Blockchain-based rough set data classification training method and system

Abstract

The application provides a rough set data classification training method and system based on a blockchain, relates to the technical field of the blockchain, and can be applied to the financial field and other fields.The method comprises the following steps: constructing a data matrix by means of a coordination coupling algorithm according to user data of a blockchain client; obtaining a sample data set by selecting the data matrix, constructing an interference data set according to the sample data set and a preset Gaussian noise; constructing a fuzzy decision system according to the interference data set, extracting an attribute reduction core of the fuzzy decision system with respect to data attributes by means of data attribute classification training; and performing attribute reduction on the interference data set by means of the attribute reduction core to obtain training data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of blockchains, and can be applied to the financial field and other fields, and particularly relates to a rough set data classification training method and system based on a blockchain. BACKGROUND

[0002] With the vigorous development of Internet big data training and recommendation applications, data training gradually presents the characteristics of being centered on user activities, locally open, decentralized and privacy risks. In order to reduce the risk of privacy leakage, in 2006, C. Dwork proposed differential privacy protection (privacy-preserving), which is a new type of information privacy protection technology. This technology hides the real data distribution of the data set by mixing random noise in the data set, and provides strong privacy protection. Therefore, when facing attackers, no matter how many times the query is submitted and how many different query combinations are arranged, or whether the attacked person is in the data set, the returned query results are not much different.

[0003] The privacy protection degree generated by differential privacy technology is often at the expense of reducing the accuracy and availability of real data, and a trade-off needs to be made between user privacy protection and data availability. Therefore, in the machine learning of the recommendation system, whether the influence of noisy data can be reduced and the training error caused by ambiguous preference data can be reduced has always been a difficulty and hotspot in the industry. At present, the k-nearest neighbor method based on content analysis is widely used to build user recommendation systems, which has the characteristics of simple application and high computational efficiency, but has poor scalability in complex scenarios with many influencing factors, and cannot meet the needs of combining auxiliary information such as implicit feedback, temporary effect or user confidence index in the recommendation system. Collaborative filtering technology solves the problem that is difficult to understand in the previous content analysis process by analyzing the association between users and products to identify new user-item relationships. A typical method is to use a matrix decomposition model to increase the activity data of users in other fields as auxiliary information to guide the realization of the correspondence between users and interest items. This method has good recommendation accuracy and scalability, but since the analysis data uses widely distributed user interest data as the basis for analysis, there are difficulties and difficulties such as difficulty in multi-party collaboration, risk of user privacy leakage, etc. Blockchain is a new technology that is decentralized, tamper-proof and traceable, and provides a good solution to decentralized privacy protection and data sharing. In the blockchain network, users participate in the construction of a multi-transaction body through transactions, and multiple transaction bodies serve as different authenticators to jointly witness transaction behavior, such as online transactions, where users consume and shop at different online merchants. Their activity data can be further decomposed into user interest items and aggregated together in a hidden manner, which not only has the characteristic of hiding the user's real identity, but also increases the extension data items and random noise, effectively solving the problem of balancing accuracy and privacy in decentralized data. Fuzzy rough set theory is a mathematical theory for handling inconsistencies in numerical data proposed by D. Dubois and H. Prade in 1990. The introduction of fuzzy rough set identification is used to calculate attribute reduction, thereby achieving the effect of reducing noise and fast classification. SUMMARY

[0004] The present application aims to provide a rough set data classification training method and system based on blockchain, which utilizes the characteristics of decentralized multi-party information sharing of blockchain to promote the safe multi-party sharing of user data, introduces the attribute reduction kernel of fuzzy rough set to reduce noise attributes, thereby assisting in the noise reduction of differential privacy data, effectively classifying user interest items, while avoiding attackers from stealing user privacy through combined queries or association algorithms, and better balancing user privacy security and recommendation accuracy.

[0005] To achieve the above purpose, the rough set data classification training method based on the blockchain provided by the present application specifically comprises: constructing a data matrix according to user data of a blockchain client through a coordination coupling algorithm; obtaining a sample data set through the data matrix, and constructing an interference data set according to the sample data set and a preset Gaussian noise; constructing a fuzzy decision system according to the interference data set, and extracting an attribute reduction core of the fuzzy decision system with respect to a data attribute through data attribute extraction of classification training; and performing attribute reduction on the interference data set by using the attribute reduction core to obtain training data.

[0006] In the rough set data classification training method based on the blockchain, optionally, constructing the fuzzy decision system according to the interference data set comprises: multiple blockchain clients consensus the constructed interference data set to a blockchain consensus accounting node, and the blockchain consensus accounting node constructs a fuzzy decision system according to the received interference data set.

[0007] In the rough set data classification training method based on the blockchain, optionally, constructing the fuzzy decision system according to the interference data set further comprises: performing cyclic iteration calculation on the interference data set according to a number of physical addresses of the blockchain client in the blockchain and a preset iteration number to obtain a shared decomposition factor gradient corresponding to the iteration number; and updating the shared decomposition factor gradient to other consensus accounting nodes of the blockchain through consensus.

[0008] In the rough set data classification training method based on the blockchain, optionally, updating the shared decomposition factor gradient to other consensus accounting nodes of the blockchain through consensus comprises: performing standardization smoothing clipping on the shared decomposition factor gradient to obtain a gradient parameter satisfying the Lyapunov condition; calculating a to-be-updated shared decomposition factor according to the gradient parameter; and updating the to-be-updated shared decomposition factor to other consensus accounting nodes of the blockchain through consensus.

[0009] In the rough set data classification training method based on the blockchain, optionally, constructing the interference data set according to the sample data set and the Gaussian noise comprises: generating Gaussian noise through a differential privacy algorithm and a binomial distribution function; and adding the Gaussian noise to the sample data set to construct the interference data set.

[0010] In the rough set data classification training method based on the blockchain, optionally, extracting the attribute reduction core of the fuzzy decision system with respect to the data attribute through data attribute extraction of classification training comprises: performing attribute relative reduction on the fuzzy decision system according to a pre-defined attribute necessity to obtain a reduction result; and obtaining the attribute reduction core of the fuzzy decision system with respect to the data attribute according to the reduction result and the data attribute.

[0011] In the above-mentioned rough set data classification training method based on blockchain, optionally, the attribute relative reduction of the fuzzy decision system according to the predefined attribute necessity to obtain the reduction result includes: obtaining multiple decision attributes according to the predefined attribute necessity; and calculating the reduction result according to the similarity relationship between the decision attributes and the fuzzy decision system.

[0012] This application also provides a blockchain-based rough set data classification training system, the system comprising a blockchain client and a blockchain consensus ledger node; the blockchain client is used to construct a data matrix based on user data of the blockchain client through a coordinated coupling algorithm, and to select a sample dataset through the data matrix; and to construct an interference dataset based on the sample dataset and preset Gaussian noise; the blockchain consensus ledger node is used to construct a fuzzy decision system based on the interference dataset; and to extract the attribute reduction kernel of the fuzzy decision system relative to the data attributes through the data attributes of the classification training, and to use the attribute reduction kernel to perform attribute reduction on the interference dataset to obtain training data.

[0013] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method.

[0014] This application also provides a computer-readable storage medium storing a computer program that performs the above-described methods.

[0015] This application also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described method.

[0016] The beneficial technical effects of this application are as follows: Leveraging the decentralized, multi-party information sharing characteristics of blockchain, it promotes secure multi-party sharing of user data while introducing attribute reduction operators based on fuzzy rough sets to reduce noisy attributes. This assists in denoising differential privacy data, effectively classifying user interest items, and preventing attackers from stealing user privacy through combinatorial queries or association algorithms, thus achieving a good balance between user privacy and recommendation accuracy. It does not rely on the participation of a third-party impartial institution, supports the anonymized sharing and aggregation of user data from multiple parties in the consortium, and provides a multi-party shared user dataset for machine learning in recommendation systems. By introducing a relation kernel setting, unnecessary attribute relationships are reduced, achieving user privacy anonymization while minimizing the impact on machine learning accuracy. Randomly sampling user dataset samples during the training process achieves interest item coupling and user privacy shielding, balancing recognition accuracy and efficiency. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of the application and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings:

[0018] Figure 1 A flowchart of a rough set data classification training method based on a blockchain provided by an embodiment of the application;

[0019] Figure 2 An architecture diagram of a blockchain network provided by an embodiment of the application;

[0020] Figure 3 A structure diagram of a blockchain client provided by an embodiment of the application;

[0021] Figure 4 A construction flowchart of a fuzzy decision system provided by an embodiment of the application;

[0022] Figure 5 A flowchart of attribute reduction core acquisition provided by an embodiment of the application;

[0023] Figure 6 A structure diagram of a blockchain consensus ledger node provided by an embodiment of the application;

[0024] Figure 7 A flowchart of attribute reduction in a rough set data classification training method provided by an embodiment of the application;

[0025] Figure 8 A flowchart of a rough set data classification training system processing provided by an embodiment of the application;

[0026] Figure 9 A structure diagram of an electronic device provided by an embodiment of the application. DETAILED DESCRIPTION

[0027] The embodiments of the present application will be described in detail below with reference to the drawings and embodiments, so that how the technical means of the present application is applied to solve the technical problems and achieve the technical effects can be fully understood and implemented. It should be noted that, unless there is a conflict, each embodiment in the present application and each feature in each embodiment can be combined with each other, and the technical solutions formed thereby are within the protection scope of the present application.

[0028] In addition, the steps shown in the flowchart of the drawings can be executed in a computer system such as a group of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0029] Reference should be made to Figure 1As shown, the rough set data classification training method based on the blockchain provided in the application specifically includes:

[0030] S101 constructing a data matrix through a coordination coupling algorithm according to user data of a blockchain client;

[0031] S102 obtaining a sample data set through the data matrix, and constructing an interference data set according to the sample data set and a preset Gaussian noise;

[0032] S103 constructing a fuzzy decision system according to the interference data set, and extracting an attribute reduction core of the fuzzy decision system with respect to a data attribute through data attribute extraction of classification training;

[0033] S104 performing attribute reduction on the interference data set by using the attribute reduction core to obtain training data.

[0034] In this embodiment, the provider of the user data can be multiple trusted subjects, and the user generates transaction data through a client and saves the transaction data in the trusted subjects, which are responsible for the authenticity of the transaction data and the reliability of the access object. The construction of the interference data set according to the sample data set and the Gaussian noise includes: generating Gaussian noise through a differential privacy algorithm and a binomial distribution function; and adding the Gaussian noise to the sample data set to construct the interference data set. The construction logic of the interference data set is mainly as follows. Let a collaborative filtering matrix M be represented as:

[0035] M=(m ij ) n×m ;

[0036] Wherein, i (i∈[1,n]) identifies an entity user identifier, j (j∈[1,m]) identifies an interest item, m ij identifies the interest degree of the user i in the interest item j. M is a sparse matrix, and the goal of the recommendation algorithm is to predict the interest degree of the user with a blank interest item.

[0037] Let M be decomposed into n (n≤N) data sets M (n) dispersed in physical locations. Each data set is approximately represented as:

[0038] M (n) ≈U (n) ×V (n) ;

[0039] Wherein, and are matrix vectors of M (n) (rank D).

[0040] For any two data sets M and Z, the similarity can be calculated by Hamming distance, given a random algorithm A, differential privacy formula (e, d)-DP, satisfies:

[0041] P(A(M) e S) < e ∈ P(A(Z) e S) + d;

[0042] Wherein, S identifies the value domain of random algorithm A, d is selected from binomial distribution function N(0, 2k t ), k t is the test step.

[0043] The sample data set is selected from M, and the Gaussian noise d (t) is added to generate the interference data set Z (n,t) , that is:

[0044] Z (n,t) ≈ U (n,t) V + d (t) ;

[0045] Suppose Z is an interference data set, A is a non-empty real attribute set, and (Z, A) is called a fuzzy information system. If the attributes in A are divided into condition attributes and decision attributes, the fuzzy information system is called a fuzzy decision system.

[0046] Then, based on the fuzzy decision system, subsequent reduction is performed to obtain training data. The specific reduction process will be described in detail in subsequent embodiments, and will not be described one by one here.

[0047] In an embodiment of the present application, constructing a fuzzy decision system according to the interference data set comprises: a plurality of blockchain clients consensus the constructed interference data set to a blockchain consensus accounting node, and the blockchain consensus accounting node constructs a fuzzy decision system according to the received interference data set.

[0048] Specifically, please refer to Figure 2As shown, in actual work, the overall blockchain network includes blockchain consensus ledger nodes 1, CA / digital identity authentication nodes 2, blockchain infrastructure cloud 3, and blockchain clients 4. The blockchain consensus ledger node 1 is a virtual computing node in the infrastructure cloud and is a basic component of the blockchain network. The blockchain network has a plurality of blockchain consensus ledger nodes, which are responsible for blockchain transaction access and processing, provide smart contract execution, transaction consensus and transaction ledger, and also provide data encryption, decryption and other security services for transaction data. The CA / identity authentication node 2 is responsible for digital identity authentication of the blockchain consensus ledger node 1, certificate distribution of transaction participants (agencies), provides digital identity security encryption public algorithm and security parameters, and is responsible for identity verification and other security services. The blockchain infrastructure cloud 3 is responsible for providing network resources, computing resources and storage resource allocation according to the networking resource needs of users, providing blockchain networking services, supporting selection of blockchain bottom layer product images according to the product form and requirements of users, configuring virtual node resources, and creating a blockchain network. The blockchain client 4 is responsible for receiving transaction requests of users accessing the blockchain, forwarding to the corresponding blockchain consensus ledger node, and receiving the smart contract execution of the blockchain consensus ledger node 1 and returning to the user. The blockchain clients are generally distributed in different physical addresses, and the physical addresses of the two blockchain clients in this example are site1 and site2.

[0049] Among them, please refer to Figure 3 As shown, the blockchain client can include a communication module 21, a ciphertext processing module 22, a secure computing module 23, and a data obfuscation storage module 24. The communication module 21 is responsible for establishing a secure channel for the blockchain terminal node and realizing the transmission and reception of user data messages of the blockchain terminal node. The ciphertext processing module 22 is responsible for encrypting and decrypting the data messages generated by the user. The secure computing module 23 is responsible for executing parsing and splitting of user data, forming a data vector and assembling a user data message. The data obfuscation storage module 24 is responsible for adding noise to user data to generate a fuzzy data set Z, which is defined as follows:

[0050] The similarity of any two data sets M and Z can be calculated by the Hamming distance. Given a random algorithm A and a differential privacy formula (e, d)-DP, it satisfies:

[0051] P(A(M) e S) < e ∈ P(A(Z) e S) + d;

[0052] Where S represents the value range of the random algorithm A. d is selected from the binomial distribution function N(0, 2k t ), and K t is the test step size.

[0053] Selecting a sample data set from M, adding Gaussian noise to generate an interference data set Z (n,t) , namely:

[0054] Z (n,t) ≈U (n,t) V+δ (t) ;

[0055] Let Z be an interference data set, A be a non-empty real-valued attribute set, and (Z, A) be called a fuzzy information system. If the attributes in A are divided into condition attributes and decision attributes, the fuzzy information system is called a fuzzy decision system.

[0056] Please refer to Figure 4 , in an embodiment of the present application, constructing a fuzzy decision system according to the interference data set further includes:

[0057] S201 performs a corresponding number of iteration calculations on the interference data set according to the number of physical addresses of the blockchain client in the blockchain and the preset number of iterations to obtain a shared decomposition factor gradient;

[0058] S202 updates the shared decomposition factor gradient to other consensus ledger nodes of the blockchain through consensus. Wherein, updating the shared decomposition factor gradient to other consensus ledger nodes of the blockchain through consensus includes: performing standardization smoothing clipping on the shared decomposition factor gradient to obtain a gradient parameter that satisfies the Lyapunov condition; calculating a to-be-updated shared decomposition factor according to the gradient parameter, and updating the to-be-updated shared decomposition factor to other consensus ledger nodes of the blockchain through consensus.

[0059] Please refer to Figure 5 , in an embodiment of the present application, the attribute reduction kernel of the fuzzy decision system relative to the data attribute extracted by the classified training data attribute includes:

[0060] S301 performs attribute relative reduction on the fuzzy decision system according to the pre-defined attribute necessity to obtain a reduction result;

[0061] S302 obtains the attribute reduction kernel of the fuzzy decision system relative to the data attribute according to the reduction result and the data attribute.

[0062] Wherein, performing attribute relative reduction on the fuzzy decision system according to the pre-defined attribute necessity to obtain a reduction result includes: obtaining a plurality of decision attributes according to the pre-defined attribute necessity; and calculating a reduction result according to the similarity relationship between the decision attributes and the fuzzy decision system.

[0063] For details, please refer to Figure 6As shown, in actual work, the blockchain consensus ledger node can include a communication module 31, a ciphertext processing module 32, a transaction processing module 33, and a rough set attribute reduction module 34:

[0064] The communication module 31 is responsible for communication between nodes of the blockchain, and completes general blockchain node communication information transmission, including transaction broadcast information, consensus information, block synchronization information, network state information, etc.

[0065] The ciphertext processing module 32 is responsible for encryption and decryption processing of data packets received by the communication module 31.

[0066] The transaction processing module 33 is responsible for generating a transaction unique identifier, a contract unique identifier, and a call parameter assembly into a transaction packet according to a transaction request provided by the ciphertext processing module 32, and calling a smart contract transaction to calculate a result. It mainly includes three steps: 1. Calculate the shared factor gradient, 2. Clip the gradient, and 3. Update the shared decomposition factor.

[0067] 1. Calculate the shared decomposition factor gradient

[0068] The shared decomposition factor V is updated in each iteration, and its updated value is called the gradient ΔV. If there are n (n = {1,...,N}) physical addresses, the gradient of each iteration t (t≤T) is represented as ΔV (n,t) , and the calculation formula is as follows:

[0069]

[0070] 2. Clip the gradient

[0071] The gradient is standardized and smoothed to ensure that the calculated gradient satisfies the Lipschitz condition, as follows:

[0072]

[0073] 3. Update the shared decomposition factor

[0074] Update the shared decomposition factor as follows:

[0075]

[0076] where is the decomposition factor step size, and γ (n,t) is a random noise matrix.

[0077] The rough set attribute reduction module 34 is responsible for rough set attribute reduction on the received random noise matrix, mainly including three steps: 1. Attribute necessity definition, 2. Attribute relative reduction, and 3. Relationship core determination. The steps are as follows:

[0078] 1. Attribute necessity definition

[0079] Let (Z, R U D) be a fuzzy decision system, Z (n,t) ≈U (n,t) V+δ (t) be a nonempty finite universe, R be a family of fuzzy T M -similarity relations derived from the condition attributes, and D be a fuzzy T M -similarity relation derived from the decision attributes. Let U / D = {D i : i = 1,..., n} be the complete fuzzy equivalence partition of D, and let:

[0080] Sim(R) = ∩{γ: γ ∈ R}

[0081] Then it is easy to prove that Sim(R) (still a T M -similarity relation) is formally:

[0082] Pos sim(R) = U Di∈D / U Sim(R) * (Di)

[0083] is called the positive region of the decision attributes D with respect to the condition attribute set R.

[0084] For γ ∈ R, if Pos sim(R) = Pos {R-{γ}} D, then γ is called unnecessary in R with respect to D, otherwise, γ is called necessary in R with respect to D. The set of all necessary elements in R with respect to D is called the core of R with respect to D, denoted by:

[0085] Core D (R) = ∩Red D (R)

[0086] where Red D (R) is the set of all reductions of R with respect to D.

[0087] 2. Attribute relative reduction

[0088] Let the granular structure of a fuzzy rough set be: where,

[0089]

[0090] Obviously, for (x λ ) R , (y λ ) R , if (x λ ) R ≠ (y λ ) R , then (x λ ) R∩(x λ ) R = φ.

[0091] Let P contains a relative reduction of R if and only if P satisfies for x∈U, λ=Sim(R)*([x] D )(x), has holds.

[0092] 3. Determination of relation core

[0093] Let the relation core of R be denoted as Core D (R), formally the following relation holds:

[0094] Core D (R)={R: c ij ={R}},1≦i,j≤n

[0095] P contains a relative reduction of R if and only if for c ij ≠ φ, P∩c ij ≠ φ.

[0096] Let R={R1,...,R l}, Red D (R)={R1,...,R l}.

[0097] For the convenience of more clearly understanding the specific implementation process of attribute reduction in the rough set data classification training method based on the blockchain provided in the present application, the above embodiment is further described with reference to the following Figure 7 :

[0098] S401: input (Z, R∪D) as a fuzzy decision system, Z^((n,t))≈U^((n,t))V+δ^((t)) is a non-empty finite domain, R is a family of fuzzy T_M-similar relations obtained by conditional attributes, D is a fuzzy T_M-similar relation obtained by decision attributes, and D is classified into U / D={D_i: i=1,...,n} by all fuzzy equivalence classes;

[0099] S402: let M_R(D) be the set of minimal elements of D, initialize M_R(D) as: M_R(D)= φ; let Red_D(R) be a relative reduction set, initialize Red_D(R) as: Red_D(R)= φ;

[0100] S403: calculate each DIS({a}) and DIS(R). Wherein, DIS({a}) represents a discernibility relation with respect to the decision attribute D, let DIS(R)=∪_(a∈R)DIS({a}); let:

[0101]

[0102] S404: For each element pair {x_i, x_j} e U*U of DIS(R), the following process is performed in a loop;

[0103] S405: Calculate s_ij and c_ij, where c_ij = {a e A: (x_i, x_j) e DIS({a})};

[0104] S406: Update the minimal element set M_R(D) and DIS(R), i.e.:

[0105] M_R(D) = M_R(D) U c_ij;

[0106] DIS(R) = DIS(R) - s_ij;

[0107] S407: Calculate the attribute core Core(D). The set of all necessary elements in R with respect to D is called the core of R with respect to D, denoted as:

[0108] Core_D(R) = n Red_D(R)

[0109] where Red_D(R) is the set of all reductions of R with respect to D.

[0110] Calculate the candidate attribute set: PS = U {c_ij: c_ij e A};

[0111] S408: If the set PS is not empty, i.e. PS ≠ φ;

[0112] S409: Select each element Cand of PS, i.e. Cand e D;

[0113] S410: If the intersection of Cand and c_ij is not empty, i.e. Cand n c_ij ≠ φ;

[0114] S411: Then RED = Core_R(D) U Cand;

[0115] S412: If c_ij e R, exclude Cand from PS, i.e. Cand n c_ij = φ;

[0116] S413: Output the reduction set RED.

[0117] The application also provides a rough set data classification training system based on a blockchain, which comprises a blockchain client and a blockchain consensus ledger node; the blockchain client is used to construct a data matrix by a coordination coupling algorithm according to user data of the blockchain client, select a sample data set through the data matrix, and construct an interference data set according to the sample data set and a preset Gaussian noise; the blockchain consensus ledger node is used to construct a fuzzy decision system according to the interference data set, extract an attribute reduction core of the fuzzy decision system with respect to a data attribute through data attribute extraction of classification training, and perform attribute reduction on the interference data set by using the attribute reduction core to obtain training data.

[0118] For details, please refer to Figure 8 As shown in the actual work, the rough set data classification training system based on the blockchain mainly realizes fuzzy data set generation.

[0119] S501: The blockchain client located at site1 initializes user data to generate a data matrix X(1);

[0120] S502: The user preference information is arranged to form a data matrix U(1), and a global data set V is set, which is formally X(1)=U(1)V;

[0121] S503: A sample data set is selected from X(1), and Gaussian noise is added to generate an interference data set Z (n,t) , that is,

[0122] Z (n,t) (1)≈U (n,t) (1)V+δ (t)

[0123] S504: The blockchain client located at site2 initializes user data to generate a data matrix X(2);

[0124] S505: The user preference information is arranged to form a data matrix U(2), and a global data set V is set, which is formally X(2)=U(2)V;

[0125] S506: A sample data set is selected from X(2), and Gaussian noise is added to generate an interference data set Z (n,t) , that is,

[0126] Z (n,t) (2)≈U (n,t) (2)V+δ (t)

[0127] A sample data set is selected from M, and Gaussian noise is added to generate an interference data set Z (n,t) , that is,

[0128] Z(n,t) ≈U (n,t) V+δ (t)

[0129] S507: Select digital samples and set detection step size;

[0130] S508: Start loop processing according to iteration parameter T, and stop and exit if the number of iterations is greater than T; start loop processing according to the number of physical addresses N, and stop and exit if the number of physical addresses is greater than N; that is, WHILE n < N AND t < T;

[0131] S509: Select digital samples and set detection step size. Start loop processing for each matrix element in Z, and stop and exit if the matrix element traversal is completed; calculate and update the privacy parameter. Let the user data set of physical address n and iteration number t be U (n,t) , and the update rule is:

[0132]

[0133] wherein,

[0134]

[0135] The following three processing steps are executed:

[0136] (1) Calculate the shared decomposition factor gradient

[0137] The shared decomposition factor V is updated in each iteration, and the updated value is called the gradient ΔV. If there are n (n = {1,..., N}) physical addresses, the gradient of each iteration t (t ≤ T) is represented as ΔV (n,t) , and the calculation formula is as follows:

[0138]

[0139] (2) Clip the gradient

[0140] Standardize and smooth clip the gradient to ensure that the calculated gradient satisfies the Lipschitz condition, as follows:

[0141]

[0142] (3) Update the shared decomposition factor

[0143] Update the shared decomposition factor as follows:

[0144]

[0145] wherein, is the decomposition factor step size, and γ (n,t) is a random noise matrix.

[0146] output update result U t and V.

[0147] S510: consensus ledger storage V.

[0148] The beneficial technical effects of the present application are that: by utilizing the characteristics of decentralized multi-party information sharing of the blockchain, the safe multi-party sharing of user data is promoted, the attribute reduction operator of fuzzy rough set is introduced to reduce noise attributes, thereby assisting in realizing the noise reduction of differential privacy data, effectively classifying user interest items, while avoiding attackers from stealing user privacy through combined queries or association algorithms, and better balancing the privacy security and recommendation accuracy of users. Not dependent on the participation of third-party fair institutions, supporting multi-party user data desensitization sharing and aggregation, providing multi-party shared user data sets for recommendation system machine learning; by introducing relationship kernel setting, reducing unnecessary attribute relationships, while realizing user privacy desensitization, reducing the impact on machine learning accuracy; randomly extracting user data set samples in the training process, realizing interest item coupling and user privacy shielding, balancing identification accuracy and efficiency.

[0149] The present application also provides an electronic device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the above method when executing the computer program.

[0150] The present application also provides a computer readable storage medium, which stores a computer program for executing the above method.

[0151] The present application also provides a computer program product, comprising computer programs / instructions, which are executed by a processor to implement the steps of the above method.

[0152] As shown in Figure 9 , the electronic device 600 can also include a communication module 110, an input unit 120, an audio processing unit 130, a display 160, and a power supply 170. It is worth noting that the electronic device 600 does not necessarily include all the components shown in Figure 9 ; in addition, the electronic device 600 can also include components not shown in Figure 9 , which can refer to prior art.

[0153] As shown in Figure 9 , the central processor 100, also known as a controller or operation control, can include a microprocessor or other processor device and / or logic device, which receives input and controls the operation of each component of the electronic device 600.

[0154] The memory 140, for example, can be one or more of a buffer, a flash memory, a hard drive, a removable media, a volatile memory, a non-volatile memory, or other suitable device. The above-mentioned information related to failure can be stored, and in addition, a program for executing the information related to failure can be stored. The central processing unit 100 can execute the program stored in the memory 140 to achieve information storage or processing, etc.

[0155] The input unit 120 provides input to the central processing unit 100. The input unit 120 is, for example, a key or a touch input device. The power supply 170 is used to supply power to the electronic device 600. The display 160 is used to display display objects such as images and characters. The display can be, for example, an LCD display, but is not limited thereto.

[0156] The memory 140 can be a solid state memory such as a read only memory (ROM), a random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and is provided with more data, and examples of such a memory are sometimes referred to as an EPROM, etc. The memory 140 can also be some other type of device. The memory 140 includes a buffer memory 141 (sometimes referred to as a buffer). The memory 140 can include an application / function storage section 142 for storing application programs and function programs or a flow for executing the operation of the electronic device 600 by the central processing unit 100.

[0157] The memory 140 can also include a data storage section (data 143) for storing data such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. A driver storage section (driver 144) of the memory 140 can include various drivers of the electronic device for a communication function and / or for performing other functions of the electronic device such as a messaging application, a contact application, etc.

[0158] The communication module 110 is a transmitter / receiver 110 that transmits and receives signals via an antenna 111. The communication module (transmitter / receiver) 110 is coupled to the central processing unit 100 to provide input signals and receive output signals, which can be the same as in the case of a conventional mobile communication terminal.

[0159] Based on different communication technologies, multiple communication modules 110, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module, can be provided in the same electronic device. The communication module (transmitter / receiver) 110 is also coupled to a speaker 131 and a microphone 132 via an audio processor 130 to provide audio output via the speaker 131 and to receive audio input from the microphone 132 to enable typical telecommunication functions. The audio processor 130 can include any suitable buffers, decoders, amplifiers, etc. In addition, the audio processor 130 is coupled to the central processor 100 to enable recording on-board via the microphone 132 and to enable playing on-board stored sounds via the speaker 131.

[0160] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, a system or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer-readable program code.

[0161] The present application is described in terms of exemplary embodiments, systems, and computer program products, in flowcharts and / or block diagrams. It will be understood that individual elements of flowcharts and / or block diagrams, and combinations of elements in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowcharts and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart or block diagram block or blocks.

[0162] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart or block diagram block or blocks.

[0163] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are generated to realize the computer-implemented processes in the computer or other programmable devices, and the instructions executed in the computer or other programmable devices provide the processes for implementing the functions specified in the flowchart Figure 1 one flow or a plurality of flows and / or the functions specified in the block Figure 1 Figure 1 one block or a plurality of blocks.

[0164] The above-described specific embodiments have further detailed the purposes, technical solutions and beneficial effects of the present application. It should be understood that the above-described is only a specific embodiment of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A rough set data classification training method based on blockchain, characterized in that, The method includes: A data matrix is ​​constructed based on user data from the blockchain client using a coordinated coupling algorithm; A sample dataset is obtained by selecting from the data matrix, and an interference dataset is constructed based on the sample dataset and preset Gaussian noise. The smart contract is invoked through the blockchain consensus ledger node to construct a fuzzy decision system based on the interference dataset. The fuzzy decision system is then subjected to relative attribute reduction based on the predefined attribute necessity to obtain the reduction result. Based on the reduction results and the data attributes trained through classification, the attribute reduction kernel of the fuzzy decision system relative to the data attributes is obtained; The smart contract is invoked through the blockchain consensus ledger node, and the attribute reduction kernel is used to perform attribute reduction on the interference dataset to obtain training data. The construction of a fuzzy decision system based on the interference dataset includes: Multiple blockchain clients will reach a consensus on the interference dataset they have constructed and transmit it to the blockchain consensus ledger node, which will then construct a fuzzy decision system based on the received interference dataset.

2. The blockchain-based rough set data classification training method according to claim 1, characterized in that, Constructing a fuzzy decision system based on the aforementioned interference dataset also includes: The smart contract is invoked through the blockchain consensus ledger node, and the shared decomposition factor gradient is obtained by iteratively calculating the interference dataset according to the number of physical addresses of blockchain clients in the blockchain and the preset number of iterations. The shared decomposition factor gradient is updated to other consensus ledger nodes in the blockchain through consensus.

3. The blockchain-based rough set data classification training method according to claim 2, characterized in that, Updating the shared factor gradient to other consensus ledger nodes in the blockchain via consensus includes: The gradient of the shared decomposition factor is standardized and smoothed to obtain gradient parameters that satisfy the Lyapunov condition; The shared decomposition factor to be updated is calculated based on the gradient parameters, and then updated to other consensus ledger nodes of the blockchain through consensus.

4. The blockchain-based rough set data classification training method according to claim 1, characterized in that, The interference dataset constructed based on the sample dataset and Gaussian noise includes: Gaussian noise is generated using a differential privacy algorithm and a binomial distribution function; The Gaussian noise is added to the sample dataset to construct an interference dataset.

5. The blockchain-based rough set data classification training method according to claim 4, characterized in that, The fuzzy decision system is subjected to attribute relative reduction based on predefined attribute necessity to obtain the reduction result, which includes: Multiple decision attributes are obtained based on predefined attribute necessity; The reduction result is calculated based on the similarity relationship between the decision attribute and the fuzzy decision system.

6. A rough set data classification training system based on blockchain, characterized in that, The system includes a blockchain client and a blockchain consensus ledger node; The blockchain client is used to construct a data matrix based on user data of the blockchain client using a coordinated coupling algorithm, select a sample dataset using the data matrix, and construct an interference dataset based on the sample dataset and preset Gaussian noise. The blockchain consensus ledger node is used to invoke a smart contract, obtain the interference dataset constructed and agreed upon by multiple blockchain clients, construct a fuzzy decision system based on the received interference dataset, and construct a fuzzy decision system based on the interference dataset, and perform attribute relative reduction on the fuzzy decision system according to predefined attribute necessity to obtain a reduction result. Based on the reduction result and the data attributes trained by classification, the attribute reduction kernel of the fuzzy decision system relative to the data attributes is obtained, and the attribute reduction kernel is used to perform attribute reduction on the interference dataset to obtain training data.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that enables a computer to execute the method of any one of claims 1 to 5.

Images