Vaccine circulation data encryption method and system based on data mining

Abstract

The present invention relates to the field of data encryption, and in particular to a vaccine circulation data encryption method and system based on data mining. The method comprises the following steps: collecting data from links in the vaccine circulation process to obtain multi-source heterogeneous circulation data, including structured circulation data, semi-structured circulation data, and unstructured circulation data; performing tensor decomposition on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the full state of vaccine circulation; constructing a circulation knowledge graph model based on the formal description of entities, relationships, and events in the heterogeneous fused circulation data; capturing temporal dynamic characteristics and long-term dependencies based on the heterogeneous fused circulation data, and establishing a time series analysis model. The present invention provides more comprehensive and accurate heterogeneous circulation data that describes the full state of vaccine circulation through in-depth analysis of vaccine circulation data.

Description

Technical Field

[0001] The present invention relates to the field of data encryption, and in particular to a vaccine circulation data encryption method and system based on data mining. Background Art

[0002] From production to the final recipient, vaccines undergo a complex supply chain management process. This process involves multiple links, including production, quality inspection, cold chain transportation, warehousing management, distribution, and ultimately, vaccination management. Each link must ensure the safety and effectiveness of the vaccine. Vaccine distribution involves a large amount of sensitive data, including patient information, vaccine batches, and transportation routes. This data must be strictly protected to prevent data leakage, tampering, or misuse. Data mining is the process of automatically discovering patterns, associations, and trends from large amounts of data. In vaccine distribution data encryption, data mining techniques can be applied to identify sensitive information in the data, such as personal identification information and drug batch numbers, and provide corresponding privacy protection measures. Encryption algorithms are important tools for protecting data privacy and security. Symmetric encryption algorithms (such as AES) or asymmetric encryption algorithms (such as RSA) can be used to encrypt sensitive data. In addition, hash algorithms (such as SHA-256) can be used to generate digital digests of the data to ensure data integrity. Traditional data encryption methods often cannot effectively handle the large amount of data generated during vaccine distribution and lack sufficient flexibility to adapt to the dynamic changes in the vaccine supply chain. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a vaccine circulation data encryption method and system based on data mining to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a vaccine circulation data encryption method based on data mining includes the following steps:

[0005] Step S1: Data is collected from all links in the vaccine circulation process to obtain multi-source heterogeneous circulation data, including structured circulation data, semi-structured circulation data, and unstructured circulation data; tensor decomposition is performed on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the entire status of vaccine circulation;

[0006] Step S2: Construct a circulation knowledge graph model based on the formal description of entities, relationships, and events in the heterogeneous fused circulation data; capture the temporal dynamic characteristics and long-term dependencies of the heterogeneous fused circulation data, and establish a time series analysis model; meta-learn the circulation knowledge graph model and the time series analysis model, and perform knowledge distillation to obtain collaborative model parameter data;

[0007] Step S3: Perform BGV homomorphic encryption on the collaborative model parameter data to obtain collaborative model parameter encrypted data; write smart contract code for storage and access control of the collaborative model parameter encrypted data, and deploy it on the blockchain platform to obtain smart contract address data; generate data access keys based on the collaborative model parameter encrypted data and the smart contract address data and enter access rights to obtain mapping relationship data between keys and access rights;

[0008] Step S4: Generate zero-knowledge proof data based on the mapping relationship data and the smart contract address data; package the mapping relationship data and the zero-knowledge proof data and upload them to the storage area of ​​the smart contract, thereby obtaining on-chain encrypted data; perform structural features based on the on-chain encrypted data, design a zero-knowledge proof protocol framework, and embed zero-knowledge characteristics to obtain proof protocol data with zero-knowledge properties;

[0009] Step S5: Use the on-chain encrypted data as input to run the proof protocol data, thereby generating verification proof data; perform security verification on the transmitted on-chain encrypted data based on the verification proof data, use TPM to build a secure isolation area, and decrypt the encrypted data on the chain in the secure isolation area to obtain a plaintext analysis report.

[0010] Through data collection, the present invention can obtain various data involved in the vaccine distribution process, including structured, semi-structured, and unstructured data. This data contains the vaccine's distribution status and related information, providing a foundation for subsequent analysis and modeling. By constructing a distribution knowledge graph model, the entities, relationships, and events in the vaccine distribution data can be formally described, establishing the connections and characteristics between them. Simultaneously, establishing a time series analysis model can capture the temporal dynamics and long-term dependencies of the data, enabling a better understanding and prediction of the vaccine distribution process. Data confidentiality and security can be ensured by encrypting and controlling storage access to collaborative model parameter data. The generated mapping relationship data between keys and access rights can be used to manage and control access to data, improving data privacy protection and access rights flexibility. By generating on-chain encrypted data from the mapping relationship data and smart contract address data, data confidentiality and integrity can be achieved on the blockchain. Furthermore, based on the structural characteristics of the on-chain encrypted data and the design of the zero-knowledge proof protocol framework, the zero-knowledge nature of the data can be ensured. Specifically, when verifying the proof data, the specific data content is not disclosed, thus protecting data privacy. By generating verification proof data and performing security verification, the integrity and correctness of the transmitted on-chain encrypted data can be verified to ensure that the data has not been tampered with. Utilizing a Trusted Platform Module (TPM) to establish a secure isolation zone provides additional data security. On-chain encrypted data is decrypted within the secure isolation zone, resulting in a plaintext analysis report providing detailed data analysis results and conclusions. Overall, the above steps include comprehensive analysis and modeling of vaccine circulation data, data encryption and privacy protection, and data security verification and decryption analysis. These steps help ensure the security, integrity, and credibility of vaccine circulation data and provide valuable data analysis results to support vaccine circulation management and decision-making.

[0011] The present invention also provides a vaccine circulation data encryption system based on data mining, which is used to execute the above-mentioned vaccine circulation data encryption method based on data mining. The vaccine circulation data encryption system based on data mining includes:

[0012] The multi-source data acquisition module is used to collect data from all links in the vaccine circulation process, thereby obtaining multi-source heterogeneous circulation data, including structured circulation data, semi-structured circulation data, and unstructured circulation data; and perform tensor decomposition on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the entire status of vaccine circulation;

[0013] The knowledge construction module is used to build a circulation knowledge graph model based on the formal description of entities, relationships, and events in heterogeneous fused circulation data; capture the temporal dynamic characteristics and long-term dependencies of the heterogeneous fused circulation data and establish a time series analysis model; meta-learn the circulation knowledge graph model and time series analysis model and perform knowledge distillation to obtain collaborative model parameter data;

[0014] The encryption and access control module is used to perform BGV homomorphic encryption on the collaborative model parameter data to obtain the collaborative model parameter encrypted data; write smart contract code for storage and access control of the collaborative model parameter encrypted data, and deploy it on the blockchain platform to obtain the smart contract address data; generate data access keys based on the collaborative model parameter encrypted data and the smart contract address data and enter access rights to obtain the mapping relationship data between the key and access rights;

[0015] The zero-knowledge proof module is used to generate zero-knowledge proof data based on the mapping relationship data and the smart contract address data; the mapping relationship data and the zero-knowledge proof data are packaged and uploaded to the storage area of ​​the smart contract to obtain on-chain encrypted data; based on the structural features of the on-chain encrypted data, a zero-knowledge proof protocol framework is designed and embedded with zero-knowledge characteristics to obtain proof protocol data with zero-knowledge properties;

[0016] The verification and decryption module is used to use the on-chain encrypted data as input to run the proof protocol data, thereby generating verification proof data; based on the verification proof data, the on-chain encrypted data after transmission is securely verified, a secure isolation area is constructed using TPM, and the encrypted data on the chain is decrypted in the secure isolation area to obtain a plaintext analysis report.

[0017] Through multi-source data collection, the present invention can acquire data from all aspects of the vaccine distribution process, including data in various formats and forms. This allows for comprehensive acquisition of distribution data, providing a more comprehensive and accurate information foundation to support subsequent data analysis and decision-making. By performing tensor decomposition on multi-source heterogeneous distribution data, the data can be converted into a higher-dimensional representation, extracting connections and features between the data. This allows for data describing the entire state of vaccine distribution, providing a more comprehensive and accurate data perspective and laying the foundation for subsequent knowledge construction and analysis. By constructing a distribution knowledge graph model, entities, relationships, and events in vaccine distribution can be formally described, establishing knowledge connections and structures. A time series analysis model can be used to capture the temporal dynamics and long-term dependencies of distribution data. This provides an in-depth understanding and analysis of the vaccine distribution process, supporting subsequent decision-making and optimization. By performing BGV homomorphic encryption on collaborative model parameter data, the confidentiality of model parameters can be protected, preventing unauthorized access. The development and deployment of smart contracts enables secure management of the storage and access of encrypted data. This ensures the security and authorized access of model parameters, protecting intellectual property and data privacy. By generating zero-knowledge proof data, data can be verified without revealing its specific content, protecting data privacy. Uploading this zero-knowledge proof data to the smart contract's storage area ensures secure data storage and verifiability. This provides trust and verifiability for on-chain encrypted data, enhancing data security and credibility. Verifying on-chain encrypted data confirms its integrity and authenticity, ensuring it has not been tampered with or forged. Decrypting on-chain encrypted data within a secure isolation zone created using a TPM yields decrypted plaintext data. Analyzing, analyzing, and visualizing this decrypted plaintext data provides in-depth understanding and insight, supporting decision-making and optimization. This ensures comprehensive data security and insight for the vaccine distribution process. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0019] Figure 1 Schematic diagram of the steps of the vaccine circulation data encryption method based on data mining of the present invention;

[0020] Figure 2 for Figure 1 Detailed step flow diagram of step S1;

[0021] Figure 3 for Figure 1 Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0022] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative work are within the scope of protection of the present invention.

[0023] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0024] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0025] To achieve this, please refer to Figures 1 to 3 The present invention provides a vaccine circulation data encryption method based on data mining, the method comprising the following steps:

[0026] Step S1: Data is collected from all links in the vaccine circulation process to obtain multi-source heterogeneous circulation data, including structured circulation data, semi-structured circulation data, and unstructured circulation data; tensor decomposition is performed on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the entire status of vaccine circulation;

[0027] Step S2: Construct a circulation knowledge graph model based on the formal description of entities, relationships, and events in the heterogeneous fused circulation data; capture the temporal dynamic characteristics and long-term dependencies of the heterogeneous fused circulation data, and establish a time series analysis model; meta-learn the circulation knowledge graph model and the time series analysis model, and perform knowledge distillation to obtain collaborative model parameter data;

[0028] Step S3: Perform BGV homomorphic encryption on the collaborative model parameter data to obtain collaborative model parameter encrypted data; write smart contract code for storage and access control of the collaborative model parameter encrypted data, and deploy it on the blockchain platform to obtain smart contract address data; generate data access keys based on the collaborative model parameter encrypted data and the smart contract address data and enter access rights to obtain mapping relationship data between keys and access rights;

[0029] Step S4: Generate zero-knowledge proof data based on the mapping relationship data and the smart contract address data; package the mapping relationship data and the zero-knowledge proof data and upload them to the storage area of ​​the smart contract, thereby obtaining on-chain encrypted data; perform structural features based on the on-chain encrypted data, design a zero-knowledge proof protocol framework, and embed zero-knowledge characteristics to obtain proof protocol data with zero-knowledge properties;

[0030] Step S5: Use the on-chain encrypted data as input to run the proof protocol data, thereby generating verification proof data; perform security verification on the transmitted on-chain encrypted data based on the verification proof data, use TPM to build a secure isolation area, and decrypt the encrypted data on the chain in the secure isolation area to obtain a plaintext analysis report.

[0031] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a vaccine circulation data encryption method based on data mining of the present invention. In this example, the vaccine circulation data encryption method based on data mining includes the following steps:

[0032] Step S1: Data is collected from all links in the vaccine circulation process to obtain multi-source heterogeneous circulation data, including structured circulation data, semi-structured circulation data, and unstructured circulation data; tensor decomposition is performed on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the entire status of vaccine circulation;

[0033] The embodiment of the present invention collects data through various links related to vaccine circulation, including structured, semi-structured and unstructured data. Data can be obtained in the following ways: Structured data: extract data from databases, API interfaces or other structured data sources. Semi-structured data: use crawler technology to extract data from sources such as web pages and log files. Unstructured data: use natural language processing technology to extract data from sources such as text, voice, and images. The collected data is cleaned, denoised and formatted to ensure the quality and consistency of the data. Use tensor decomposition technology (such as principal component analysis, singular value decomposition, etc.) to convert multi-source heterogeneous circulation data into heterogeneous fused circulation data that describes the full state of vaccine circulation. Tensor decomposition can extract the association and features of data and convert high-dimensional data into low-dimensional representations.

[0034] Step S2: Construct a circulation knowledge graph model based on the formal description of entities, relationships, and events in the heterogeneous fused circulation data; capture the temporal dynamic characteristics and long-term dependencies of the heterogeneous fused circulation data, and establish a time series analysis model; meta-learn the circulation knowledge graph model and the time series analysis model, and perform knowledge distillation to obtain collaborative model parameter data;

[0035] The embodiment of the present invention constructs a circulation knowledge graph model based on the entities, relationships and events in the heterogeneous fused circulation data. A graph database or knowledge representation technology (such as RDF, OWL, etc.) can be used to represent and store the knowledge graph. A time series analysis model is established by capturing the temporal dynamic characteristics and long-term dependencies of the heterogeneous fused circulation data. Statistical methods, machine learning methods or deep learning methods can be used to model and predict time series data. Meta-learning is performed on the knowledge graph model and the time series analysis model to learn their parameters and structures. Then, the knowledge of these models is transferred to the collaborative model using knowledge distillation technology to obtain more efficient and accurate model parameter data.

[0036] Step S3: Perform BGV homomorphic encryption on the collaborative model parameter data to obtain collaborative model parameter encrypted data; write smart contract code for storage and access control of the collaborative model parameter encrypted data, and deploy it on the blockchain platform to obtain smart contract address data; generate data access keys based on the collaborative model parameter encrypted data and the smart contract address data and enter access rights to obtain mapping relationship data between keys and access rights;

[0037] In this embodiment of the present invention, BGV homomorphic encryption is performed on collaborative model parameter data to protect the confidentiality of the model parameters. BGV homomorphic encryption is a cryptographic method that enables computation in an encrypted state. Smart contract code is developed to store and manage the storage and access control of the encrypted collaborative model parameter data. The smart contract is then deployed on a blockchain platform to obtain the smart contract address data. Data access keys are generated based on the encrypted collaborative model parameter data and the smart contract address data. Access rights are then entered into the smart contract, establishing a mapping between the key and access rights. This ensures that only authorized users can access the encrypted data.

[0038] Step S4: Generate zero-knowledge proof data based on the mapping relationship data and the smart contract address data; package the mapping relationship data and the zero-knowledge proof data and upload them to the storage area of ​​the smart contract, thereby obtaining on-chain encrypted data; perform structural features based on the on-chain encrypted data, design a zero-knowledge proof protocol framework, and embed zero-knowledge characteristics to obtain proof protocol data with zero-knowledge properties;

[0039] This embodiment of the present invention generates zero-knowledge proof data based on mapping relationship data and smart contract address data. Zero-knowledge proof is a cryptographic technique that can prove a statement is true without revealing specific information. The mapping relationship data and zero-knowledge proof data are packaged and uploaded to the smart contract's storage area to obtain on-chain encrypted data. This data is then stored in encrypted form on the blockchain, ensuring its confidentiality and security. Structural feature analysis is performed based on the on-chain encrypted data to extract data features and patterns. A zero-knowledge proof protocol framework is then designed to embed zero-knowledge properties into the proof protocol data, ensuring that sensitive information is not leaked during the verification process.

[0040] Step S5: Use the on-chain encrypted data as input to run the proof protocol data, thereby generating verification proof data; perform security verification on the transmitted on-chain encrypted data based on the verification proof data, use TPM to build a secure isolation area, and decrypt the encrypted data on the chain in the secure isolation area to obtain a plaintext analysis report.

[0041] This embodiment of the present invention uses on-chain encrypted data as input and runs the proof protocol data to generate verification proof data. This verification proof data verifies the correctness and integrity of the on-chain encrypted data. The transmitted on-chain encrypted data is then securely verified based on the verification proof data to ensure that the data has not been tampered with during transmission. A secure isolation zone is then established using a trusted execution environment (such as a TPM) to decrypt the on-chain encrypted data and obtain a plaintext analysis report.

[0042] Through data collection, the present invention can obtain various data involved in the vaccine distribution process, including structured, semi-structured, and unstructured data. This data contains the vaccine's distribution status and related information, providing a foundation for subsequent analysis and modeling. By constructing a distribution knowledge graph model, the entities, relationships, and events in the vaccine distribution data can be formally described, establishing the connections and characteristics between them. Simultaneously, establishing a time series analysis model can capture the temporal dynamics and long-term dependencies of the data, enabling a better understanding and prediction of the vaccine distribution process. Data confidentiality and security can be ensured by encrypting and controlling storage access to collaborative model parameter data. The generated mapping relationship data between keys and access rights can be used to manage and control access to data, improving data privacy protection and access rights flexibility. By generating on-chain encrypted data from the mapping relationship data and smart contract address data, data confidentiality and integrity can be achieved on the blockchain. Furthermore, based on the structural characteristics of the on-chain encrypted data and the design of the zero-knowledge proof protocol framework, the zero-knowledge nature of the data can be ensured. Specifically, when verifying the proof data, the specific data content is not disclosed, thus protecting data privacy. By generating verification proof data and performing security verification, the integrity and correctness of the transmitted on-chain encrypted data can be verified to ensure that the data has not been tampered with. Utilizing a Trusted Platform Module (TPM) to establish a secure isolation zone provides additional data security. On-chain encrypted data is decrypted within the secure isolation zone, resulting in a plaintext analysis report providing detailed data analysis results and conclusions. Overall, the above steps include comprehensive analysis and modeling of vaccine circulation data, data encryption and privacy protection, and data security verification and decryption analysis. These steps help ensure the security, integrity, and credibility of vaccine circulation data and provide valuable data analysis results to support vaccine circulation management and decision-making.

[0043] Preferably, step S1 includes the following steps:

[0044] Step S11: collecting temperature, humidity, and GPS information at different stages of vaccine circulation to obtain structured circulation data;

[0045] Step S12: using RFID tags and sensor network devices to collect semi-structured circulation data during the vaccine circulation process;

[0046] Step S13: Collecting pictures, texts, and video information through video surveillance to obtain unstructured circulation data;

[0047] Step S14: performing data cleaning on the structured circulation data, semi-structured circulation data, and unstructured circulation data, thereby obtaining multi-source heterogeneous circulation data;

[0048] Step S15: Perform tensor decomposition on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the entire state of vaccine circulation.

[0049] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in the embodiment of the present invention, step S1 includes the following steps:

[0050] Step S11: collecting temperature, humidity, and GPS information at different stages of vaccine circulation to obtain structured circulation data;

[0051] In this embodiment, sensors or other monitoring devices are used to collect real-time temperature, humidity, and GPS location information at different stages of vaccine distribution. These devices can be installed in transport containers, warehouses, transport vehicles, and other locations. The collected structured data, including temperature, humidity, and GPS information, is stored in a database or other data storage system for subsequent processing and analysis.

[0052] Step S12: using RFID tags and sensor network devices to collect semi-structured circulation data during the vaccine circulation process;

[0053] In this embodiment of the present invention, RFID tags and sensor network devices are used to collect semi-structured circulation data during the vaccine distribution process. RFID tags can be attached to vaccine packaging, and sensor network devices can be installed in locations such as warehouses and transport containers. RFID readers and sensor network devices collect semi-structured data, such as vaccine packaging identification information and status information, in real time. This data is stored in a database or other data storage system for subsequent processing and analysis.

[0054] Step S13: Collecting pictures, texts, and video information through video surveillance to obtain unstructured circulation data;

[0055] Embodiments of the present invention use video surveillance equipment to monitor and record the vaccine distribution process. These devices can be installed in locations such as warehouses and transport vehicles. The video surveillance equipment captures images, text, and video information during the vaccine distribution process. Image and text processing techniques can be used to extract key information from the video. The collected unstructured data, such as images, text, and video information, is stored in a database or other data storage system for subsequent processing and analysis.

[0056] Step S14: performing data cleaning on the structured circulation data, semi-structured circulation data, and unstructured circulation data, thereby obtaining multi-source heterogeneous circulation data;

[0057] This embodiment of the present invention cleans and denoises structured, semi-structured, and unstructured circulation data to remove errors, missing values, and outliers. The cleaned data is then integrated to form multi-source heterogeneous circulation data. The integrated data is formatted to ensure consistency and standardization for subsequent processing and analysis.

[0058] Step S15: Perform tensor decomposition on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the entire state of vaccine circulation.

[0059] The embodiments of the present invention use tensor decomposition techniques, such as principal component analysis (PCA) and singular value decomposition (SVD), to decompose multi-source heterogeneous circulation data. These techniques can extract the relevance and characteristics of the data and convert high-dimensional data into low-dimensional representations. Through tensor decomposition, heterogeneous fused circulation data describing the full state of vaccine circulation is obtained. This data will contain information from each source and comprehensively reflect the overall state of vaccine circulation.

[0060] The present invention installs sensor devices at various stages of vaccine distribution, such as production, transportation, warehousing, and distribution, to collect environmental parameters such as temperature and humidity, as well as GPS information to track logistics routes. This data is recorded and organized into a structured format for easy analysis and processing. By acquiring temperature, humidity, and GPS information, the environmental conditions and location of vaccines during distribution can be monitored. This structured distribution data provides real-time monitoring of vaccine storage and transportation conditions, helping to ensure vaccine quality and safety. RFID tags are attached to vaccine packaging or containers, and sensor network devices are used to collect vaccine-related information such as batch number, production date, and supplier information. This data can be semi-structured, as it has a standardized format, but the format may vary between different suppliers or production lines. Using RFID tags and sensor network devices, vaccine identification information and related data can be obtained in real time. This semi-structured distribution data provides tracking and tracing capabilities for vaccines, helping to monitor vaccine distribution routes and information. During the vaccine distribution process, video surveillance equipment is used to record actual scenes, including vaccine storage areas, transport vehicles, and operational processes. This video surveillance data can be converted into images, text, and video information for further analysis and processing. Video surveillance can capture the actual scenes and conditions of vaccine distribution. The collected images, text, and video information can provide more intuitive and comprehensive data, such as vaccine packaging status and operational procedures. This unstructured distribution data helps provide a more comprehensive understanding of the vaccine distribution environment and operational processes. The collected structured, semi-structured, and unstructured distribution data is processed and cleaned. Data cleaning includes removing duplicate data, fixing errors, and filling missing values ​​to ensure data consistency and accuracy. After cleaning, different types of data are integrated to form multi-source heterogeneous distribution data. Data cleaning can be used to remove noise, remove duplicates, and format the collected distribution data to improve data quality and consistency. The resulting multi-source heterogeneous distribution data is integrated into different types of data, providing a more accurate and comprehensive data foundation for subsequent analysis and modeling. Tensor decomposition is performed on the multi-source heterogeneous distribution data. Tensor decomposition is a data dimensionality reduction and feature extraction technique that transforms high-dimensional distribution data into a low-dimensional representation while preserving the data's key features. Through tensor decomposition, we can extract heterogeneous, fused circulation data that describes the full state of vaccine distribution, encompassing multiple aspects and dimensions of the vaccine distribution process. By performing tensor decomposition on multi-source, heterogeneous distribution data, we can reduce its dimensionality and extract features, resulting in a more concise and representative data representation. This heterogeneous, fused circulation data describing the full state of vaccine distribution integrates various types of distribution information, encompassing multiple aspects and dimensions of the vaccine distribution process, and helping to fully understand the status and characteristics of vaccine distribution.This data representation can provide a more reliable and effective foundation for subsequent analysis, modeling, and decision-making. In summary, the detailed explanation of the steps above demonstrates the process of collecting and processing vaccine circulation data. These steps help obtain more accurate, comprehensive, and usable vaccine circulation data, improve the ability to monitor vaccine quality and safety, and provide a more reliable basis for subsequent analysis, modeling, and decision-making.

[0061] Preferably, step S15 includes the following steps:

[0062] Step S151: integrating the multi-source heterogeneous data of circulation to generate full circulation feature vector data describing a single data entity;

[0063] The embodiment of the present invention integrates multi-source heterogeneous circulation data from different sources. This can include structured circulation data, semi-structured circulation data, and unstructured circulation data. The goal of the integration is to associate the information of each data source to generate full circulation feature vector data that describes a single data entity (such as a vaccine). Key features are extracted from the integrated data. This can include statistical features, time series features, spatial features, etc. The goal of feature extraction is to capture useful information in the data and convert it into a quantifiable feature vector. The extracted features are combined into a feature vector to describe the full circulation characteristics of a single data entity. The feature vector can be a numerical, binary, or other form of vector representation.

[0064] Step S152: performing tensor decomposition on the full-circulation feature vector data to obtain latent semantic space data;

[0065] Embodiments of the present invention use tensor decomposition techniques, such as principal component analysis (PCA) and singular value decomposition (SVD), to decompose fully circulated feature vector data. These techniques can transform high-dimensional data into low-dimensional representations and extract the latent semantic information in the data. By performing tensor decomposition, a representation describing the data entities in a latent semantic space is obtained. The latent semantic space data retains the important features of the original data while removing redundant information.

[0066] Step S153: Extract fusion features of the entire vaccine circulation state from the latent semantic space data to obtain heterogeneous fusion circulation data.

[0067] Embodiments of the present invention perform feature extraction on latent semantic space data to capture key characteristics of the overall vaccine distribution status. This may include utilizing machine learning algorithms, deep learning models, or other feature extraction methods to extract useful features. Features extracted from the latent semantic space data are combined into a heterogeneous fused distribution data set. This data integrates information from multiple sources to provide a holistic description of the overall vaccine distribution status.

[0068] The present invention integrates circulation data from different data sources to generate full-circulation feature vector data that describes a single data entity. These feature vector data can be numerical or discrete, and are used to describe various aspects and attributes of vaccine circulation data, such as temperature, humidity, time, location, supply chain stage, etc. Data integration and generation of feature vector data help to unify the formats and representations of different data sources, and provide a comprehensive description of a single data entity. Through full-circulation feature vector data, the key features of vaccine circulation data can be better captured, providing a more accurate and comprehensive data foundation for subsequent analysis and modeling. Tensor decomposition is performed on the full-circulation feature vector data to convert it into latent semantic space data. Tensor decomposition is a dimensionality reduction technology that converts high-dimensional feature vector data into a low-dimensional representation and retains the important features of the data. Latent semantic space data can better reflect the intrinsic structure and correlation of the data. Through tensor decomposition, the dimension of the data can be reduced, redundant information can be reduced, and potential semantic features can be extracted. Latent semantic space data has a more compact and representative representation, which helps to reduce the storage and computational complexity of the data, and can better capture the hidden patterns and correlations of vaccine circulation data. Feature extraction is performed on latent semantic space data to obtain heterogeneous fused circulation data that describes the full status of vaccine circulation. These heterogeneous fused circulation data contain features from multiple aspects and dimensions, such as vaccine quality, transportation routes, supply chain reliability, etc. By performing fused feature extraction on latent semantic space data, multiple aspects and dimensions of vaccine circulation data can be comprehensively considered to extract more representative and comprehensive features. These heterogeneous fused circulation data can more comprehensively describe the circulation status and characteristics of vaccines, providing a more reliable and effective foundation for subsequent analysis, modeling, and decision-making. In summary, the above steps include integrating data, generating full-circulation feature vector data, reducing dimensionality and extracting latent semantic space data, and fusing feature extraction to generate heterogeneous fused circulation data. These steps help reduce data redundancy and dimensionality, extract key features, capture the inherent structure and hidden patterns of the data, and provide a more comprehensive and integrated data foundation for describing the circulation status of vaccines.

[0069] Preferably, step S151 includes the following steps:

[0070] Step S1511: extracting data changes in the time series of the semi-structured circulation data and expressing it in the form of a tensor, thereby obtaining semi-structured vector data;

[0071] This embodiment of the present invention preprocesses semi-structured circulation data, including steps such as data cleaning, denoising, and standardization. The semi-structured circulation data is sorted chronologically and its changes over time are extracted. This can include calculating the degree of data change using techniques such as differencing and moving averages. The extracted data change information is represented as a tensor. Multidimensional arrays or matrices can be used to represent data changes, with each time point assigned a dimension.

[0072] Step S1512: performing pixel value digital representation on the image data in the unstructured circulation data and performing preset range mapping to obtain image vector data;

[0073] Embodiments of the present invention process image data in unstructured circulation data, converting the image into a digital representation. Image processing techniques, such as grayscale conversion, edge detection, and feature extraction, can be used to convert the image into numerical data. Preset range mapping is performed on the digitally represented image data, mapping pixel values ​​to a specific numerical range. This can be achieved through methods such as linear transformation, normalization, or standardization. The digitally represented and pre-set range-mapped image data is converted into a vector form, with each pixel value serving as a dimension of the vector.

[0074] Step S1513: extracting features from the text data in the unstructured circulation data using natural language processing technology to obtain text vector data;

[0075] In an embodiment of the present invention, text data in unstructured circulation data is preprocessed, including steps such as word segmentation, stop word removal, and stemming. Natural language processing techniques are used to extract features from the preprocessed text data. This can include methods such as bag-of-words models, TF-IDF (term frequency-inverse document frequency) weights, and word embeddings to represent text features. The extracted text features are represented in vector form. Methods such as word vector representation and one-hot encoding can be used to map text features to a vector space.

[0076] Step S1514: extracting frames from the video data in the unstructured circulation data, digitally representing the pixel values ​​of the video frames, and performing preset range mapping to obtain video vector data;

[0077] An embodiment of the present invention extracts key frames from video data in unstructured circulation data. Specific frames in the video can be selected as representations based on time intervals or a key frame extraction algorithm. The extracted video frames are digitally represented. Image processing techniques, such as grayscale, edge detection, and feature extraction, can be used to convert the video frames into numerical data. Preset range mapping is performed on the digitally represented video frames to map pixel values ​​into a specific numerical range. The video frame data that has undergone frame extraction, digital representation, and preset range mapping is converted into a vector form, with the pixel value of each frame serving as a dimension of the vector.

[0078] Step S1515: perform vector merging on the structured circulation data, semi-structured vector data, image vector data, text vector data, and video vector data, thereby generating full circulation feature vector data describing a single data entity.

[0079] The embodiment of the present invention represents structured circulation data as feature vectors, and the original numerical data can be used directly. The above-mentioned semi-structured vector data, image vector data, text vector data and video vector data are merged. Different types of vector data can be merged using methods such as vector splicing and feature combination. The feature vector of the structured circulation data and the merged unstructured vector data are merged to obtain the full circulation feature vector data describing a single data entity (such as a vaccine). The individual vectors can be connected according to the dimension to form a complete feature vector.

[0080] The present invention targets semi-structured circulation data by first extracting data changes in a time series. This involves sorting the data chronologically and extracting trends and patterns in the data over time. The extracted data changes are then represented as tensors and converted into multidimensional arrays for easier processing and analysis. This results in semi-structured vector data. By extracting data changes in a time series from semi-structured circulation data, dynamic trends in the data can be captured, including periodicity, trends, or other patterns. Representing data as tensors facilitates mathematical calculations and pattern recognition, providing a more effective foundation for subsequent feature extraction and data analysis. Regarding image data in unstructured circulation data, the image data is first converted into a digital representation of pixel values, representing each pixel in the image as a numerical value. The pixel values ​​are then mapped according to a preset range, confining them to a specific numerical range. Ultimately, vector data representing the image is obtained. By digitally representing the pixel values ​​and mapping them to a preset range, the image data can be converted into numerical vector data, allowing for unified processing and analysis with other types of data. This allows for better utilization of image information in circulation data, such as image quality, identifiers, and degree of damage, providing more comprehensive visual information for subsequent feature extraction and data integration. Natural language processing techniques are used to extract features from text data within unstructured circulation data. This can include techniques such as bag-of-words models, word embedding, and topic modeling, converting text data into numerical vector representations. By extracting features such as keywords, contextual information, and themes from the text, vector data describing the text is generated. Feature extraction from text data transforms unstructured text information into vector representations that computers can understand and process. This allows for better utilization of semantic and contextual information within the text data, such as information describing vaccine attributes, manufacturer, and batch number, providing more comprehensive text information for subsequent feature fusion and data analysis. For video data within unstructured circulation data, key frames are first extracted from the video to capture static information. Next, pixel values ​​are digitally represented for each video frame, representing each pixel in the frame as a numerical value. Finally, pixel values ​​are mapped according to a preset range, confining them to a specific numerical value. This results in vector data representing the video. By extracting frames from video data, representing pixel values ​​digitally, and mapping them to preset ranges, we can convert video data into numerical vector data, allowing it to be processed and analyzed uniformly with other types of data. This allows for better utilization of video information in circulation data, such as video surveillance, transportation processes, and quality inspections, providing more comprehensive visual information for subsequent feature extraction and data integration. By merging the various previously extracted vector data types, we generate full-circulation feature vector data describing a single data entity.This can be achieved by splicing, combining or superimposing different types of vector data according to certain rules. The final fully circulated feature vector data contains the characteristic information of multiple types of data, which can comprehensively describe the circulation status and characteristics of a single data entity. By merging different types of vector data, multiple aspects and dimensions of vaccine circulation data can be comprehensively considered to extract more representative and comprehensive features. Full circulation feature vector data can more comprehensively describe the circulation status and characteristics of vaccines, providing a more reliable and effective basis for subsequent analysis, modeling and decision-making. This comprehensive feature vector data helps to reveal the inherent correlations and hidden patterns between data, and provides a stronger data foundation for the application of machine learning, data mining and other technologies.

[0081] Preferably, step S1 includes the following steps:

[0082] Step S21: Identify key entities, relationships between entities, and important events in the vaccine circulation process based on heterogeneous fusion circulation data, thereby obtaining circulation knowledge graph data;

[0083] Step S22: constructing a circulation knowledge graph model based on the circulation knowledge graph data;

[0084] Step S23: performing time series characteristic analysis on the heterogeneous fused circulation data to obtain circulation time characteristic data;

[0085] Step S24: extracting long-term dependency relationships from the circulation time feature data to obtain dependency relationship data;

[0086] Step S25: establishing a time series analysis model based on the circulation time characteristic data and the dependency relationship data;

[0087] Step S26: Meta-learn the general knowledge graph model and the time series analysis model, and perform knowledge distillation to obtain collaborative model parameter data.

[0088] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 1 Detailed step flow diagram of step S2 in the embodiment of the present invention, step S2 includes the following steps:

[0089] Step S21: Identify key entities, relationships between entities, and important events in the vaccine circulation process based on heterogeneous fusion circulation data, thereby obtaining circulation knowledge graph data;

[0090] The embodiments of the present invention collect heterogeneous fusion circulation data, including structured data, semi-structured data and unstructured data. The data is preprocessed, cleaned and standardized to ensure the consistency and quality of the data. Natural language processing (NLP) and named entity recognition (NER) technologies are used to identify key entities in vaccine circulation data, such as vaccine name, manufacturer, batch number, location, etc. Through text analysis and relationship extraction technology, the relationship between entities in vaccine circulation data, such as production relationship, supply chain relationship, sales relationship, etc., is identified. Using technologies such as event extraction and text classification, important events in vaccine circulation data are identified, such as vaccine recalls, supply shortages, vaccine batch problems, etc. The identified key entities, relationships between entities and important events are organized into the form of a knowledge graph. The knowledge graph can be represented using a graph database or a graph structure, where entities are nodes and relationships are edges.

[0091] Step S22: constructing a circulation knowledge graph model based on the circulation knowledge graph data;

[0092] The embodiment of the present invention determines the model structure and properties of the knowledge graph. For example, define entity types, relationship types, attribute types, etc. Import the constructed circulation knowledge graph data into a graph database or other suitable storage system. Ensure the integrity and consistency of the data. Based on the constructed knowledge graph model, implement query and reasoning operations of the graph database. Query and reasoning can be performed using a query language (such as SPARQL) or the API of the graph database to obtain relevant information about vaccine circulation. In order to better understand and analyze the circulation knowledge graph, visualization tools can be used to visualize the knowledge graph. This can help users intuitively explore and discover relationships and patterns in vaccine circulation data.

[0093] Step S23: performing time series characteristic analysis on the heterogeneous fused circulation data to obtain circulation time characteristic data;

[0094] The embodiments of the present invention preprocess heterogeneous fused circulation data, including steps such as data cleaning, denoising, and standardization. Time series analysis methods, such as stationarity testing, autocorrelation analysis, and periodicity analysis, are used to extract time series characteristics from the circulation data. Various statistical indicators, volatility indicators, and periodicity indicators can be calculated. Based on the extracted time series characteristics, feature engineering is performed to construct a feature set for modeling. These can include lag features, moving average features, trend features, and the like. The feature-engineered time series characteristics are represented as vectors, with each feature vector representing the circulation time characteristic data for a time point or time period.

[0095] Step S24: extracting long-term dependency relationships from the circulation time feature data to obtain dependency relationship data;

[0096] This embodiment of the present invention defines the concept of long-term dependencies based on the characteristics of circulation time feature data and application requirements. Time series analysis methods, such as autoregressive models (AR) and hidden Markov models (HMMs), are used to extract long-term dependencies from circulation time feature data. The extracted dependencies are represented as a graph or matrix, where nodes represent time series features and edges represent dependencies. The extracted dependency data is then processed and optimized, including noise removal and pruning of insignificant relationships.

[0097] Step S25: establishing a time series analysis model based on the circulation time characteristic data and the dependency relationship data;

[0098] The present invention selects an appropriate time series analysis model, such as ARIMA or LSTM, based on the characteristics of the circulation time feature data and dependency data. The selected time series analysis model is trained using the circulation time feature data and dependency data. Common machine learning algorithms can be used for model training. Model parameters are tuned based on the model's performance metrics to improve the model's accuracy and predictive power. The trained time series analysis model is evaluated using evaluation metrics such as root mean square error (RMSE) and mean absolute error (MAE).

[0099] Step S26: Meta-learn the general knowledge graph model and the time series analysis model, and perform knowledge distillation to obtain collaborative model parameter data.

[0100] The embodiments of the present invention use meta-learning methods, such as model-independent optimization (MAML) and variational autoencoders (VAE), to meta-learn the general knowledge graph model and the time series analysis model to obtain the initial parameters of the model. The model parameters obtained by meta-learning are applied to the general knowledge graph model and the time series analysis model to distill the model knowledge. Knowledge distillation methods such as model distillation and parameter distillation can be used. The distilled model parameter data is applied to the general knowledge graph model and the time series analysis model to achieve synergy between the two. This can be achieved through model integration, shared parameters, etc.

[0101] The present invention analyzes and processes heterogeneous and integrated vaccine circulation data to identify key entities, such as vaccine batches, manufacturers, and supply chain nodes. It also identifies relationships between entities, such as supply relationships and logistics routes. Furthermore, it identifies important events, such as transactions, transport disruptions, and quality issues. Through these analyses, knowledge graph data describing vaccine circulation relationships is generated. By constructing a circulation knowledge graph, key entities, relationships between entities, and important events in the vaccine circulation process can be structured. This allows for a better understanding and analysis of the vaccine circulation process, revealing the connections between entities and circulation routes. The knowledge graph provides an intuitive way to represent and query key information about vaccine circulation, laying the foundation for subsequent analysis, reasoning, and decision-making. Based on the resulting circulation knowledge graph data, a circulation knowledge graph model is constructed. This model uses graph theory and graph analysis methods to model and represent the entities, relationships, and events in the knowledge graph. This modeling allows for a better understanding and analysis of the knowledge graph data in the vaccine circulation process, extracting useful information and patterns from it. By constructing a circulation knowledge graph model, the knowledge and patterns within vaccine circulation data can be more deeply explored. This model can help understand the structure and dynamics of vaccine distribution, uncovering hidden relationships and patterns. This provides a more accurate and comprehensive understanding of the vaccine distribution process, providing stronger support for subsequent analysis and decision-making. Time series characteristic analysis is performed on heterogeneous, integrated distribution data. This includes statistical analysis, trend analysis, and periodicity analysis of the temporal dimension. These analyses can extract indicators and features that describe the temporal characteristics of distribution data. Time series characteristic analysis of distribution data can reveal patterns and regularities in data changes over time. This helps understand the temporal properties of vaccine distribution, including seasonal demand, cyclical fluctuations, and trend changes. Extracting temporal characteristic data allows for better description and analysis of the dynamics of vaccine distribution, providing a useful data foundation for subsequent time series modeling and analysis. Long-term dependencies are extracted from distribution time characteristic data. Long-term dependencies refer to potential temporal dependencies between data points, meaning that current data points may be affected by points in the distant past. By analyzing the long-term dependencies of time series data, temporal correlations and dependencies can be identified. Extracting long-term dependencies from circulation time feature data can help better understand and model the temporal dynamics of vaccine distribution. These dependencies can reveal the temporal continuity and correlation of vaccine distribution data, helping to predict and explain future trends and behaviors. By extracting dependency data, important features in time series data can be more accurately captured and represented, providing a more reliable foundation for subsequent time series analysis and modeling. A time series analysis model was established based on the obtained circulation time feature data and dependency data.Time series analysis is a technique that uses statistical and mathematical methods to study time series data, aiming to uncover patterns, trends, and periodicity within the data and to enable prediction and analysis. Building a time series analysis model can help provide a deeper understanding and prediction of the temporal dynamics of vaccine distribution. Modeling and analyzing time series data can reveal patterns and regularities, including trends, seasonality, and periodicity. This can provide predictions and insights into the future behavior of the vaccine distribution process, informing subsequent planning and decision-making. Meta-learning and knowledge distillation are performed on the previously constructed distribution knowledge graph model and time series analysis model. Meta-learning is a machine learning method used to learn how to learn and adapt to different tasks. Meta-learning can optimize a model's learning and generalization capabilities, improving its performance on new tasks. Knowledge distillation is a method that transfers knowledge from a large model to a smaller model, reducing model complexity and improving inference efficiency. Meta-learning and knowledge distillation on the knowledge graph model and time series analysis model yield collaborative model parameter data. This collaborative model parameter data combines the strengths of the knowledge graph model and the time series analysis model, offering enhanced comprehensive analysis and inference capabilities. The collaborative model can better utilize knowledge graph data and time series data to provide more accurate and comprehensive vaccine circulation analysis and prediction.

[0102] Preferably, step S26 includes the following steps:

[0103] Step S261: Structural encoding is performed on the circulation knowledge graph model, and node structural features are extracted to obtain node structural feature data;

[0104] The embodiment of the present invention performs structural encoding on the circulation knowledge graph model and converts the nodes and the relationships between them into vector form. Structural encoding can be performed using methods such as Graph Neural Network (GNN). The structural features of the nodes are extracted from the structurally encoded circulation knowledge graph model. Aggregation functions (such as pooling operations, attention mechanisms, etc.) can be used to extract features from the neighbors of the node. The extracted node structural features are represented as vectors, with each node as a dimension of the vector.

[0105] Step S262: Using meta-path encoding technology to perform vector space representation learning on the time series analysis model, thereby obtaining time series feature embedding data;

[0106] This embodiment of the present invention defines a metapath for describing a time series analysis model, that is, describing the trajectory of a node sequence. The metapath can be designed based on the characteristics of the time series data and the requirements of the problem. Metapath encoding technology is used to convert the time series analysis model into a vector space representation. Metapath encoding can be performed using methods such as Node2Vec and DeepWalk. The resulting time series features are embedded and represented as a vector, with each time point or time period as a dimension of the vector.

[0107] Step S263: performing relationship learning processing on the node structure feature data and the temporal feature embedding data based on the attention mechanism, thereby obtaining an attention relationship matrix;

[0108] An embodiment of the present invention defines an attention mechanism for learning the relationship between node structure feature data and time series feature embedding data. The attention mechanism can measure the importance of different features through attention weights. Using the defined attention mechanism, the relationship between the node structure feature data and the time series feature embedding data is learned. The relationship learning process can be performed using a method such as a graph attention network (GAT). The learned attention relationship is represented in a matrix form, where the rows of the matrix represent the node structure feature data and the columns represent the time series feature embedding data.

[0109] Step S264: performing attention relationship connection on the circulation knowledge graph model and the time series analysis model according to the attention relationship matrix, thereby obtaining a knowledge graph-time series full model;

[0110] In this embodiment of the present invention, an attention relationship is established between the circulation knowledge graph model and the time series analysis model based on an attention relationship matrix. This connection can be achieved using operations such as matrix multiplication. The connected model is represented as a knowledge graph-time series full model, which combines information from the circulation knowledge graph model and the time series analysis model.

[0111] Step S265: Randomly sample knowledge graph nodes and time series steps in the knowledge graph-time series full model to obtain sample index data;

[0112] In an embodiment of the present invention, some nodes are randomly selected from the knowledge graph-time series full model as target nodes for sampling. A random sampling algorithm (such as uniform sampling, probabilistic sampling, etc.) can be used for node sampling. Some time series steps are randomly selected from the knowledge graph-time series full model as target steps for sampling. Random sampling can be performed based on the length and distribution of the time series. The sampled knowledge graph nodes and time series steps are represented as index data for subsequent model calculation and update.

[0113] Step S266: Calculate the information received by each node and step according to the knowledge graph-time series full model, and update its own representation according to the sampling index data to obtain the collaborative model parameter data.

[0114] This embodiment of the present invention calculates the information received by each node and step based on the knowledge graph-time series full model. Methods such as graph neural networks can be used for information transmission and aggregation. Based on the calculated received information, the representations of nodes and steps are updated using sampled index data. Model parameters can be updated using methods such as gradient descent. The updated model parameters are represented as collaborative model parameter data, including node and step representations. These parameters can be used for further analysis, prediction, or decision-making.

[0115] The present invention can represent the nodes in the knowledge graph as vector representations with structural information by performing structural encoding and node structure feature extraction on the circulation knowledge graph model. These node structure feature data can capture the associations and dependencies between nodes, which helps to reveal the functions and importance of nodes in the vaccine circulation process. By using meta-path encoding technology to perform vector space representation learning on the time series analysis model, the time series data can be converted into an embedded representation with semantic information. These time series feature embedding data can better capture the patterns and trends of time series data, which helps to understand and predict the temporal dynamics of the vaccine circulation process. By performing relational learning processing on the node structure feature data and the time series feature embedding data based on the attention mechanism, important relationships and dependencies between nodes can be identified. The obtained attention relationship matrix can highlight key nodes and related temporal features, providing attention and concentration on important information in the vaccine circulation process. By relationally connecting the circulation knowledge graph model and the time series analysis model according to the attention relationship matrix, the information of the two can be fused and interacted. The obtained knowledge graph-time series full model combines the advantages of the knowledge graph model and the time series analysis model, and can more comprehensively describe and analyze the key information and temporal dynamics in the vaccine circulation process. By randomly sampling knowledge graph nodes and time series steps in the knowledge graph-time series full model, a portion of sample data can be obtained for model training and updating. Sampling index data can help the model better utilize limited data samples and provide diverse observations and analyses of the vaccine circulation process. By calculating the information received by each node and step based on the knowledge graph-time series full model and updating its own representation based on the sampling index data, collaborative learning and parameter updates between the knowledge graph model and the time series analysis model can be achieved. The resulting collaborative model parameter data combines the advantages of the two models, has better comprehensive analysis and reasoning capabilities, and can more accurately describe and predict various situations and trends in the vaccine circulation process. This helps to improve the understanding and prediction capabilities of vaccine circulation and provide useful support for vaccine supply chain management and decision-making.

[0116] Preferably, step S3 includes the following steps:

[0117] Step S31: performing BGV homomorphic encryption on the collaborative model parameter data to obtain collaborative model parameter encrypted data;

[0118] This embodiment of the present invention selects and implements a BGV homomorphic encryption algorithm, such as the Brakerski-Gentry-Vaikuntanathan (BGV) algorithm. This algorithm supports addition and multiplication operations in an encrypted state. The BGV homomorphic encryption algorithm is used to encrypt collaborative model parameter data, ensuring that the encrypted data can still support the required computational operations.

[0119] Step S32: Writing smart contract code for storing and access controlling the encrypted data of collaborative model parameters, thereby obtaining a smart contract code number;

[0120] This embodiment of the present invention designs a smart contract that defines a data structure for storing encrypted collaborative model parameter data. This can be written in Solidity or other applicable smart contract programming languages. The smart contract code is written to store the encrypted collaborative model parameter data on the blockchain and provide data update functionality. Variables, functions, and events within the contract can be used to manage data storage and updates. Access control logic is added to the smart contract to ensure that only authorized users can read and modify the collaborative model parameter data. Access control can be implemented using permission modifiers or conditional statements.

[0121] Step S33: Deploy the smart contract code data on the blockchain platform to obtain the smart contract address data;

[0122] This embodiment of the present invention selects a blockchain platform that meets the requirements, such as Ethereum or Hyperledger Fabric. Using the tools and commands provided by the blockchain platform, the smart contract code is deployed to the blockchain network. This generates a smart contract address, which uniquely identifies the contract on the blockchain. The smart contract address is recorded for future use. The address can be saved in a database or other data storage.

[0123] Step S34: Generate access key data based on the collaborative model parameter encryption data and the smart contract address data, thereby obtaining access key data;

[0124] This embodiment of the present invention selects a suitable key generation algorithm, such as HMAC or RSA. The encrypted collaborative model parameter data and the smart contract address data are passed as input to the key generation algorithm. The selected key generation algorithm is used to generate an access key. The access key is used to access the smart contract and decrypt the encrypted collaborative model parameter data. The generated access key data is saved for subsequent use. The key can be stored in secure storage.

[0125] Step S35: Enter the access permission according to the access key data, thereby obtaining mapping relationship data between the key and the access permission.

[0126] Embodiments of the present invention determine the definition and level of access rights. For example, different levels of permissions can be defined, such as read, write, execute, etc. Based on the access key data, each key is associated with the corresponding access right. This can be achieved by recording the correspondence between the key and the permission in a database or other data storage. The generated mapping relationship data between the key and the access right is saved. This can be used to subsequently verify the user's access rights and determine the operations that can be performed based on the key. When a user attempts to access collaborative model parameters or execute smart contract operations, the access key is used for permission verification. Based on the mapping relationship data between the key and the access right, it is confirmed whether the user has the requested permission.

[0127] By performing BGV homomorphic encryption on collaborative model parameter data, this invention converts the parameter data into an encrypted form, protecting the confidentiality and security of the model parameters. Encrypted data can be stored, transmitted, and processed without exposing sensitive information, helping to protect privacy and prevent unauthorized access. By writing smart contract code to store and control access to encrypted collaborative model parameter data, a secure data storage and access mechanism is implemented. Smart contract code can define data read and write rules, ensuring that only authorized users or contracts can access and modify data, thereby improving data confidentiality and integrity. By deploying the smart contract code on a blockchain platform, smart contracts can be executed and managed. After deployment, a smart contract is assigned a unique address that identifies its location on the blockchain. The smart contract address data can be used as a public identifier for interacting with and accessing the contract. Authorized access to the encrypted data is achieved by generating an access key based on the encrypted collaborative model parameter data and the smart contract address data. The access key is a secure credential that only users holding the correct key can decrypt and access the encrypted data, ensuring data confidentiality and restricting access rights. Access rights are entered using the access key data to establish a mapping between keys and access rights. This ensures that only users with the appropriate access rights can obtain the corresponding access keys, thereby enabling authorized access management for encrypted data. The mapping relationship between keys and access rights provides fine-grained control and auditing of access control, ensuring data security and compliance.

[0128] Preferably, step S4 includes the following steps:

[0129] Step S41: Using the zk-SNARKs algorithm to perform a zero-knowledge proof algorithm operation based on the mapping relationship data and the smart contract address data, thereby generating zero-knowledge proof data;

[0130] This embodiment of the present invention selects and implements the zk-SNARKs algorithm. zk-SNARKs is a zero-knowledge proof protocol that can prove a statement is true without revealing any information about the statement. Mapping data is prepared, including the mapping relationship between keys and access rights and the smart contract address data. Based on this mapping data and the smart contract address data, the zk-SNARKs algorithm is used to perform a zero-knowledge proof algorithm operation to generate zero-knowledge proof data.

[0131] Step S42: Pack the mapping relationship data and zero-knowledge proof data and upload them to the storage area of ​​the smart contract to obtain on-chain encrypted data;

[0132] This embodiment of the present invention packages mapping relationship data and zero-knowledge proof data for storage and transmission within a smart contract. Data structures (such as JSON) can be used to organize the data and add necessary metadata. The packaged data is uploaded to the smart contract's storage area using the smart contract's upload function. The uploaded data is stored in encrypted form on the blockchain, ensuring data security and privacy.

[0133] Step S43: Analyze the structural characteristics of the encrypted data on the chain to obtain the structural characteristic data of the data on the chain;

[0134] Embodiments of the present invention obtain on-chain encrypted data from smart contracts. The encrypted data can be retrieved using the smart contract's query function. The on-chain encrypted data is analyzed to extract structural features. This may involve decrypting and parsing the encrypted data to understand its organization and content. Useful on-chain data structural features are extracted from the data obtained through structural feature analysis.

[0135] Step S44: Design a zero-knowledge proof protocol framework based on the on-chain data structure feature data, thereby obtaining zero-knowledge proof protocol framework data;

[0136] This embodiment of the present invention designs a suitable zero-knowledge proof protocol framework based on the structural characteristics of on-chain data. This includes determining the protocol's interaction steps, information transmission methods, and proof calculation rules. Based on the designed zero-knowledge proof protocol framework, the corresponding protocol framework data is generated. This may involve defining the protocol's language, constraints, and conventions, as well as generating the relevant protocol code.

[0137] Step S45: embed zero-knowledge characteristics according to the zero-knowledge proof protocol framework data, thereby obtaining proof protocol data with zero-knowledge properties.

[0138] The embodiments of the present invention embed zero-knowledge proof features into the protocol based on the zero-knowledge proof protocol framework data. This may involve improving the information transmission and calculation processes within the protocol to meet the requirements of zero-knowledge proof. Based on the protocol embedded with zero-knowledge features, proof protocol data with zero-knowledge properties is generated. This data will include the proof's calculation steps, interaction process, and related proof information.

[0139] The zk-SNARKs proposed in this invention are a zero-knowledge proof algorithm that can prove the truth of a statement without disclosing sensitive data. Using the zk-SNARKs algorithm, zero-knowledge proof data can be generated based on mapping relationship data and smart contract address data. This proof data can be used to prove that a user has specific access rights without revealing specific access keys or sensitive data, thereby protecting user privacy and data confidentiality. By packaging the mapping relationship data and zero-knowledge proof data and uploading them to the smart contract's storage area, the relevant data can be securely stored on the blockchain. The decentralized and tamper-proof nature of the blockchain ensures data security and trustworthiness. On-chain encrypted data ensures data transparency and verifiability while protecting data confidentiality. By analyzing the structural features of the on-chain encrypted data, structural feature information can be extracted. This on-chain data structural feature data can be used for further analysis and reasoning, helping to understand and reveal patterns, associations, and attributes in the data. This helps discover potential patterns and trends, supporting in-depth data analysis and decision-making. By designing a zero-knowledge proof protocol framework based on the on-chain data structural feature data, a zero-knowledge proof protocol can be constructed for specific data structures. This protocol framework data defines the structure and rules of proofs, ensuring their correctness and security. The zero-knowledge proof protocol framework data provides the foundation and guidance for further generating specific zero-knowledge proofs. By embedding zero-knowledge features within the zero-knowledge proof protocol framework data, proof protocol data with zero-knowledge properties can be generated. This proof protocol data can be used to verify user access rights without leaking sensitive data, achieving the goal of zero-knowledge proofs. Zero-knowledge proof protocol data provides verifiability and privacy protection, providing greater security and trust for data access and authorization.

[0140] Preferably, step S5 includes the following steps:

[0141] Step S51: Using the encrypted data on the chain as input, the proof protocol data is run to generate verification proof data;

[0142] This embodiment of the present invention obtains encrypted on-chain data from a smart contract. Proof protocol data is prepared, including input parameters for the proof protocol and rules for proof calculation. The prepared proof protocol data is fed into the proof protocol using the on-chain encrypted data as input. During execution, the proof protocol performs calculations based on the input data and generates verification proof data.

[0143] Step S52: Perform security verification on the transmitted on-chain encrypted data according to the verification certificate data, thereby obtaining verification result data;

[0144] This embodiment of the present invention prepares the data and algorithms required for the verification process. This includes verification proof data, encrypted data transmitted on-chain, and the corresponding security verification algorithm. Security verification is performed using the verification proof data and the encrypted data transmitted on-chain. The security verification algorithm verifies the proof data and checks its validity. The verification result data indicates the success or failure of the verification.

[0145] Step S53: Use TPM to build a secure isolation area, obtain a decryption key from a trusted third party, and decrypt the encrypted data on the chain in the secure isolation area to obtain the decrypted plaintext data;

[0146] This embodiment of the present invention uses a Trusted Platform Module (TPM) to create a secure, isolated environment for decryption operations. A TPM is a hardware module that provides security and isolation protection. Decryption keys are obtained from a trusted third party. This trusted third party may be an authority or a trusted key management system. Within the secure, isolated area, the encrypted data on the chain is decrypted using the obtained decryption key. The decryption process converts the encrypted result into decrypted plaintext data.

[0147] Step S54: Statistically analyze the decrypted plaintext data and the verification result data, and display them visually to obtain a plaintext analysis report.

[0148] This embodiment of the present invention performs statistical analysis on decrypted plaintext data and verification result data, utilizing various statistical methods and techniques, such as frequency analysis and data clustering, to extract useful information about the data. The results of the statistical analysis are visualized, for example, using charts, graphs, or other visualization tools. This facilitates a better understanding and interpretation of data characteristics and trends. Based on the statistical analysis and visualization results, a plaintext analysis report is generated. The report includes a detailed explanation, conclusions, and recommendations for the decrypted plaintext data and verification result data.

[0149] The present invention uses on-chain encrypted data as input and runs the proof protocol data to generate verification proof data. The verification proof data verifies the legitimacy and integrity of the on-chain encrypted data, ensuring that the data has not been tampered with or damaged during transmission. This increases trust and reliability in the data, ensuring its security and integrity. By performing security verification on the transmitted on-chain encrypted data based on the verification proof data, the authenticity and integrity of the on-chain data can be verified. The verification result data confirms the validity of the on-chain encrypted data, ensuring that the data has not been tampered with or forged. This ensures data integrity, reliability, and verifiability. By utilizing a TPM (Trusted Platform Module) to establish a secure isolation zone, a trusted execution environment is provided to protect the keys and data during the decryption process. Obtaining the decryption key from a trusted third party ensures the key's security and authenticity. Decrypting the on-chain encrypted data within the secure isolation zone yields the decrypted plaintext data, enabling access and use of the encrypted data. Statistical analysis of the decrypted plaintext data and the verification result data can extract useful information and insights, revealing data characteristics, trends, and correlations. Visualizations can intuitively present data analysis results and findings, helping users understand the data and make decisions. Plain text analysis reports can provide in-depth understanding and insights into the data, supporting the promotion and optimization of business and decision-making.

[0150] The present invention also provides a vaccine circulation data encryption system based on data mining, which is used to execute the above-mentioned vaccine circulation data encryption method based on data mining. The vaccine circulation data encryption system based on data mining includes:

[0151] The multi-source data acquisition module is used to collect data from all links in the vaccine circulation process, thereby obtaining multi-source heterogeneous circulation data, including structured circulation data, semi-structured circulation data, and unstructured circulation data; and perform tensor decomposition on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the entire status of vaccine circulation;

[0152] The knowledge construction module is used to build a circulation knowledge graph model based on the formal description of entities, relationships, and events in heterogeneous fused circulation data; capture the temporal dynamic characteristics and long-term dependencies of the heterogeneous fused circulation data and establish a time series analysis model; meta-learn the circulation knowledge graph model and time series analysis model and perform knowledge distillation to obtain collaborative model parameter data;

[0153] The encryption and access control module is used to perform BGV homomorphic encryption on the collaborative model parameter data to obtain the collaborative model parameter encrypted data; write smart contract code for storage and access control of the collaborative model parameter encrypted data, and deploy it on the blockchain platform to obtain the smart contract address data; generate data access keys based on the collaborative model parameter encrypted data and the smart contract address data and enter access rights to obtain the mapping relationship data between the key and access rights;

[0154] The zero-knowledge proof module is used to generate zero-knowledge proof data based on the mapping relationship data and the smart contract address data; the mapping relationship data and the zero-knowledge proof data are packaged and uploaded to the storage area of ​​the smart contract to obtain on-chain encrypted data; based on the structural features of the on-chain encrypted data, a zero-knowledge proof protocol framework is designed and embedded with zero-knowledge characteristics to obtain proof protocol data with zero-knowledge properties;

[0155] The verification and decryption module is used to use the on-chain encrypted data as input to run the proof protocol data, thereby generating verification proof data; based on the verification proof data, the on-chain encrypted data after transmission is securely verified, a secure isolation area is constructed using TPM, and the encrypted data on the chain is decrypted in the secure isolation area to obtain a plaintext analysis report.

[0156] Through multi-source data collection, the present invention can acquire data from all aspects of the vaccine distribution process, including data in various formats and forms. This allows for comprehensive acquisition of distribution data, providing a more comprehensive and accurate information foundation to support subsequent data analysis and decision-making. By performing tensor decomposition on multi-source heterogeneous distribution data, the data can be converted into a higher-dimensional representation, extracting connections and features between the data. This allows for data describing the entire state of vaccine distribution, providing a more comprehensive and accurate data perspective and laying the foundation for subsequent knowledge construction and analysis. By constructing a distribution knowledge graph model, entities, relationships, and events in vaccine distribution can be formally described, establishing knowledge connections and structures. A time series analysis model can be used to capture the temporal dynamics and long-term dependencies of distribution data. This provides an in-depth understanding and analysis of the vaccine distribution process, supporting subsequent decision-making and optimization. By performing BGV homomorphic encryption on collaborative model parameter data, the confidentiality of model parameters can be protected, preventing unauthorized access. The development and deployment of smart contracts enables secure management of the storage and access of encrypted data. This ensures the security and authorized access of model parameters, protecting intellectual property and data privacy. By generating zero-knowledge proof data, data can be verified without revealing its specific content, protecting data privacy. Uploading this zero-knowledge proof data to the smart contract's storage area ensures secure data storage and verifiability. This provides trust and verifiability for on-chain encrypted data, enhancing data security and credibility. Verifying on-chain encrypted data confirms its integrity and authenticity, ensuring it has not been tampered with or forged. Decrypting on-chain encrypted data within a secure isolation zone created using a TPM yields decrypted plaintext data. Analyzing, analyzing, and visualizing this decrypted plaintext data provides in-depth understanding and insight, supporting decision-making and optimization. This ensures comprehensive data security and insight for the vaccine distribution process.

[0157] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0158] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A vaccine circulation data encryption method based on data mining, characterized in that: The following steps are involved: Step S1: Data collection is performed on the links in the vaccine circulation process to obtain multi-source heterogeneous circulation data, including structured circulation data, semi-structured circulation data and unstructured circulation data; Perform tensor decomposition on multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the full status of vaccine circulation; Step S2: Construct a circulation knowledge graph model based on the formal description of entities, relationships, and events in the heterogeneous fused circulation data; capture the temporal dynamic characteristics and long-term dependency relationships based on the heterogeneous fused circulation data, and establish a time series analysis model; Perform meta-learning on the circulation knowledge graph model and time series analysis model, and perform knowledge distillation to obtain collaborative model parameter data; Step S3: Perform BGV homomorphic encryption on the collaborative model parameter data to obtain collaborative model parameter encrypted data; Write smart contract code to store and control access to encrypted data of collaborative model parameters, and deploy it on the blockchain platform to obtain smart contract address data; Generate data access keys based on the collaborative model parameter encryption data and smart contract address data, and enter access rights to obtain the mapping relationship data between keys and access rights; Step S4: Generate zero-knowledge proof data based on the mapping relationship data between the key and the access permission and the smart contract address data; The mapping relationship data between the key and access rights and the zero-knowledge proof data are packaged and uploaded to the storage area of ​​the smart contract to obtain the encrypted data on the chain; Based on the structural features of the on-chain encrypted data, a zero-knowledge proof protocol framework is designed and embedded with zero-knowledge features, thus obtaining proof protocol data with zero-knowledge properties; Step S5: Using the encrypted data on the chain as input, the zero-knowledge proof protocol data is run to generate verification proof data; The encrypted data on the chain after transmission is securely verified based on the verification proof data, a secure isolation area is built using TPM, and the encrypted data on the chain is decrypted in the secure isolation area to obtain a plaintext analysis report.

2. The vaccine circulation data encryption method based on data mining according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: collecting temperature, humidity, and GPS information at different stages of vaccine circulation to obtain structured circulation data; Step S12: using RFID tags and sensor network devices to collect semi-structured circulation data during the vaccine circulation process; Step S13: Collecting pictures, texts, and video information through video surveillance to obtain unstructured circulation data; Step S14: performing data cleaning on the structured circulation data, semi-structured circulation data, and unstructured circulation data, thereby obtaining multi-source heterogeneous circulation data; Step S15: Perform tensor decomposition on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the entire state of vaccine circulation.

3. The vaccine circulation data encryption method based on data mining according to claim 2 is characterized in that: Step S15 includes the following steps: Step S151: integrating the multi-source heterogeneous data of circulation to generate full circulation feature vector data describing a single data entity; Step S152: performing tensor decomposition on the full-circulation feature vector data to obtain latent semantic space data; Step S153: Extract fusion features of the entire vaccine circulation state from the latent semantic space data to obtain heterogeneous fusion circulation data.

4. The vaccine circulation data encryption method based on data mining according to claim 3 is characterized in that: Step S151 includes the following steps: Step S1511: extracting data changes in the time series of the semi-structured circulation data and expressing it in the form of a tensor, thereby obtaining semi-structured vector data; Step S1512: performing pixel value digital representation on the image data in the unstructured circulation data and performing preset range mapping to obtain image vector data; Step S1513: extracting features from the text data in the unstructured circulation data using natural language processing technology to obtain text vector data; Step S1514: extracting frames from the video data in the unstructured circulation data, digitally representing pixel values ​​of the video frames, and performing preset range mapping to obtain video vector data; Step S1515: perform vector merging on the structured circulation data, semi-structured vector data, image vector data, text vector data, and video vector data, thereby generating full circulation feature vector data describing a single data entity.

5. The vaccine circulation data encryption method based on data mining according to claim 4 is characterized in that: Step S1 includes the following steps: Step S21: Identify key entities, relationships between entities, and important events in the vaccine circulation process based on heterogeneous fusion circulation data, thereby obtaining circulation knowledge graph data; Step S22: constructing a circulation knowledge graph model based on the circulation knowledge graph data; Step S23: performing time series characteristic analysis on the heterogeneous fused circulation data to obtain circulation time characteristic data; Step S24: extracting long-term dependency relationships from the circulation time feature data to obtain dependency relationship data; Step S25: establishing a time series analysis model based on the circulation time characteristic data and the dependency relationship data; Step S26: Meta-learn the circulation knowledge graph model and the time series analysis model, and perform knowledge distillation to obtain collaborative model parameter data.

6. The vaccine circulation data encryption method based on data mining according to claim 5 is characterized in that: Step S26 includes the following steps: Step S261: Structural encoding is performed on the circulation knowledge graph model, and node structural features are extracted to obtain node structural feature data; Step S262: Using meta-path encoding technology to perform vector space representation learning on the time series analysis model, thereby obtaining time series feature embedding data; Step S263: performing relationship learning processing on the node structure feature data and the temporal feature embedding data based on the attention mechanism, thereby obtaining an attention relationship matrix; Step S264: performing attention relationship connection on the circulation knowledge graph model and the time series analysis model according to the attention relationship matrix, thereby obtaining a knowledge graph-time series full model; Step S265: Randomly sample knowledge graph nodes and time series steps in the knowledge graph-time series full model to obtain sample index data; Step S266: Calculate the information received by each node and step according to the knowledge graph-time series full model, and update its own representation according to the sampling index data to obtain the collaborative model parameter data.

7. The vaccine circulation data encryption method based on data mining according to claim 6 is characterized in that: Step S3 includes the following steps: Step S31: performing BGV homomorphic encryption on the collaborative model parameter data to obtain collaborative model parameter encrypted data; Step S32: Writing smart contract code for storing and access controlling the encrypted data of collaborative model parameters, thereby obtaining a smart contract code number; Step S33: Deploy the smart contract code data on the blockchain platform to obtain the smart contract address data; Step S34: Generate access key data based on the collaborative model parameter encryption data and the smart contract address data, thereby obtaining access key data; Step S35: Enter the access permission according to the access key data, thereby obtaining mapping relationship data between the key and the access permission.

8. The vaccine circulation data encryption method based on data mining according to claim 7 is characterized in that: Step S4 includes the following steps: Step S41: Using the zk-SNARKs algorithm, a zero-knowledge proof algorithm is performed based on the mapping relationship data between the key and the access permission and the smart contract address data, thereby generating zero-knowledge proof data; Step S42: Packaging the mapping relationship data between the key and access rights and the zero-knowledge proof data, and uploading them to the storage area of ​​the smart contract, thereby obtaining on-chain encrypted data; Step S43: Analyze the structural characteristics of the encrypted data on the chain to obtain the structural characteristic data of the data on the chain; Step S44: Design a zero-knowledge proof protocol framework based on the on-chain data structure feature data, thereby obtaining zero-knowledge proof protocol framework data; Step S45: embed zero-knowledge characteristics according to the zero-knowledge proof protocol framework data, thereby obtaining proof protocol data with zero-knowledge properties.

9. The vaccine circulation data encryption method based on data mining according to claim 8 is characterized in that: Step S5 includes the following steps: Step S51: Using the on-chain encrypted data as input, the zero-knowledge proof protocol data is run to generate verification proof data. Step S52: Perform security verification on the transmitted on-chain encrypted data according to the verification certificate data, thereby obtaining verification result data; Step S53: Use TPM to build a secure isolation area, obtain a decryption key from a trusted third party, and decrypt the encrypted data on the chain in the secure isolation area to obtain the decrypted plaintext data; Step S54: Statistically analyze the decrypted plaintext data and the verification result data, and display them visually to obtain a plaintext analysis report.

10. A vaccine circulation data encryption system based on data mining, characterized in that: For executing the vaccine circulation data encryption method based on data mining according to claim 1, the vaccine circulation data encryption system based on data mining comprises: The multi-source data acquisition module is used to collect data from all links in the vaccine circulation process, thereby obtaining multi-source heterogeneous circulation data, including structured circulation data, semi-structured circulation data, and unstructured circulation data; and perform tensor decomposition on the multi-source heterogeneous circulation data to obtain heterogeneous fused circulation data that describes the entire status of vaccine circulation; The knowledge construction module is used to construct a circulation knowledge graph model based on the formal description of entities, relationships, and events in heterogeneous fused circulation data; capture the temporal dynamic characteristics and long-term dependencies of the heterogeneous fused circulation data and establish a time series analysis model; meta-learn the circulation knowledge graph model and time series analysis model and perform knowledge distillation to obtain collaborative model parameter data; The encryption and access control module is used to perform BGV homomorphic encryption on the collaborative model parameter data to obtain the collaborative model parameter encrypted data; write smart contract code for storage and access control of the collaborative model parameter encrypted data, and deploy it on the blockchain platform to obtain the smart contract address data; generate data access keys based on the collaborative model parameter encrypted data and the smart contract address data and enter access rights to obtain the mapping relationship data between the key and access rights; The zero-knowledge proof module is used to generate zero-knowledge proof data based on the mapping relationship data between keys and access rights and the smart contract address data; the mapping relationship data between keys and access rights and the zero-knowledge proof data are packaged and uploaded to the storage area of ​​the smart contract to obtain on-chain encrypted data; based on the structural features of the on-chain encrypted data, a zero-knowledge proof protocol framework is designed and embedded with zero-knowledge characteristics to obtain proof protocol data with zero-knowledge properties; The verification and decryption module is used to use the on-chain encrypted data as input to run the zero-knowledge proof protocol data, thereby generating verification proof data; based on the verification proof data, the on-chain encrypted data after transmission is securely verified, a secure isolation area is constructed using the TPM, and the encrypted results of the on-chain encrypted data are decrypted in the secure isolation area to obtain a plaintext analysis report.

Images