CN113972012A - Infectious disease prevention and control cooperative system based on alliance chain and public chain - Google Patents

Infectious disease prevention and control cooperative system based on alliance chain and public chain Download PDF

Info

Publication number
CN113972012A
CN113972012A CN202111251762.5A CN202111251762A CN113972012A CN 113972012 A CN113972012 A CN 113972012A CN 202111251762 A CN202111251762 A CN 202111251762A CN 113972012 A CN113972012 A CN 113972012A
Authority
CN
China
Prior art keywords
data
infectious disease
chain
training
public chain
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111251762.5A
Other languages
Chinese (zh)
Inventor
宋明宇
章永龙
李斌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Yangzhou University
Original Assignee
Yangzhou University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Yangzhou University filed Critical Yangzhou University
Priority to CN202111251762.5A priority Critical patent/CN113972012A/en
Publication of CN113972012A publication Critical patent/CN113972012A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/80—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for detecting, monitoring or modelling epidemics or pandemics, e.g. flu
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/60—Protecting data
    • G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
    • G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
    • G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Biomedical Technology (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Public Health (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Medical Informatics (AREA)
  • Bioethics (AREA)
  • Pathology (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Computer Hardware Design (AREA)
  • Computer Security & Cryptography (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

The invention discloses an infectious disease prevention and control cooperative system based on a alliance chain and a public chain, which is used for constructing an alliance chain network and a public chain network, wherein communities, enterprises and scientific research institutions are accessed to the public chain network, a virtual institution consisting of a prevention and control center, a medical institution and the public chain is accessed to the alliance chain network, each institution on the alliance chain correspondingly serves as a local client node for federal learning, an infectious disease risk assessment model is trained in a local data center, parameters obtained by training are uploaded to a federal service terminal to perform safety parameter aggregation by using a federal learning method, and a final global infectious disease risk assessment model is obtained after iteration of federal learning. The invention utilizes the block chain as a basic technical framework, can ensure the authenticity of data and effectively stimulate multiple parties to cooperate and share real data on the basis of fully ensuring the safety, privacy, rights and interests of the data, and finally realizes the evaluation of the risk level of the infectious diseases.

Description

Infectious disease prevention and control cooperative system based on alliance chain and public chain
Technical Field
The invention relates to the technical field of block chains, in particular to the field of infectious disease prevention and control, and constructs an infectious disease prevention and control cooperative system based on a alliance chain and a public chain.
Background
Currently, the data stored in a data center of a single medical institution is extremely limited, and more infectious disease data is collected and shared by each institution. In the process of cooperatively preventing and controlling infectious diseases, related institutions tend to visit related medical data of the disease states of more patients, but the medical data of the patients relate to various privacy, the privacy of personal information of the patients is involved, the competition among various organizations is also concerned, the organizations worry that the data sharing may improve the service quality of competitors, the competitiveness of the competitors is further improved, the organizations are sensitive topics, the sharing of the medical data of various organizations is realized on the basis of protecting the personal sensitive data of the patients, the uniformity of the accuracy and the timeliness of infectious disease monitoring is achieved, and the problem to be solved urgently is solved.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the problem that information data sharing of medical treatment, infectious diseases and the like cannot be realized by medical institutions, disease prevention and control centers, infectious disease research centers and the like due to a vertical reporting mode of a traditional infectious disease reporting system, the invention provides an infectious disease prevention and control cooperative system based on a alliance chain and a public chain. Model parameters of the uploaded medical data of each organization data center are encrypted, transmitted and fused through a federal learning method. Finally, the data privacy of patients of all institutions is guaranteed, and meanwhile, the prediction and evaluation of the infectious disease risk level are achieved.
The technical scheme is as follows: in order to achieve the purpose, the invention adopts the following technical scheme:
a communicable disease prevention and control cooperative system based on a alliance chain and a public chain comprises an alliance chain network and a public chain network, wherein users accessed to the public chain network comprise communities, enterprises and scientific research institutions, the users accessed to the alliance chain network comprise a virtual institution consisting of a prevention and control center, a medical institution and the public chain, and each institution in the alliance chain is correspondingly used as a local client node for federal learning; infectious disease monitoring reports from communities, enterprises and scientific research institutions are recorded on the public chain, screening and classification are carried out, and reward punishment related users are traced according to preset rules;
the method comprises the steps that medical data of each node mechanism on a alliance chain are subjected to infectious disease risk assessment model training in a local data center, parameters obtained through training are uploaded to a federal server side to conduct safety parameter aggregation by using a federal learning method, the federal server side returns the aggregated parameters to each node mechanism, and the node mechanisms receive the parameters and update respective model parameters of the node mechanisms by using decrypted gradients to conduct training; and completing model training after a plurality of iterative processes of federal learning to obtain a final global infectious disease risk assessment model, equivalently obtaining the model obtained by carrying out calculation after concentrating infectious disease data of each organization on the alliance chain.
Furthermore, each node on the alliance chain comprises a medical institution, a prevention and control center and a virtual institution consisting of a public chain, each node is provided with a corresponding data center, and the data center stores privacy medical data of each institution; each node on the federation chain is equivalent to a local client for federated learning and participates in the iterative process of federated learning; the data center participates in the local training, encryption and decryption process of the local client side of the federal study.
Furthermore, an intelligent contract exciting component and an intelligent contract monitoring component which are made by an organization party and are used for reward and punishment measures and business logic conversion are issued on the public chain network; the intelligent contract monitoring component is used for judging, screening and classifying infectious disease monitoring reports uploaded by each node in the public chain, wherein input data of the intelligent contract monitoring component is structured decision data with a standard format, and judgment fusion is carried out at a semantic level or weighted fusion is carried out at a data level during processing; the intelligent contract incentive component is used for carrying out economic incentive on users reporting real and effective infectious disease monitoring data and carrying out economic punishment on the users reporting false infectious disease monitoring data.
Furthermore, the intelligent contract monitoring component comprises a decision selection algorithm and a decision fusion algorithm, wherein the decision selection algorithm is used for setting a decision screening rule and filtering invalid or malicious decisions; the decision fusion algorithm is used for setting a decision fusion rule to obtain a consensus decision.
Further, virtual currency is distributed in public chain networks, economic incentives are issued and circulated in the form of blockchain currency, sources including wrong decision penalties, malicious user penalties, and government regular incentives.
Further, the specific process of federal learning fusion includes:
each node on the alliance chain is regarded as a local client for federal learning, and local training is carried out in a data center; before local training, feature extraction and fusion are carried out on medical data including infectious disease pathology, medical resources and cases stored in a data center;
the data center carries out local training, updates parameters of a local model, extracts training parameters, encrypts and sends the parameters to the federal server;
the federal service side carries out aggregation on parameters of all the clients, records the set of the randomly selected and updated clients as S, and for the client x belonging to S, t is the model updating turn, and the updated parameters of the client x are
Figure BDA0003318944730000021
Performing security aggregation on all clients at the federated server to update wt+1，
Figure BDA0003318944730000022
Wherein m isxIs the sample size of client x, m is the total number of training samples,
Figure BDA0003318944730000031
updating the result of the parameters of the client x in the t +1 round;
the federal service terminal aggregates the result wt+1And returning to the data center of each local client, and updating and training the respective model by using the obtained parameters by the local client.
Furthermore, the federal server does not store any data, only serves as a trusted third party to assist calculation, and after training is completed, the updated parameter wt+1The participating parties respectively hold the summarized operation results.
As a possible implementation manner, the manner of performing feature extraction and fusion on the medical data is as follows: defining a medical resource and pathological feature set, expressing the medical resource and case features used by the case by using vectors, and finally expressing the fusion features of the case as the sum of the products of the medical resource and case features used by the case and the hyperparameters.
As another possible implementation, the method for extracting and fusing the features of the medical data includes: introducing a deep learning model to respectively extract the characteristics of different types of medical data: extracting characteristics of pathological data of a case by adopting a DeepFM model; for medical resource data with case relevance, extracting relevance characteristics among the data by using a node2vec model; for case data with a time sequence, an LSTM model is adopted to depict the evolution characteristics of the data; and finally fusing the extracted different features through a feature fusion function.
Further, the infectious disease risk assessment model enables the fused features to pass through multiple MLPs and softmax layers to obtain an infectious disease risk grade score.
Has the advantages that: by adopting the technical scheme, the invention utilizes the alliance chain as a basic technical framework and cooperates with an incentive mechanism on the public chain to establish an infectious disease prevention and control cooperative system on the basis of fully ensuring data security, privacy, rights and interests and the like, the system not only ensures the uniformity of the accuracy and timeliness of infectious disease monitoring, but also utilizes an infectious disease risk assessment model based on multi-dimensional feature fusion to predict the risk level of infectious diseases, solves the problem of privacy data of each organization in infectious disease data sharing, and can effectively stimulate each organization to cooperate and share while ensuring the authenticity of the infectious disease data by adopting the incentive mechanism in the public chain.
Drawings
FIG. 1 is a schematic diagram of the structure of an infectious disease prevention and control system based on a alliance chain and a public chain;
FIG. 2 is a multi-dimensional feature fusion-based model for risk assessment of infectious diseases according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1, an embodiment of the present invention provides an infectious disease prevention and control collaboration system based on a alliance chain and an alliance chain, which mainly includes an alliance chain network and a public chain network, where users accessing the public chain network include communities, enterprises, scientific research institutions, and the like, users accessing the alliance chain network include prevention and control centers, medical institutions, and the like, and also include virtual institutions formed by the public chain. Infectious disease monitoring reports from communities, enterprises, scientific research institutions and the like are recorded on the public chain, screening and classification are carried out, and reward punishment related users are traced according to preset rules. Each organization (a control center, a medical organization, a virtual organization and the like) on the federation chain correspondingly serves as a local client node for federated learning, the infectious disease risk assessment model training is carried out in a local data center, parameters obtained by training are uploaded to a federated server to carry out safety parameter aggregation by using a federated learning method, and a final global infectious disease risk assessment model is obtained after federated learning iteration. The system construction and operation of the embodiment of the invention mainly comprise the following steps:
step 1, a system builder jointly deploys a alliance chain network with a plurality of organizations, and each organization on the chain is correspondingly used as a local client node for federal learning;
step 2, forming a public chain network by communities, enterprises, scientific research institutions and the like, recording infectious disease monitoring reports from the society, the enterprises and the scientific research institutions through an intelligent contract monitoring component on the chain, screening and classifying the infectious disease monitoring reports, submitting all verification and evaluation results to an intelligent contract credit incentive component, and tracing and punishing related users by the component according to preset rules, wherein the public chain network is equivalent to a virtual mechanism and is finally connected to a alliance chain and is a node of the alliance chain;
and 3, performing infectious disease risk assessment model training on the medical data of each node institution in the alliance chain in a local data center, wherein the training comprises extracting case samples related to infectious diseases, calculating required characteristics, performing characteristic fusion, classification assessment and the like. Uploading the parameters obtained by training to a federal server side by using a federal learning method to perform safety parameter aggregation, returning the aggregated parameters to each node mechanism by the federal server side, and updating the respective model parameters of the node mechanisms by using decrypted gradients after the node mechanisms receive the parameters to perform training;
and 4, iterating the step 3, iterating the process for multiple times until the loss function is converged, obtaining a final global model, and finishing the model training process. The product is substantially equivalent to a model obtained by concentrating infectious disease data of each organization in the alliance chain and then performing calculation.
In step 1, the construction of the alliance link network comprises the following specific processes:
step 1-1, each node on the alliance chain comprises a medical institution, a prevention and control center, a virtual institution and the like consisting of a public chain, each node is provided with a corresponding data center, and the data center stores private medical data of each institution;
step 1-2, each node on the alliance chain is equivalent to a local client for federal learning in step 3 and participates in the iterative process of federal learning in step 4;
the data center participates in the local training, encryption and decryption process of the local client side of the federal study.
In step 2, the specific process of the excitation mechanism of the public link network of the virtual mechanism includes:
and 2-1, each organization such as a community, an enterprise, a scientific research institution and the like accesses the existing medical data management system into a public chain sharing system, which is different from various organizations in the prior alliance chain, and the organizations in the alliance chain have access thresholds and higher authority and can be used for establishing mutual trust among all large organizations. The public chain is commonly maintained by countless mechanisms which are not known mutually, and the mechanisms on the public chain have no admission thresholds, so that the public chain has original tokens which are used for exciting everyone to participate in sharing;
step 2-2, an organization side formulates incentive punishment measures and business logic, converts the incentive punishment measures and business logic into an intelligent contract incentive component and an intelligent contract monitoring component, and issues the intelligent contract incentive component and the intelligent contract monitoring component to a public chain network;
and 2-3, uploading infectious disease monitoring reports of all nodes in the public chain to an intelligent contract monitoring component, and judging, screening and classifying the infectious disease monitoring reports, wherein input data of the intelligent contract monitoring component is structured decision data with a standard format, and discrimination fusion can be carried out at a semantic level or weighting fusion can be carried out at a data level during processing.
Specifically, the intelligent contract monitoring component generally comprises two key algorithms, namely a decision selection algorithm and a decision fusion algorithm, the decision selection algorithm sets a decision screening rule, screening indexes such as credit of a submitter, deposit of the submitter, algorithm quality, data quality and the like can be combined in the decision screening rule, and invalid or malicious decisions in massive decisions are quickly filtered; the decision fusion algorithm sets a decision fusion rule, and the decision fusion rule sets fusion weight of algorithm decision according to factors such as evidence-raising score, prediction precision, quality score and the like so as to obtain accurate consensus decision. The accuracy and timeliness of disease monitoring can be considered by adjusting the fusion weight of the two. All verification and evaluation results are submitted to an intelligent contract credit incentive component, and the component traces back the reward and punishment related users according to a preset rule;
step 2-4, for the users reporting the real and effective infectious disease monitoring data, the economic incentive can be obtained by adopting reward measures through an intelligent contract incentive component, and on the contrary, for the clients reporting the false infectious disease monitoring data, the corresponding economic penalty can be obtained;
where a virtual currency is published over a public link network, economic incentives are issued and circulated in the form of blockchain currency, and sources may include false decision penalties, malicious user penalties, and government periodic incentives. The user or institution uses the acquired reward virtual currency consuming institution provided health services or other services.
And 3, in the step 3, the federal learning fusion comprises the following specific processes:
step 3-1, regarding each node on the alliance chain as a local client for federal learning, and performing local training in a data center; before local training, a preprocessing process needs to be carried out on medical data, the medical data stored in a data center comprises attributes such as infectious disease pathological characteristics, medical resources and cases, wherein: the pathological characteristics mainly describe the symptoms of infectious diseases, such as pulmonary tuberculosis with symptoms of cough, fever, diarrhea and the like; the medical resources comprise resource items such as various medicines and the like; the case includes personal information such as age, sex, home address, etc. of the patient. In addition, there is a certain correlation between pathological features, medical resources, and attributes of cases, and it is necessary to extract and fuse features of these data.
Two ways of feature fusion are proposed here, the first being the original fusion:
assuming that t types of medical resources are shared in a medical institution, a medical resource set R ═ { R ═ R is defined1，r2，...，rt}; there are p cases, the case set is C ═ C1，c2，...，cp}; extracting characteristic words from medical records as pathological characteristics, and defining the pathological characteristic set as F ═ F under the assumption that q pathological characteristic words are shared in medical institutions in a certain area1，f2，...，fq}. For example, if case a uses r1，r2Two medical resources having f1，f3Two pathological features, then the medical resource used by it can be expressed as Ra1, 1, 0, 0, … …, 0, the pathological features of which can be denoted as FaThe sample of case a is denoted as c, 1, 0, … …, 0a，
cx＝μRa+λFa
Wherein λ and μ are respectively defined as the hyperparameters describing the medical resources used in case a of infectious disease and the pathological characteristics of infectious disease, RaAnd FaThe collection of medical resources and pathological features used by the infectious disease case a, respectively.
Due to RaAnd FaD-max { t, q } after fusion, wherein the missing vector is supplemented with a null vector.
Secondly, the second mode is to introduce the existing deep learning model to perform feature fusion on various neural networks. The structure of the proposed model is shown in fig. 2, the extracted features are different for different types of medical data, and the model adopts different deep learning models or algorithms, so that the overall features of infectious diseases are more accurately, more comprehensively and more generally depicted, and the accurate infectious disease risk assessment is realized.
Firstly, respectively extracting the characteristics of different types of medical data:
1) for pathological data of symptoms such as cough, fever and diarrhea, a DeepFM model is adopted, the model comprises an FM (factor mechanisms) part and a DNN (Deep Neural Networks) part, and low-order and high-order characteristics of the pathological data can be extracted simultaneously.
2) Medical resource data with relevance such as medicines and equipment are subjected to relevance depiction by using a node2vec model, and relevance features among the data are extracted.
3) For case data having a time series including onset, LSTM (Long Short-Term Memory) was used
The model characterizes the evolution of the data.
Then, the extracted different features are fused through a feature fusion function (such as vector splicing, vector addition, vector inner product and the like).
The fused features are passed through multiple layers of MLP (Multi-layer Perceptin) and sofimax layers to obtain infectious disease risk grade scores.
The model training process is as follows:
dividing the data samples into batches with the size of B, wherein the total number of local updates is E, i is one turn of the local updates, and when each epoch i belongs to (1, E), performing the following parameter training:
Figure BDA0003318944730000071
w is a local model parameter, where the model parameter w includes model parameters fused by the above-mentioned feature-extracted models such as deep fm, node2vec, LSTM, etc., η is a learning rate, h is one batch in the total number B of batches, and l () is a loss function, where the parameter w of the local client is updated by performing a local SGD random gradient descent method in batches.
And 3-2, extracting training parameters from the local data center, and encrypting the result by adopting an encryption method, wherein the encryption method comprises homomorphic encryption, differential privacy or secret sharing technology to cover gradient and the like.
Step 3-3, sending the encrypted result to a federal server, wherein all parameters execute safety aggregation at the federal server without learning information about any local client;
the parameters of all the clients are aggregated at the federal server, and the aggregation process is as follows:
first, initialize the parameter server parameter w0(ii) a Secondly, recording the set of the randomly selected and updated clients as S, regarding the client x belonging to S, t as the model updating turn, and the updated parameter of the x client is
Figure BDA0003318944730000072
Figure BDA0003318944730000073
Wherein, the client update function represents the function of local update of the client parameter in step 3-1.
Finally, all clients are executed on the federal service endRow security aggregation to update wt+1，
Figure BDA0003318944730000074
Wherein m isxIs the sample size of client x, m is the total number of training samples,
Figure BDA0003318944730000075
the result is updated for the client x parameters in the t +1 round.
Step 3-4, the federal service terminal aggregates the result wt+1Returning to the data center of each local client, and updating and training respective models by using the obtained parameters by the local clients;
wherein, the clients participating in the training communicate and summarize the operation results. The federal service end does not store any data, only serves as a trusted third party to assist calculation, and after training is completed, the updated parameter wt+1The calculation results held and summarized by each party are actually the adjustment amount of the parameters after the multiple parties summarize. And after the adjustment quantity is obtained, the adjustment quantity is sent back to the participant to help the participant update the parameter w of the model, and each parameter of the model is dynamically adjusted according to the data of the participant, so that the model can obtain better accuracy.
In the step 4, the iterative model comprises the following specific processes:
and 4-1, iterating the step 3, and dynamically adjusting each parameter of the model according to the infectious disease data to enable the model to obtain better accuracy.
And 4-2, after the federal learning training is finished, all participants share the model parameters, and finally, the result of infectious disease monitoring fusion is obtained.
The final product is substantially equivalent to a model obtained by collecting medical data of multiple infectious diseases and then performing calculation.
The invention utilizes the alliance chain as a basic technical framework and cooperates with an incentive mechanism on the public chain to establish an infectious disease prevention and control cooperative system on the basis of fully ensuring data security, privacy, rights and interests and the like, the system not only ensures the uniformity of the accuracy and timeliness of infectious disease monitoring, but also utilizes an infectious disease risk assessment model based on multi-dimensional feature fusion to predict the risk level of infectious diseases, solves the problem of privacy data of each organization in infectious disease data sharing, and adopts the incentive mechanism in the public chain, so that each organization can be effectively encouraged to cooperate and share while ensuring the authenticity of the infectious disease data.
The foregoing shows and describes the basic principles, principal steps and advantages of the present invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that various changes and modifications may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims (10)

1. A communicable disease prevention and control cooperative system based on a alliance chain and a public chain is characterized by comprising an alliance chain network and a public chain network, wherein users accessed to the public chain network comprise communities, enterprises and scientific research institutions; infectious disease monitoring reports from communities, enterprises and scientific research institutions are recorded on the public chain, screening and classification are carried out, and reward punishment related users are traced according to preset rules;
the method comprises the steps that medical data of each node mechanism on a alliance chain are subjected to infectious disease risk assessment model training in a local data center, parameters obtained through training are uploaded to a federal server side to conduct safety parameter aggregation by using a federal learning method, the federal server side returns the aggregated parameters to each node mechanism, and the node mechanisms receive the parameters and update respective model parameters of the node mechanisms by using decrypted gradients to conduct training; and completing model training after a plurality of iterative processes of federal learning to obtain a final global infectious disease risk assessment model, equivalently obtaining the model obtained by carrying out calculation after concentrating infectious disease data of each organization on the alliance chain.
2. An infectious disease prevention and control cooperative system based on a alliance chain and a public chain as claimed in claim 1, wherein each node in the alliance chain comprises a medical institution, a prevention and control center, and a virtual institution composed of the public chain, each node has a corresponding data center, and the data center stores private medical data of each institution; each node on the federation chain is equivalent to a local client for federated learning and participates in the iterative process of federated learning; the data center participates in the local training, encryption and decryption process of the local client side of the federal study.
3. An infectious disease control and coordination system based on a alliance chain and a public chain as claimed in claim 1, wherein the public chain network issues an intelligent contract incentive component and an intelligent contract monitoring component which are manufactured by reward and punishment measures and business logic conversion established by an organization party; the intelligent contract monitoring component is used for judging, screening and classifying infectious disease monitoring reports uploaded by each node in the public chain, wherein input data of the intelligent contract monitoring component is structured decision data with a standard format, and judgment fusion is carried out at a semantic level or weighted fusion is carried out at a data level during processing; the intelligent contract incentive component is used for carrying out economic incentive on users reporting real and effective infectious disease monitoring data and carrying out economic punishment on the users reporting false infectious disease monitoring data.
4. An infectious disease prevention and control cooperative system based on a alliance chain and a public chain as claimed in claim 3, wherein the intelligent contract monitoring component comprises a decision selection algorithm and a decision fusion algorithm, the decision selection algorithm is used for setting decision screening rules and filtering invalid or malicious decisions; the decision fusion algorithm is used for setting a decision fusion rule to obtain a consensus decision.
5. An infection control and coordination system according to claim 3, wherein virtual currency is distributed in public chain network, economic incentives are distributed and circulated in the form of block chain general certificates, and the sources include wrong decision penalties, malicious user penalties and government regular incentives.
6. A federation chain and public chain-based infectious disease control and coordination system according to claim 1, wherein the specific process of federal learning fusion includes:
each node on the alliance chain is regarded as a local client for federal learning, and local training is carried out in a data center; before local training, feature extraction and fusion are carried out on medical data including infectious disease pathology, medical resources and cases stored in a data center;
the data center carries out local training, updates parameters of a local model, extracts training parameters, encrypts and sends the parameters to the federal server;
the federal service side carries out aggregation on parameters of all the clients, records the set of the randomly selected and updated clients as S, and for the client x belonging to S, t is the model updating turn, and the updated parameters of the client x are
Figure FDA0003318944720000021
Performing security aggregation on all clients at the federated server to update wt+1，
Figure FDA0003318944720000022
Wherein m isxIs the sample size of client x, m is the total number of training samples,
Figure FDA0003318944720000023
updating the result of the parameters of the client x in the t +1 round;
the federal service terminal aggregates the result wt+1And returning to the data center of each local client, and updating and training the respective model by using the obtained parameters by the local client.
7. An epidemic prevention and control coordination system based on alliance chain and public chain as claimed in claim 6 wherein the federal service end does not save any data, but only acts as a trusted third party, assists in computation, and after training is completed, updates parameter wt+1The participating parties respectively hold the summarized operation results.
8. An infectious disease prevention and control cooperative system based on a alliance chain and a public chain as claimed in claim 6, wherein the medical data is extracted and fused by the following means: defining a medical resource and pathological feature set, expressing the medical resource and case features used by the case by using vectors, and finally expressing the fusion features of the case as the sum of the products of the medical resource and case features used by the case and the hyperparameters.
9. An infectious disease prevention and control cooperative system based on a alliance chain and a public chain as claimed in claim 6, wherein the medical data is extracted and fused by the following means: introducing a deep learning model to respectively extract the characteristics of different types of medical data: extracting characteristics of pathological data of a case by adopting a DeepFM model; for medical resource data with case relevance, extracting relevance characteristics among the data by using a node2vec model; for case data with a time sequence, an LSTM model is adopted to depict the evolution characteristics of the data; and finally fusing the extracted different features through a feature fusion function.
10. An infection prevention and control cooperative system based on alliance chain and public chain as claimed in claim 6 wherein the infection risk assessment model passes the fused features through multiple layers of MLP and softmax layer to obtain the infection risk grade score.
CN202111251762.5A 2021-10-25 2021-10-25 Infectious disease prevention and control cooperative system based on alliance chain and public chain Pending CN113972012A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111251762.5A CN113972012A (en) 2021-10-25 2021-10-25 Infectious disease prevention and control cooperative system based on alliance chain and public chain

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111251762.5A CN113972012A (en) 2021-10-25 2021-10-25 Infectious disease prevention and control cooperative system based on alliance chain and public chain

Publications (1)

Publication Number Publication Date
CN113972012A true CN113972012A (en) 2022-01-25

Family

ID=79588732

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111251762.5A Pending CN113972012A (en) 2021-10-25 2021-10-25 Infectious disease prevention and control cooperative system based on alliance chain and public chain

Country Status (1)

Country Link
CN (1) CN113972012A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114564731A (en) * 2022-02-28 2022-05-31 大连理工大学 Intelligent wind power plant wind condition prediction method based on transverse federal learning
CN115622777A (en) * 2022-10-09 2023-01-17 郑州大学 Multi-center federal learning data sharing method based on alliance chain
CN117094420A (en) * 2023-10-20 2023-11-21 浙江大学 Model training method, device, power prediction method, equipment and medium

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114564731A (en) * 2022-02-28 2022-05-31 大连理工大学 Intelligent wind power plant wind condition prediction method based on transverse federal learning
CN114564731B (en) * 2022-02-28 2024-06-04 大连理工大学 Intelligent wind power plant wind condition prediction method based on transverse federal learning
CN115622777A (en) * 2022-10-09 2023-01-17 郑州大学 Multi-center federal learning data sharing method based on alliance chain
CN117094420A (en) * 2023-10-20 2023-11-21 浙江大学 Model training method, device, power prediction method, equipment and medium
CN117094420B (en) * 2023-10-20 2024-02-06 浙江大学 Model training method, device, power prediction method, equipment and medium

Similar Documents

Publication Publication Date Title
CN114417421B (en) Shared information privacy protection method and related device based on meta universe
CN111866869A (en) Federal learning indoor positioning privacy protection method facing edge calculation
CN112101403B (en) Classification method and system based on federal few-sample network model and electronic equipment
CN113972012A (en) Infectious disease prevention and control cooperative system based on alliance chain and public chain
CN112101404B (en) Image classification method and system based on generation countermeasure network and electronic equipment
Li et al. Contract-theoretic pricing for security deposits in sharded blockchain with Internet of Things (IoT)
Zhong et al. Privacy-protected blockchain system
Malik et al. Building a secure platform for digital governance interoperability and data exchange using blockchain and deep learning-based frameworks
Bandara et al. Skunk—A blockchain and zero trust security enabled federated learning platform for 5G/6G network slicing
Kanamori et al. Privacy-preserving federated learning for detecting fraudulent financial transactions in japanese banks
CN114844652A (en) Cloud authentication service system based on block chain and big data mining method
Lv et al. Research on modeling of e-banking fraud account identification based on federated learning
Bandara et al. Bassa-ml—a blockchain and model card integrated federated learning provenance platform
Wang et al. What if properties are owned by no one or everyone? foundation of blockchain enabled engineered ownership
Ali et al. Incentive-driven federated learning and associated security challenges: A systematic review
CN101242410B (en) Grid subjective trust processing method based on simple object access protocol
Ali et al. A systematic review of federated learning incentive mechanisms and associated security challenges
CN112085051B (en) Image classification method and system based on weighted voting and electronic equipment
CN114723012B (en) Calculation method and device based on distributed training system
Alenizi et al. Enhancing secure financial transactions through the synergy of blockchain and artificial intelligence
Belej et al. Ensuring accuracy of personal data processed in blockchain systems
Mosqueda González et al. Secure and efficient decentralized machine learning through group-based model aggregation
Yang et al. Federated Deep Learning Algorithm For Target Recognition Based On Trusted Network
Pentland et al. 10. Towards an Ecosystem of Trusted Data and AI
CN118300855A (en) Credit data security management system based on cloud service

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination