supply and demand transaction records.

[0013] Furthermore, the first acquisition module includes: The first acquisition unit is used to convert the agricultural production data, market transaction data, and consumer behavior data obtained from IoT sensors, trading platforms, and consumer feedback systems using a standardized protocol to obtain a data set in a unified format; The second acquisition unit is used to store data sets in a unified format through a distributed database, and use distributed storage technology to partition and manage IoT data, transaction data, and consumer data to obtain an integrated agricultural data set.

[0014] Furthermore, the second acquisition module includes: The third acquisition unit is used to obtain a unified format data set from the agricultural data integration, and uses distributed computing technology to partition the multi-source data. According to the data source, it is divided into farmer production data partition, market transaction data partition and consumer behavior data partition to obtain the partitioned agricultural data set; a fourth acquisition unit, configured to, if the partitioned agricultural data set includes farmer production data, encrypt the farmer production data partition using differential privacy technology, add noise using a preset differential privacy parameter, and obtain the encrypted farmer production data partition; The fifth acquisition unit is used to use the federated learning algorithm to train in a distributed computing environment based on the encrypted farmer production data partitions and other data partitions, generate model parameters for the supply and demand forecast requirements of each region, and obtain an initial regionalized supply and demand forecast model; The sixth acquisition unit is used to perform prediction calculations on the partitioned agricultural data set through the initial regionalized supply and demand prediction model, obtain the supply and demand prediction results of each region and adjust the model parameters to obtain a privacy-protected prediction model.

[0015] Furthermore, the second acquisition module includes: The seventh acquisition unit is used to obtain supply and demand forecast results from the privacy-preserving forecast model, combine them with real-time market transaction data, use distributed computing technology to partition multi-source data, generate partitioned data sets based on regional divisions, and obtain regionalized supply and demand forecast data; an eighth acquisition unit, configured to process real-time market transaction data using a time series analysis algorithm based on regionalized supply and demand forecast data, calculate price trends for agricultural products in each region, and obtain regionalized price trend data; a ninth acquisition unit configured to calculate the logistics demand of each region using a linear regression algorithm in combination with the regionalized supply and demand forecast data if the regionalized price trend data meets a preset threshold, thereby obtaining regionalized logistics demand data; The tenth acquisition unit is used to integrate the regionalized logistics demand data and regionalized price trend data using distributed computing technology to generate a real-time supply and demand matching solution for each region.

[0016] The beneficial effects achieved by the present invention are: The present invention provides an artificial intelligence-based rural e-commerce service method and system. By integrating multi-source heterogeneous data, federated learning and differential privacy technologies are used to build a privacy-protected regional supply and demand forecasting model. Based on the forecast results and real-time market data, time series analysis is used to calculate price trends and logistics demand, and a real-time supply and demand matching solution is generated. Furthermore, the present invention applies graph neural networks to optimize supply chain management paths and dynamically adjusts the weights of transportation nodes to reduce logistics losses. Finally, transaction data is recorded through blockchain smart contracts to achieve full traceability. The present invention effectively solves problems such as information asymmetry between agricultural production and market demand, low supply chain efficiency, etc., improves the accuracy of agricultural product supply and demand matching and logistics distribution efficiency, ensures data security and transaction credibility, and provides intelligent decision-making support for all links in the agricultural industry chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of an embodiment of an artificial intelligence-based rural e-commerce service method of the present invention. DETAILED DESCRIPTION

[0018] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0019] like Figure 1 As shown, the first embodiment of the present invention proposes an artificial intelligence-based rural e-commerce service method, comprising the following steps: Step S100: Acquire multi-source heterogeneous data from IoT sensors, trading platforms, and consumer feedback systems, convert agricultural production data, market transaction data, and consumer behavior data into a unified format using standardized protocols, and obtain an integrated agricultural data set through distributed database storage.

[0020] The consumer feedback system is an intelligent management platform that collects and analyzes user feedback data to optimize product and service experience and improve customer satisfaction. Its design must integrate multi-source data integration, real-time processing, and closed-loop management mechanisms.

[0021] Multi-source heterogeneous data refers to complex data forms composed of data sets with different sources, structures and forms.

[0022] Standardization agreements are regulatory documents developed through consultation by recognized organizations to unify technical specifications and ensure system compatibility and security. Their core goal is to eliminate technical barriers, improve resource efficiency, and establish a common operating benchmark within a specific field. These agreements, based on scientific and technical experience and reusable, serve as fundamental technical rules supporting industrial production, information technology, and social governance.

[0023] A distributed database (DDB) is a database system physically dispersed across multiple network nodes (such as servers and data centers) but logically managed in a unified manner. It uses distributed technologies and protocols to enable collaborative data storage, querying, and transaction processing, offering high performance, scalability, and fault tolerance. Its core principle is data distribution transparency, meaning users can complete operations without having to know the specific storage location of their data.

[0024] Agricultural data collection refers to multidimensional data sets systematically collected and integrated through various technical means during agricultural production, management, and research. These data cover areas such as crop growth, environmental parameters, farming practices, and market trends. They aim to support precision agriculture decision-making, optimize resource utilization, and improve industry efficiency. Their core value lies in empowering traditional agriculture through data-driven technologies (such as AI and the Internet of Things), enabling the transition from experience-based to scientific management.

[0025] Step S200: Based on the integrated agricultural data set, a federated learning algorithm is used to train a regional supply and demand forecasting model. If the data source of the integrated agricultural data set includes farmers' production data, the data is encrypted using differential privacy technology to obtain a privacy-protected forecasting model.

[0026] Federated Learning (FL) is a distributed machine learning framework that allows multiple participants (such as devices, organizations, or edge nodes) to collaboratively train a global model to share knowledge and optimize the model without sharing local data. Its core goal is to address the "data silo" problem and improve model generalization while meeting data privacy requirements (such as GDPR and HIPAA) and network communication efficiency constraints.

[0027] The Regional Supply and Demand Forecasting Model is a quantitative analysis tool that integrates regional economic characteristics, resource distribution, and market demand dynamics. It is used to predict the changing trends in the supply and demand relationship between goods, services, or resources within a specific geographic area. By integrating multi-dimensional data on the region's economy, policies, and population, the Regional Supply and Demand Forecasting Model constructs a dynamic equilibrium framework to support industrial chain optimization, resource allocation, and policy formulation.

[0028] Farmer production data refers to the quantitative records and statistical information generated by farmers, with families as the basic unit, throughout the entire agricultural production process, including planting, breeding, and processing. It covers dimensions such as resource input, output benefits, technological application, and environmental impact, and is a key basis for evaluating agricultural production efficiency, rationality of resource allocation, and sustainable development capabilities.

[0029] Differential Privacy (DP) is a mathematically based data privacy protection technology. Its core goal is to inject controllable noise or perturbations into a dataset, making it nearly impossible to reverse engineer the analysis results to reveal individual user information. This approach achieves a strict balance between data utility and privacy protection. 13 This technology quantifies the risk of privacy leakage, ensuring that the presence or absence of any single individual in a dataset has minimal impact on the output.

[0030] Encryption is the technical process of converting raw data (plaintext) into unreadable ciphertext using a specific algorithm to ensure data confidentiality, integrity, and authentication. Its core goal is to ensure that information cannot be deciphered by unauthorized parties, even if intercepted, during transmission or storage. It is widely used in network security, communication protocols, and privacy protection.

[0031] Privacy-preserving prediction models refer to computing frameworks that use specific technical means to protect the privacy of sensitive user or system data when building and deploying prediction models. Their core goal is to achieve accurate predictions while ensuring the privacy of raw data, intermediate computational processes, and prediction results.

[0032] Step S300: Obtain supply and demand forecast results from the privacy-preserving forecast model, combine them with real-time market transaction data, and use a time series analysis algorithm to calculate agricultural product price trends and logistics demand to obtain a real-time supply and demand matching solution.

[0033] Supply and demand forecasts are quantitative assessments of the matching state of resource supply and demand in specific future scenarios, achieved by integrating historical data, dynamic parameters, and multi-dimensional models. Their core goal is to optimize resource allocation efficiency, balance supply and demand, and reduce systemic risk. They are widely used in fields such as energy, manufacturing, and fluid control.

[0034] Time series analysis algorithms are a collection of mathematical methods that employ statistical modeling and machine learning techniques to identify patterns, predict trends, or detect anomalies in chronologically ordered observational data. Their core goal is to extract trends, seasonality, cyclical patterns, and randomness from the data, and to build dynamic models to reveal patterns in data evolution and accurately predict future states.

[0035] Agricultural product price trends refer to the regular and directional fluctuations in agricultural product market prices over a specific timeframe, influenced by supply and demand, production costs, and the external environment. At its core, this analysis decomposes long-term trends, cyclical fluctuations, and seasonal variations to reveal the underlying patterns of price evolution and the expected future direction.

[0036] Logistics demand refers to the demand for logistics services such as transportation, warehousing, packaging, and handling, driven by socioeconomic activities, to efficiently allocate physical goods such as raw materials, semi-finished products, and finished products within a specific time and space. Essentially, logistics operations fulfill the spatial and temporal transfer needs of goods across production, distribution, and consumption, while also encompassing value-added services such as information processing.

[0037] A real-time supply-demand matching solution utilizes dynamic data collection and analysis to instantly adjust resource allocation strategies, achieving efficient and precise alignment of upstream and downstream supply and demand across time, space, and quantity. Its core goal is to minimize the dual losses of idle resources and unmet demand by driving decisions through algorithmic models based on real-time business conditions (such as inventory fluctuations, order surges, or changes in shipping capacity).

[0038] Step S400: Based on the real-time supply and demand matching plan, the graph neural network algorithm is used to optimize the supply chain management path. If the logistics loss rate exceeds the preset threshold, the transportation node weight is adjusted to obtain an optimized logistics scheduling plan.

[0039] A Graph Neural Network (GNN) is a deep learning model designed for graph-structured data. By recursively aggregating node neighborhood information and updating node representations, it enables joint modeling and analysis of topological relationships, node attributes, and edge features in graph data. Its core approach is to leverage message passing to capture the complex interconnectedness of non-Euclidean spaces, such as social networks and molecular structures.

[0040] The supply chain management approach involves designing and optimizing processes across the entire supply chain, from raw material procurement to end-consumer consumption, based on the business relationships and resource distribution of each participant in the supply chain network (suppliers, manufacturers, distributors, etc.). This approach aims to achieve efficient resource allocation and minimize operating costs by coordinating information, logistics, and capital flows. Its core approach is to achieve a balance between quality, efficiency, and customer satisfaction through dynamic adjustments to collaborative mechanisms across planning, procurement, production, warehousing, and distribution.

[0041] Logistics loss rate refers to the percentage of goods lost in weight or quality during transportation, warehousing, and distribution due to natural factors (such as evaporation and spoilage) or technical operations (such as loading and unloading collisions and sorting errors). Its core metric is the percentage of lost goods to the initial total, reflecting the efficiency and control level of the supply chain during physical delivery.

[0042] Transport node weight refers to a quantitative assessment of the importance of specific hubs (such as warehouses, distribution centers, and transportation hubs) in a logistics network, based on their functional attributes, traffic carrying capacity, and network connectivity. This weight measures the impact of a node on overall logistics efficiency and is typically calculated from a combination of parameters such as transport volume, service coverage, resource processing efficiency, and network topology. It serves as a core basis for optimizing route planning and resource allocation.

[0043] A logistics scheduling solution is an execution strategy that efficiently matches transport capacity and optimizes transportation routes through systematic planning and dynamic adjustments based on cargo attributes (such as weight, specifications, and timeliness requirements), transportation resource distribution (vehicles, personnel), and external constraints (road conditions, weather). Its core goal is to balance transportation efficiency, cost control, and service quality, encompassing an integrated operational framework encompassing vehicle allocation, warehouse coordination, route optimization, and real-time monitoring.

[0044] Step S500: Obtain scheduling instructions from the optimized logistics scheduling plan, record the transaction data between the production end and the consumer end on the blockchain platform through smart contract technology, determine the integrity of the transaction data, and obtain traceable supply and demand transaction records.

[0045] Dispatch instructions are mandatory and authoritative operational instructions issued by higher-level dispatching agencies or on-duty dispatchers to subordinate units or personnel. These instructions are designed to dynamically adjust resource allocation, respond to emergencies, and ensure system security and operational efficiency. Their core function is to ensure the standardization and controllability of task execution in scenarios such as production, transportation, and energy systems through standardized operational commands.

[0046] Smart contract technology is a digital protocol system based on blockchain technology, which uses pre-programmed code to automatically execute, verify, and manage contract terms. Its core logic transforms the traditional contract execution process into a programmable conditional response mechanism, ensuring trusted transactions and asset transfers without third-party intervention.

[0047] Blockchain platforms are an open technology ecosystem built on distributed ledger technology, integrating consensus mechanisms, encryption algorithms, and smart contracts. They support the development of decentralized applications (DApps), asset digitization, and the execution of multi-party collaborative agreements. Essentially, they utilize a network of nodes and algorithmic rules to build a collaborative infrastructure that eliminates the need for trusted intermediaries, ensuring data immutability, transparent ownership traceability, and automated processes.

[0048] Supply and demand transaction records are on-chain data sets that record the exchange of resources or services between suppliers and demanders via blockchain networks within decentralized applications (DApps). Their core feature is the automated execution of transaction terms through smart contracts, with transaction details (such as time, quantity, price, and participant addresses) stored in an immutable form on a distributed ledger. These records are verified through a decentralized verification mechanism and transparently queried by all nodes in the network, eliminating the information asymmetry and trust issues inherent in traditional transactions.

[0049] Furthermore, the artificial intelligence-based rural e-commerce service method provided in this embodiment includes step S100: Step S110: Using standardized protocols to convert the agricultural production data, market transaction data, and consumer behavior data obtained from IoT sensors, trading platforms, and consumer feedback systems into a unified format data set.

[0050] In one possible implementation, standardized protocols are used to convert agricultural production data, market transaction data, and consumer behavior data. For example, soil moisture data collected by IoT sensors might be output in JSON format, market price data from trading platforms might be stored in CSV format, and evaluation data from consumer feedback systems might be recorded in XML format.

[0051] To achieve a unified format, JSON-LD can be used as a standardized protocol, converting all data into a unified key-value structure. Specifically, soil moisture data can be converted into JSON-LD objects containing fields such as "sensor_id," "timestamp," and "humidity." Transaction data can be converted into objects containing fields such as "product_type," "price," and "trade_time." Consumer data can include fields such as "user_id," "rating," and "comment." This approach ensures that data from different sources maintains consistency in semantics and structure, facilitating subsequent integration and analysis.

[0052] Step S120: Storing the data set in a unified format through a distributed database, using distributed storage technology to perform partition management on IoT data, transaction data, and consumer data to obtain an integrated agricultural data set.

[0053] For example, distributed database storage solutions can be implemented with partition management using Apache Cassandra (an open-source distributed database management system). Apache Cassandra is based on distributed hash table technology and can efficiently handle high-concurrency writes and queries.

[0054] In one embodiment, IoT data is partitioned by geographic region, with sensor data from Northeast China stored in one partition and data from South China in another. Transaction data is partitioned by time, such as storing data for the first quarter of 2025 in a separate node. Consumer data is partitioned by user ID range. This partitioning approach improves data retrieval efficiency and reduces single-point load. For example, when querying soil moisture data for Northeast China, only the specific partition needs to be accessed, reducing response time from seconds to milliseconds.

[0055] Specifically, distributed storage technology enables high data availability and fault tolerance through the Hadoop Distributed File System (HDFS). For example, a file containing soil moisture data in an agricultural dataset might be 10GB in size, transaction data 5GB, and consumer data 2GB. HDFS stores this data in shards across multiple nodes, with each shard being 128MB in size and creating three replicas for each shard. If a node fails, the system can quickly recover the data from the remaining replicas.

[0056] This approach has the beneficial effect of ensuring data reliability while supporting large-scale parallel data processing. For example, when analyzing national wheat price trends, HDFS can read transaction data in parallel, significantly improving processing speed.

[0057] It's important to note that the integrated agricultural data set can be further optimized using a data lake architecture. For example, a Delta Lake can be used to store the integrated data and support ACID (the abbreviation for the four essential elements of correct execution of database transactions) transactions to ensure data consistency.

[0058] In one example, the agricultural data set stored in the data lake contains 10 million records covering production, transactions, and consumer feedback. Analysts can use SQL queries to quickly extract correlations between wheat yield and price in a specific region and generate visual reports. This integration approach has the beneficial effect of breaking down data silos and enabling cross-domain analysis. For example, by combining soil moisture and consumer rating data, agricultural companies can optimize planting strategies and enhance their market competitiveness.

[0059] Implementation of the aforementioned technical subject matter should preferably consider data privacy. For example, differential privacy techniques can be used to add noise to consumer behavior data during conversion and storage to protect user identities. When storing data in a distributed manner, data encryption and access control ensure that only authorized users can access specific partitions. This approach has the beneficial effect of improving data security and complying with regulatory requirements without compromising analytical accuracy.

[0060] Furthermore, the artificial intelligence-based rural e-commerce service method provided in this embodiment includes step S200: Step S210: Obtain a data set in a unified format from the agricultural data integration, use distributed computing technology to partition the multi-source data, and divide it into farmer production data partitions, market transaction data partitions, and consumer behavior data partitions according to the data source to obtain the partitioned agricultural data set.

[0061] For example, when obtaining a unified format data set from agricultural data integration, multi-source data can be partitioned and processed through a distributed computing framework.

[0062] Specifically, farmers' production data includes information such as planting area, crop types and yields, market transaction data includes price fluctuations and transaction volume, and consumer behavior data covers purchasing preferences and feedback scores.

[0063] One possible implementation uses Apache Spark as a distributed computing framework, partitioning data into three partitions based on source. Farmer production data is partitioned and stored in a specific HDFS directory, market transaction data is partitioned by time series, and consumer behavior data is partitioned by user group. This partitioning facilitates parallel processing and improves computational efficiency. For example, when processing 10 million pieces of farmer production data, Apache Spark can distribute the data across 10 nodes for parallel computation, significantly reducing processing time.

[0064] Step S220: If the partitioned agricultural data set contains farmer production data, the farmer production data partition is encrypted using differential privacy technology, and noise is added using preset differential privacy parameters to obtain the encrypted farmer production data partition.

[0065] In one embodiment, when performing differential privacy encryption on partitions of farmer production data, sensitive information, such as farmer yield data, can be protected by adding Laplace noise. The default differential privacy parameter ε is set to 0.1 to ensure strong privacy protection. For example, if a farmer's wheat yield is 5,000 kilograms, adding noise may reduce the data to 4,980 or 5,020 kilograms. This approach protects privacy while preserving the statistical properties of the data, facilitating subsequent analysis.

[0066] It should be noted that differential privacy can be implemented through open source libraries such as Opacus and is suitable for distributed environments.

[0067] Step S230: Based on the encrypted farmer production data partitions and other data partitions, a federated learning algorithm is used to perform training in a distributed computing environment, and model parameters are generated for the supply and demand forecasting needs of each region to obtain an initial regionalized supply and demand forecasting model.

[0068] For example, when using the federated learning algorithm for model training, the data partitions of each region do not need to be stored centrally. Instead, the model parameters are updated through local calculations and finally aggregated to generate the initial regionalized supply and demand forecast model.

[0069] Specifically, a sub-model is trained locally based on the production data of farmers in Northeast China. After generating supply and demand forecast parameters, only the parameters, not the original data, are uploaded to the central server. This approach effectively protects data privacy. For example, the predicted wheat supply and demand gap in Northeast China is 100,000 tons, while in South China it is 50,000 tons. These parameters are then aggregated to form a national model.

[0070] The advantage of federated learning is that it reduces the risk of data leakage and supports cross-regional collaboration. In one embodiment, the initial regionalized supply and demand forecasting model can be used to perform forecast calculations on regional data, combining time series analysis and machine learning algorithms. For example, an LSTM model can be used to predict wheat supply and demand trends for the first quarter of 2025.

[0071] The model uses historical production and market price data from the Northeast region as input. After outputting a forecast, it adjusts parameters based on actual supply and demand deviations. For example, if the forecasted supply-demand gap is 80,000 tons but the actual gap is 90,000 tons, the model optimizes weights through backpropagation to generate a privacy-preserving forecast model. This approach ensures forecast accuracy while protecting data security.

[0072] Step S240: Perform forecast calculations on the partitioned agricultural data set using the initial regionalized supply and demand forecast model, obtain the supply and demand forecast results for each region, and adjust the model parameters to obtain a privacy-preserving forecast model.

[0073] For example, privacy-preserving prediction models can support dynamic supply and demand management after adjusting model parameters.

[0074] Specifically, agricultural companies can adjust their planting plans based on forecasts for the Northeast region, such as increasing wheat plantings by 10%. Market transaction data from specific regions guides pricing strategies, while consumer behavior data helps optimize product positioning. This multi-region collaborative forecasting approach improves agricultural supply chain efficiency and facilitates accurate decision-making.

[0075] Furthermore, the artificial intelligence-based rural e-commerce service method provided in this embodiment includes step S300: Step S310: Obtain supply and demand forecast results from the privacy-preserving forecast model, combine them with real-time market transaction data, use distributed computing technology to partition multi-source data, generate partitioned data sets based on regional divisions, and obtain regionalized supply and demand forecast data.

[0076] For example, when obtaining supply and demand forecast results from a privacy-preserving forecasting model, multi-source data can be processed through a distributed computing framework.

[0077] Specifically, suppose an agricultural supply chain system needs to predict the supply and demand of wheat in the second quarter of 2025. The forecasting model outputs that the supply and demand gap in Northeast China is 120,000 tons and in South China is 60,000 tons.

[0078] Real-time market trading data includes daily price fluctuations and trading volume. For example, the spot price of wheat in Northeast China is 2,500 yuan per ton, and the trading volume is 5,000 tons.

[0079] This data is stream-processed using Apache Flink and divided into regions such as Northeast China and South China to generate regional supply and demand forecasts. The partitioned data sets are stored in a distributed file system, such as HDFS, in region-specific directories to facilitate parallel computing.

[0080] Step S320: Based on the regionalized supply and demand forecast data, a time series analysis algorithm is used to process the real-time market transaction data, and price trends of agricultural products in each region are calculated to obtain regionalized price trend data.

[0081] In one possible implementation, a time series analysis algorithm is used to process real-time market transaction data to calculate price trends. For example, an ARIMA model can be used to analyze historical wheat price data in Northeast China and, combined with real-time transaction data, predict price trends for the next 30 days. Suppose the model predicts that prices in Northeast China will rise by 5% to 2,625 yuan per ton, while prices in South China will remain stable at 2,400 yuan per ton.

[0082] Step S330: If the regionalized price trend data meets the preset threshold, the regionalized supply and demand forecast data is combined with a linear regression algorithm to calculate the logistics demand of each region to obtain regionalized logistics demand data.

[0083] If the preset threshold for price fluctuation exceeds 3%, the Northeast region meets the criteria, triggering subsequent logistics demand calculations. The ARIMA model (autoregressive integrated moving average) analyzes the time series characteristics of prices to ensure the accuracy of trend forecasts. For example, a linear regression algorithm is used to calculate logistics demand based on regional supply and demand forecasts and price trend data.

[0084] Specifically, using a 120,000-ton supply-demand gap and a 5% price increase in Northeast China as input features, the linear regression model predicts a need for approximately 1,000 additional truckloads, each carrying 120 tons. In South China, due to stable prices, the increase in logistics demand is smaller, at approximately 500 truckloads. This approach optimizes logistics resource allocation by quantifying the relationship between supply, demand, and price.

[0085] Step S340: Using regionalized logistics demand data and regionalized price trend data, distributed computing technology is used to integrate the data to generate a real-time supply and demand matching solution for each region.

[0086] One possible implementation involves integrating regional logistics demand and price trend data through distributed computing technology to generate real-time supply and demand matching solutions. For example, Apache Kafka could be used as a message queue to transmit logistics demand and price trend data from the Northeast region to a central server in real time.

[0087] The server runs a matching algorithm and generates a plan: Northeast China prioritizes allocating 50,000 tons of local inventory, with the remaining 70,000 tons transferred from neighboring regions. South China meets demand through local production. Once the plan is generated, it is distributed to agricultural enterprises for adjustments to transportation plans and inventory management. This approach improves supply chain responsiveness through data integration and real-time matching.

[0088] It's important to note that distributed computing technologies must ensure data consistency and low latency when integrating multi-source data. For example, Kafka partitions store data streams from different regions, and Flink processes nodes in parallel to compute matching solutions, shortening response times. The entire process, from prediction to matching, forms a closed loop, ensuring accurate and efficient regional supply and demand management.

[0089] Furthermore, the artificial intelligence-based rural e-commerce service method provided in this embodiment includes step S400: Step S410: Based on the real-time supply and demand matching solution, a graph neural network algorithm is used to calculate the logistics path in supply chain management, an initial path set is generated for the transportation nodes in each region, and regionalized logistics path data is obtained.

[0090] For example, in supply chain management, when using graph neural network algorithms to calculate logistics paths, an initial path set can be generated by analyzing the relationships between transportation nodes.

[0091] Specifically, suppose an agricultural supply chain needs to optimize the soybean logistics route in the third quarter of 2025. The nodes include Harbin and Qiqihar in the Northeast, and Shanghai and Nanjing in the East China.

[0092] Graph neural networks generate an initial set of routes based on features such as node distance, transportation costs, and traffic conditions. For example, a route from Harbin to Shanghai passes through Shenyang, with a transportation distance of 2,000 kilometers and a cost of 300 yuan per ton. This approach captures the complex relationships between nodes and ensures that the set of routes covers a wide range of possibilities.

[0093] Step S420: Calculate the logistics loss rate of each path using statistical analysis tools based on regionalized logistics path data. If the logistics loss rate exceeds a preset threshold, redistribute the connection strength of the transport node to obtain adjusted node weight data.

[0094] In one possible implementation, based on regionalized logistics path data, statistical analysis tools are used to calculate the logistics loss rate of each path.

[0095] Taking transportation from Northeast China to East China as an example, statistical analysis tools processed historical data and found that the Harbin-Shanghai route had a loss rate of 5% due to the high number of transfers, exceeding the preset threshold of 3%. Other routes, such as Qiqihar-Nanjing, had a loss rate of 2.5%, meeting the requirement.

[0096] High loss rates may be due to storage conditions at transit points or weather conditions during transportation. Statistical analysis quantifies loss and provides a basis for subsequent optimization. For example, for routes with excessive loss rates, the connection strength of transport nodes can be reallocated to adjust node weights.

[0097] Specifically, due to the high loss rate on the Harbin-Shanghai route, the weight of the Shenyang transit node was reduced, while the weight of the direct route was increased. After this adjustment, the weight of the direct route from Harbin to Shanghai was increased from 0.3 to 0.5, generating new node weight data. This adjustment optimizes the network structure by reducing the impact of high-loss nodes.

[0098] Step S430: Based on the adjusted node weight data, a graph calculation tool is used to replan the logistics path, generate an optimized path set for the distribution demand of each region, and determine the regionalized optimized path data.

[0099] In one possible implementation, a graph computing tool is used to replan the logistics route based on the adjusted node weight data.

[0100] Optimally, a graph calculation tool is used to recalculate the route from Harbin to Shanghai, generating a new set of routes. For example, a direct route with a distance of 1,800 kilometers and a cost of 320 yuan per ton reduces the loss rate to 2.8%. Other routes in East China, such as Qiqihar to Nanjing, remain unchanged. This approach improves transportation efficiency by dynamically adjusting routes.

[0101] Step S440: By using the regionalized optimized path data, combined with inventory allocation and regional division, a scheduling algorithm is used to integrate the impact of distribution efficiency and price fluctuations to generate a logistics scheduling plan for each region.

[0102] For example, by combining regionalized optimized path data with inventory allocation and regional division, a scheduling algorithm is used to generate a logistics scheduling plan.

[0103] Specifically, Northeast China's soybean inventory is concentrated in Harbin, while East China's demand is 100,000 tons. The scheduling algorithm comprehensively optimizes route data and price fluctuations to generate a plan: 70,000 tons will be shipped directly from Harbin to Shanghai, and 30,000 tons will be shipped from Qiqihar to Nanjing.

[0104] If price fluctuations exceed 2%, the path with the lowest loss rate is prioritized. This approach balances efficiency and cost by integrating multi-dimensional data.

[0105] It's important to note that scheduling algorithms must consider regional characteristics when integrating delivery efficiency and price fluctuations. For example, in East China, peak demand occurs in the summer, so scheduling prioritizes faster routes. In Northeast China, when inventory is abundant, batch shipping can reduce costs. This approach improves supply chain responsiveness through refined scheduling.

[0106] Preferably, the artificial intelligence-based rural e-commerce service method provided in this embodiment includes step S500: Step S510: Obtain scheduling instructions from the optimized logistics scheduling plan, use data analysis tools to extract the distribution paths and time windows of each area, combine the supply and demand matching data of the production end and the consumption end, generate an instruction set including the path, time and cargo volume, and obtain structured scheduling instruction data.

[0107] In one possible implementation, optimizing logistics scheduling involves generating dispatch instructions, and data analysis tools can extract delivery routes and time windows. For example, an agricultural supply chain needs to schedule soybean shipments for the third quarter of 2025. The route from Harbin in Northeast China to Shanghai in East China is a direct route, covering a distance of 1,800 kilometers, and the delivery time window is within 48 hours.

[0108] Data parsing tools analyze supply and demand matching data, revealing Harbin's inventory of 80,000 tons and Shanghai's demand for 60,000 tons. This generates a set of instructions: a direct route from Harbin to Shanghai, with a cargo volume of 60,000 tons, and a time window from July 15th to 17th. Structured dispatch instruction data records the route, cargo volume, and time in a tabular format to ensure clear instructions.

[0109] Step S520: Based on the structured dispatch instruction data, a transaction contract is generated on the blockchain platform using a smart contract generation tool. Based on the volume and price fluctuation data of the production and consumption ends, a contract record containing transaction terms is generated to determine executable transaction contract data.

[0110] For example, based on dispatch instruction data, a smart contract generation tool generates a transaction contract on the blockchain platform. To account for fluctuations in volume and price, the contract records the following terms: 60,000 tons of soybeans, priced at 2,000 yuan per ton, with a fluctuation range of ±2%. In one possible implementation, when generating the contract, the tool automatically embeds the identity information of the Harbin supplier (production side) and the Shanghai purchaser (consumption side), generating a record with the contract number 20250715-01. The blockchain platform ensures that the terms cannot be tampered with, enhancing transaction trust.

[0111] Step S530: Use the executable transaction contract data and a data verification tool to verify the timestamp, quantity, and price fields of the transaction record. If the timestamp, quantity, and price fields of the transaction record are consistent with the dispatch instruction data, the integrity of the transaction data is confirmed, and the verified transaction record data is obtained.

[0112] As you can understand, data verification tools verify the integrity of transaction records. For example, the verification tool compares the transaction record's timestamp (2025-07-15 10:00), cargo volume (60,000 tons), and price (2,000 yuan per ton) with the dispatch instruction data to see if they are consistent. If they are, the transaction record is considered valid. This verification process is completed through the blockchain's distributed ledger, ensuring data transparency. Verified transaction records provide the basis for subsequent tracing.

[0113] Step S540: Based on the verified transaction record data, a blockchain query tool is used to generate traceable records including the production end, consumption end, and distribution path. For the supply and demand matching data of each region, a transaction record containing a timestamp and contract number is generated to determine a traceable supply and demand transaction data set.

[0114] In one possible implementation, a blockchain query tool generates traceable records, including the production location in Harbin, the consumption location in Shanghai, a direct delivery route, the timestamp 2025-07-15 10:00, and the contract number 20250715-01. For example, the query tool generates a supply and demand transaction dataset documenting the flow of 60,000 tons of soybeans from Northeast China to East China, which can be queried by regulators or businesses. This dataset can be searched by timestamp or contract number, facilitating the tracing of goods. For example, the above process forms a closed-loop management loop through data parsing, contract generation, data verification, and blockchain querying. The data parsing tool ensures instruction accuracy, the smart contract enhances transaction security, the verification tool ensures data consistency, and the query tool improves traceability efficiency. This approach optimizes supply chain scheduling, reduces transaction risks, and improves response time.

[0115] The present invention relates to an artificial intelligence-based rural e-commerce service system, which is used to implement the above-mentioned artificial intelligence-based rural e-commerce service method. The artificial intelligence-based rural e-commerce service system includes a first acquisition module, a second acquisition module, a third acquisition module, a fourth acquisition module and a fifth acquisition module, wherein the first acquisition module is used to acquire multi-source heterogeneous data from Internet of Things sensors, trading platforms and consumer feedback systems, and convert agricultural production data, market transaction data and consumer behavior data into a unified format using a standardized protocol, and obtain an integrated agricultural data set through distributed database storage; the second acquisition module is used to train a regional supply and demand forecasting model based on the integrated agricultural data set using a federated learning algorithm. If the data source of the integrated agricultural data set includes farmers' production data, the integrated agricultural data set is used to train a regional supply and demand forecasting model based on the integrated agricultural data set. The data is encrypted through differential privacy technology to obtain a privacy-protected prediction model; the third acquisition module is used to obtain supply and demand forecast results from the privacy-protected prediction model, and combined with real-time market transaction data, a time series analysis algorithm is used to calculate the price trend of agricultural products and logistics demand to obtain a real-time supply and demand matching plan; the fourth acquisition module is used to optimize the supply chain management path according to the real-time supply and demand matching plan using a graph neural network algorithm. If the logistics loss rate exceeds the preset threshold, the transportation node weight is adjusted to obtain an optimized logistics scheduling plan; the fifth acquisition module is used to obtain scheduling instructions from the optimized logistics scheduling plan, and record the transaction data of the production and consumption ends on the blockchain platform through smart contract technology, judge the integrity of the transaction data, and obtain traceable supply and demand transaction records.

[0116] Furthermore, the artificial intelligence-based rural e-commerce service system provided in this embodiment can The first acquisition module includes a first acquisition unit and a second acquisition unit, wherein the first acquisition unit is used to use a standardized protocol to convert the format of agricultural production data, market transaction data and consumer behavior data obtained from Internet of Things sensors, trading platforms and consumer feedback systems to obtain a data set in a unified format; the second acquisition unit is used to store the data set in a unified format through a distributed database, and use distributed storage technology to partition and manage Internet of Things data, transaction data and consumer data to obtain an integrated agricultural data set.

[0117] Preferably, the artificial intelligence-based rural e-commerce service system provided in this embodiment, the second acquisition module includes a third acquisition unit, a fourth acquisition unit, a fifth acquisition unit and a sixth acquisition unit, wherein the third acquisition unit is used to obtain a data set in a unified format from the agricultural data integration, and use distributed computing technology to partition the multi-source data, and divide it into farmer production data partitions, market transaction data partitions and consumer behavior data partitions according to the data source to obtain the partitioned agricultural data set; the fourth acquisition unit is used to encrypt the farmer production data partition by differential privacy technology if the partitioned agricultural data set contains farmer production data, and add noise by using preset differential privacy parameters to obtain encrypted farmer production data partitions; the fifth acquisition unit is used to train in a distributed computing environment based on the encrypted farmer production data partitions and other data partitions using a federated learning algorithm, generate model parameters for the supply and demand forecasting needs of each region, and obtain an initial regionalized supply and demand forecasting model; the sixth acquisition unit is used to perform predictive calculations on the partitioned agricultural data set by the initial regionalized supply and demand forecasting model, obtain the supply and demand forecast results of each region and adjust the model parameters to obtain a privacy-protected forecasting model.

[0118] Furthermore, the artificial intelligence-based rural e-commerce service system provided in this embodiment, the second acquisition module includes a seventh acquisition unit, an eighth acquisition unit, a ninth acquisition unit and a tenth acquisition unit, wherein the seventh acquisition unit is used to obtain supply and demand forecast results from a privacy-protected prediction model, combine real-time market transaction data, use distributed computing technology to partition multi-source data, generate partitioned data sets according to regional divisions, and obtain regionalized supply and demand forecast data; the eighth acquisition unit is used to process real-time market transaction data according to the regionalized supply and demand forecast data using a time series analysis algorithm, calculate price trends for agricultural products in each region, and obtain regionalized price trend data; the ninth acquisition unit is used to calculate the logistics demand of each region using a linear regression algorithm in combination with the regionalized supply and demand forecast data if the regionalized price trend data meets a preset threshold, and obtain regionalized logistics demand data; the tenth acquisition unit is used to integrate data through regionalized logistics demand data and regionalized price trend data using distributed computing technology to generate real-time supply and demand matching solutions for each region.

[0119] Compared with the existing technology, the artificial intelligence-based rural e-commerce service method and system provided in this embodiment integrates multi-source heterogeneous data and uses federated learning and differential privacy technologies to build a privacy-protected regional supply and demand forecasting model. Based on the forecast results and real-time market data, time series analysis is used to calculate price trends and logistics demand to generate a real-time supply and demand matching solution. Furthermore, this embodiment applies graph neural networks to optimize supply chain management paths and dynamically adjusts the weights of transportation nodes to reduce logistics losses. Finally, transaction data is recorded through blockchain smart contracts to achieve full traceability. This embodiment effectively solves problems such as information asymmetry between agricultural production and market demand, low supply chain efficiency, etc., improves the accuracy of agricultural product supply and demand matching and logistics distribution efficiency, ensures data security and transaction credibility, and provides intelligent decision-making support for all links in the agricultural industry chain.

[0120] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the invention. Thus, the present invention is intended to include such changes and modifications as fall within the scope of the claims and their equivalents.

Claims

1. A rural e-commerce service method based on artificial intelligence, characterized in that: The following steps are involved: Acquire multi-source heterogeneous data from IoT sensors, trading platforms, and consumer feedback systems, convert agricultural production data, market transaction data, and consumer behavior data into a unified format using standardized protocols, and store the integrated agricultural data set in a distributed database. Based on the integrated agricultural data set, a federated learning algorithm is used to train a regional supply and demand forecasting model. If the data source of the integrated agricultural data set includes farmer production data, the data is encrypted using differential privacy technology to obtain a privacy-preserving forecasting model. Obtain supply and demand forecast results from a privacy-preserving forecasting model, combine them with real-time market transaction data, and use time series analysis algorithms to calculate agricultural product price trends and logistics demand to obtain a real-time supply and demand matching solution; Based on the real-time supply and demand matching solution, a graph neural network algorithm is used to optimize the supply chain management path. If the logistics loss rate exceeds a preset threshold, the transportation node weight is adjusted to obtain an optimized logistics scheduling solution. Obtain scheduling instructions from the optimized logistics scheduling plan, record transaction data between the production and consumption ends on the blockchain platform through smart contract technology, determine the integrity of the transaction data, and obtain traceable supply and demand transaction records.

2. The rural e-commerce service method based on artificial intelligence according to claim 1, characterized in that: The steps of acquiring multi-source heterogeneous data from IoT sensors, trading platforms, and consumer feedback systems, converting agricultural production data, market transaction data, and consumer behavior data into a unified format using standardized protocols, and storing the integrated agricultural data set in a distributed database include: Using standardized protocols to convert the agricultural production data, market transaction data, and consumer behavior data obtained from IoT sensors, trading platforms, and consumer feedback systems to obtain the data set in the unified format; The data sets in a unified format are stored in a distributed database, and the IoT data, transaction data and consumer data are partitioned and managed using distributed storage technology to obtain an integrated agricultural data set.

3. The rural e-commerce service method based on artificial intelligence according to claim 1, characterized in that: The steps of training a regional supply and demand forecasting model using a federated learning algorithm based on the integrated agricultural data set, and encrypting the data using differential privacy technology to obtain a privacy-protected forecasting model if the data source of the integrated agricultural data set includes farmer production data, include: The unified format data set is obtained from agricultural data integration, and distributed computing technology is used to partition the multi-source data into farmer production data partitions, market transaction data partitions, and consumer behavior data partitions according to the data sources, thereby obtaining a partitioned agricultural data set; If the partitioned agricultural data set contains farmer production data, the farmer production data partition is encrypted using differential privacy technology, and noise is added using preset differential privacy parameters to obtain the encrypted farmer production data partition; Based on the encrypted farmer production data partitions and other data partitions, a federated learning algorithm is used to train in a distributed computing environment to generate model parameters for the supply and demand forecast requirements of each region, and obtain an initial regionalized supply and demand forecast model; The initial regionalized supply and demand forecasting model is used to perform forecast calculations on the partitioned agricultural data set, obtain supply and demand forecast results for each region, and adjust model parameters to obtain a privacy-preserving forecasting model.

4. The artificial intelligence-based rural e-commerce service method according to claim 1, characterized in that: The steps of obtaining supply and demand forecast results from the privacy-preserving forecast model, combining them with real-time market transaction data, and using a time series analysis algorithm to calculate agricultural product price trends and logistics demand to obtain a real-time supply and demand matching solution include: Obtain supply and demand forecast results from a privacy-preserving forecast model, combine it with real-time market transaction data, and use distributed computing technology to partition multi-source data. Generate partitioned data sets based on regional divisions to obtain regionalized supply and demand forecast data. Based on regional supply and demand forecast data, time series analysis algorithms are used to process real-time market transaction data, calculate price trends for agricultural products in each region, and obtain regional price trend data; If the regionalized price trend data meets the preset threshold, the regionalized supply and demand forecast data is combined with a linear regression algorithm to calculate the logistics demand of each region to obtain regionalized logistics demand data; By using regionalized logistics demand data and regionalized price trend data, distributed computing technology is used to integrate the data to generate a real-time supply and demand matching solution for each region.

5. The rural e-commerce service method based on artificial intelligence according to claim 1, characterized in that: According to the real-time supply and demand matching solution, a graph neural network algorithm is used to optimize the supply chain management path. If the logistics loss rate exceeds a preset threshold, the transportation node weight is adjusted. The steps of obtaining an optimized logistics scheduling solution include: Based on the real-time supply and demand matching solution, the graph neural network algorithm is used to calculate the logistics path in supply chain management, generate an initial path set for the transportation nodes in each region, and obtain regionalized logistics path data; Using regionalized logistics path data, statistical analysis tools are used to calculate the logistics loss rate of each path. If the logistics loss rate exceeds a preset threshold, the connection strength of the transportation node is redistributed to obtain adjusted node weight data; Based on the adjusted node weight data, graph calculation tools are used to re-plan logistics routes, generate optimized route sets based on the distribution needs of each region, and determine regional optimized route data; By optimizing regionalized route data, combining inventory allocation and regional division, and using scheduling algorithms to integrate the impact of distribution efficiency and price fluctuations, logistics scheduling plans for each region are generated.

6. The rural e-commerce service method based on artificial intelligence according to claim 1, characterized in that: The steps of obtaining scheduling instructions from the optimized logistics scheduling plan, recording transaction data between the production end and the consumer end on the blockchain platform through smart contract technology, determining the integrity of the transaction data, and obtaining traceable supply and demand transaction records include: Obtain dispatch instructions from the optimized logistics dispatch plan, use data analysis tools to extract the delivery routes and time windows for each region, combine the supply and demand matching data from the production and consumption ends, generate an instruction set containing the route, time, and quantity, and obtain structured dispatch instruction data; Based on structured dispatch instruction data, a smart contract generation tool is used to generate a transaction contract on the blockchain platform. Based on the volume and price fluctuation data of the production and consumption ends, a contract record containing transaction terms is generated to determine the executable transaction contract data. Using the executable transaction contract data, a data verification tool is used to verify the timestamp, quantity, and price fields of the transaction record. If the timestamp, quantity, and price fields of the transaction record are consistent with the dispatch instruction data, the transaction data integrity is confirmed, and the verified transaction record data is obtained; Based on the verified transaction record data, blockchain query tools are used to generate traceable records including the production end, consumption end and distribution path. For the supply and demand matching data of each region, transaction records containing timestamps and contract numbers are generated to determine the traceable supply and demand transaction data set.

7. An artificial intelligence-based rural e-commerce service system, used to implement the artificial intelligence-based rural e-commerce service method according to any one of claims 1 to 6, characterized in that: The artificial intelligence-based rural e-commerce service system includes: The first acquisition module is used to obtain multi-source heterogeneous data from IoT sensors, trading platforms, and consumer feedback systems. It uses standardized protocols to convert agricultural production data, market transaction data, and consumer behavior data into a unified format and stores the integrated agricultural data set in a distributed database. The second acquisition module is used to train a regional supply and demand forecasting model based on the integrated agricultural data set using a federated learning algorithm. If the data source of the integrated agricultural data set includes farmer production data, the data is encrypted using differential privacy technology to obtain a privacy-preserving forecasting model; The third acquisition module is used to obtain supply and demand forecast results from the privacy-preserving prediction model. Combined with real-time market transaction data, it uses a time series analysis algorithm to calculate agricultural product price trends and logistics demand to obtain a real-time supply and demand matching solution. A fourth acquisition module is configured to optimize the supply chain management path using a graph neural network algorithm based on the real-time supply and demand matching solution. If the logistics loss rate exceeds a preset threshold, the transportation node weight is adjusted to obtain an optimized logistics scheduling solution. The fifth acquisition module is used to obtain scheduling instructions from the optimized logistics scheduling plan, record the transaction data between the production and consumption ends on the blockchain platform through smart contract technology, determine the integrity of the transaction data, and obtain traceable supply and demand transaction records.

8. The artificial intelligence-based rural e-commerce service system according to claim 7, characterized in that: The first acquisition module includes: A first acquisition unit is configured to convert the agricultural production data, market transaction data, and consumer behavior data obtained from IoT sensors, trading platforms, and consumer feedback systems using a standardized protocol to obtain a data set in a unified format; The second acquisition unit is used to store data sets in a unified format through a distributed database, and use distributed storage technology to partition and manage IoT data, transaction data, and consumer data to obtain an integrated agricultural data set.

9. The artificial intelligence-based rural e-commerce service system according to claim 7, characterized in that: The second acquisition module includes: A third acquisition unit is configured to acquire the unified format data set from the agricultural data integration, partition the multi-source data using distributed computing technology, and divide the data into farmer production data partitions, market transaction data partitions, and consumer behavior data partitions according to the data sources, thereby obtaining a partitioned agricultural data set; a fourth acquisition unit, configured to, if the partitioned agricultural data set includes farmer production data, encrypt the farmer production data partition using differential privacy technology, add noise using a preset differential privacy parameter, and obtain an encrypted farmer production data partition; The fifth acquisition unit is used to use the federated learning algorithm to train in a distributed computing environment based on the encrypted farmer production data partitions and other data partitions, generate model parameters for the supply and demand forecast requirements of each region, and obtain an initial regionalized supply and demand forecast model; The sixth acquisition unit is used to perform prediction calculations on the partitioned agricultural data set using the initial regionalized supply and demand prediction model, obtain supply and demand prediction results for each region, and adjust model parameters to obtain a privacy-protected prediction model.

10. The artificial intelligence-based rural e-commerce service system according to claim 7, characterized in that: The second acquisition module includes: The seventh acquisition unit is used to obtain supply and demand forecast results from the privacy-preserving forecast model, combine them with real-time market transaction data, use distributed computing technology to partition multi-source data, generate partitioned data sets based on regional divisions, and obtain regionalized supply and demand forecast data; an eighth acquisition unit, configured to process real-time market transaction data using a time series analysis algorithm based on regionalized supply and demand forecast data, calculate price trends for agricultural products in each region, and obtain regionalized price trend data; a ninth acquisition unit configured to calculate the logistics demand of each region using a linear regression algorithm in combination with the regionalized supply and demand forecast data if the regionalized price trend data meets a preset threshold, thereby obtaining regionalized logistics demand data; The tenth acquisition unit is used to integrate the regionalized logistics demand data and the regionalized price trend data using distributed computing technology to generate a real-time supply and demand matching solution for each region.

Images