Block chain user transaction object recommendation method, system and device

Abstract

The invention discloses a block chain user transaction object recommendation method, system and device. The method comprises the steps of obtaining on-chain data of candidate transaction objects in a block chain network and obtaining feature vectors through feature extraction; clustering the preprocessed feature vectors based on a Gaussian mixture model so as to obtain similarity scores of the users; further, based on labeled historical transaction object data, a user risk model is obtained through training, and reasoning is performed through feature vectors to obtain a risk value score of a user; and constructing a weighted comprehensive scoring function, sorting and screening the candidate transaction objects in combination with the user similarity score and the user risk value score, generating a comprehensive recommendation table, and feeding back the comprehensive recommendation table. According to the method, the security and accuracy of block chain user transaction object recommendation can be effectively improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of machine learning and blockchain, and particularly relates to a blockchain user transaction object recommendation method, system and device. BACKGROUND

[0002] In recent years, with the continuous expansion of the on-chain user scale, the transaction behavior is increasingly complex, and users face problems such as information asymmetry, difficulty in establishing trust, and insufficient identification of potential risk nodes when selecting appropriate transaction objects. The credit scoring method has a lag in responding to abnormal behaviors of high-reputation accounts; the method based on social relationships is ineffective in the anonymous address scenario and is vulnerable to Sybil attacks; the collaborative filtering method has the problems of cold start and data sparsity, and ignores transaction security; the rule or blacklist filtering can only identify known high-risk objects and lacks the ability to predict unknown risks. SUMMARY

[0003] The present application provides a blockchain user transaction object recommendation method, system and device to solve the problems in the prior art.

[0004] To solve the above technical problems, the present application solves the problems by the following technical solutions: Obtain the on-chain data of the blockchain network transaction object and perform feature extraction and preprocessing to obtain a feature vector; Pre-set an initial clustering number, obtain the model parameters of the corresponding Gaussian model, combine the Bayesian information criterion to obtain the corresponding overall profile coefficient, and then obtain the group label of the user, and perform similarity calculation based on the group label to obtain the similarity score of the user; Construct a user risk pre-training model, train the user risk pre-training model through historical transaction object data, obtain the user risk model, and perform inference through the feature vector to obtain the risk value score of the user; Input the similarity score of the user and the risk value score of the user into a weighted scoring function to obtain a comprehensive recommendation table; Obtain the actual transaction rate of the comprehensive recommendation table and feed it back to the on-chain data.

[0005] As an implementable manner, the obtaining of the on-chain data of the blockchain network transaction object and the feature extraction and preprocessing to obtain the feature vector comprises the following steps: Use the data transparency feature of the public chain to obtain the on-chain data through the blockchain node or the third-party blockchain data interface, wherein the on-chain data includes continuous feature data and discrete feature data; Fill the continuous feature data with the mean interpolation method and perform Z-Score standardization to obtain a continuous feature vector; The discrete feature data is filled with similar samples and one-hot encoded to obtain a discrete feature vector; The continuous feature vector and the discrete feature vector are aggregated and preprocessed to obtain a feature vector, wherein the preprocessing includes one or more of differential cleaning, format unification, missing value processing, and vectorization encoding.

[0006] As an implementable manner, the similarity score of the user is obtained by the following steps: The initial parameters of the Gaussian distribution are calculated according to the initial number of clusters, and the initial parameters of the Gaussian distribution are iteratively updated to obtain the log-likelihood value of the Gaussian distribution model; A difference threshold is preset, and the difference between the comparison results of adjacent log-likelihood values is made to be less than the difference threshold; The model parameters are calculated by the Gaussian model of each initial cluster number, and the BIC value is obtained based on the Bayesian information criterion and the model parameters; The n initial cluster numbers corresponding to the minimum value of the BIC value are selected as candidate cluster numbers, and the overall silhouette coefficient under each cluster number is calculated by the candidate cluster number; The final cluster number is obtained according to the overall silhouette coefficient, and the similarity score of the user is obtained based on the final cluster number and the feature vector.

[0007] As an implementable manner, the similarity score of the user is obtained based on the final cluster number and the feature vector, including the following steps: The posterior probability of the user belonging to each cluster is obtained based on the final cluster number and the feature vector, and the group label of the current user is obtained based on the cluster label corresponding to the posterior probability; According to the same group label of the user, the first similarity between the target user and the same group center is obtained, and the second similarity between the target user and the candidate transaction object is calculated; The first similarity and the second similarity are weighted and fused to obtain the similarity score of the blockchain user.

[0008] As an implementable manner, the user risk pre-training model is constructed, the user risk pre-training model is trained by the historical transaction object data, the user risk model is obtained, and the risk value score of the user is obtained by the inference of the feature vector, including the following steps: The user is divided into normal users and tainted users, a user threshold is preset, and the feature vector is divided and calculated to obtain a first probability that the normal user is greater than the user threshold and a second probability that the tainted user is greater than the user threshold; According to the eigenvector, the first probability and the second probability, a posteriori probability is calculated through a Bayesian formula, a user risk pre-training model is trained, and a user risk model is obtained, wherein the user risk pre-training model comprises a Bayesian model; Based on the user risk model, the eigenvector is inferred to obtain a risk value score of the user.

[0009] As an implementable manner, the similarity score of the user and the risk value score of the user are input into a weighted scoring function to obtain a comprehensive recommendation table, comprising the following steps: The similarity score of the user and the risk value score of the user are weighted and fused to obtain a comprehensive recommendation score of the candidate transaction object; A preset transaction threshold is set, the candidate transaction objects are sorted in descending order according to the comprehensive recommendation score, and a comprehensive recommendation table is obtained based on the transaction objects meeting the transaction threshold, wherein the comprehensive recommendation table comprises a user ID, a comprehensive score, a group label and a matching degree.

[0010] As an implementable manner, the weight in the comprehensive scoring function is dynamically adjusted according to the actual transaction rate of the comprehensive recommendation table.

[0011] A blockchain user transaction object recommendation system comprises a feature extraction module, a clustering analysis module, a risk calculation module, a comprehensive recommendation module and an output feedback module: The feature extraction module acquires chain data of a blockchain network transaction object and performs feature extraction and preprocessing to obtain an eigenvector; The clustering analysis module presets an initial clustering number, acquires model parameters of a corresponding Gaussian model, obtains a corresponding overall contour coefficient by combining a Bayesian information criterion, and further obtains a group label of the user, and performs similarity calculation based on the group label to obtain a similarity score of the user; The risk calculation module constructs a user risk pre-training model, trains the user risk pre-training model through historical transaction object data, obtains a user risk model and performs inference through the eigenvector to obtain a risk value score of the user; The comprehensive recommendation module inputs the similarity score of the user and the risk value score of the user into a weighted scoring function to obtain a comprehensive recommendation table; The output feedback module acquires an actual transaction rate of the comprehensive recommendation table and feeds back to the chain data.

[0012] A computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the following method: Chain data of a blockchain network transaction object is acquired and feature extraction and preprocessing are performed to obtain an eigenvector; A preset initial clustering number is obtained, model parameters of a corresponding Gaussian model are obtained, an overall profile coefficient corresponding to the Bayesian information criterion is obtained, and then a group label of the user is obtained, similarity calculation is performed based on the group label to obtain a similarity score of the user; A user risk pre-training model is constructed, the user risk pre-training model is trained through historical transaction object data, a user risk model is obtained, and a risk value score of the user is obtained through inference of the feature vector; The similarity score of the user and the risk value score of the user are input into a weighted score function to obtain a comprehensive recommendation table; An actual transaction rate of the comprehensive recommendation table is obtained and fed back to the on-chain data.

[0013] A blockchain user transaction object recommendation device comprises a memory, a processor and a computer program stored in the memory and running on the processor, characterized in that the processor implements the following method when executing the computer program: On-chain data of a blockchain network transaction object is obtained, feature extraction and preprocessing are performed, and a feature vector is obtained; A preset initial clustering number is obtained, model parameters of a corresponding Gaussian model are obtained, an overall profile coefficient corresponding to the Bayesian information criterion is obtained, and then a group label of the user is obtained, similarity calculation is performed based on the group label to obtain a similarity score of the user; A user risk pre-training model is constructed, the user risk pre-training model is trained through historical transaction object data, a user risk model is obtained, and a risk value score of the user is obtained through inference of the feature vector; The similarity score of the user and the risk value score of the user are input into a weighted score function to obtain a comprehensive recommendation table; An actual transaction rate of the comprehensive recommendation table is obtained and fed back to the on-chain data.

[0014] The present application has the following technical effects: (1) Compared with the technical scheme of the prior art, the blockchain user transaction object recommendation method proposed in the present application combines clustering algorithm and Bayesian algorithm, so that the method can not only provide accurate preference recommendation when the user searches for a transaction object, but also provide certain security for the transaction, and the data after the transaction is used for real-time dynamic adjustment, which better overcomes the limitations of the prior art; (2) With the continuous expansion of the blockchain transaction group size, the types of users in the platform are complex and diverse, and there are potential risk users and new users lack of transaction experience. New users have difficulties in selecting suitable transaction objects and obtaining relevant user information on the chain, resulting in low transaction matching efficiency, low transaction success rate and unstable transaction quality. In view of this market pain point, the technical solution provided by the present application is self-evident in its urgency and necessity; (3) The method combines clustering analysis and Bayesian classification, taking into account group behavior and individual risk characteristics; using the transparency and real-time nature of on-chain data, a closed-loop dynamic optimization is achieved. Improve the user transaction matching efficiency and success rate of the blockchain transaction platform; reduce the occurrence of potential risk transactions, improve the overall security and credibility of the platform; strong scalability, can adapt to different types of blockchain networks and transaction characteristics.

[0015] Other advantages, objects and features of the present application will be apparent from the following description, and will be understood by those skilled in the art through study and practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The flowchart of the method of the present application; Figure 2 The module diagram of the system of the present application; Figure 3 The flowchart of the present application. DETAILED DESCRIPTION

[0017] In order to clearly set forth the present application, make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, so that those skilled in the art can implement it according to the description.

[0018] In the context of the present application, the trusted execution environment provides an isolated running environment from the perspective of underlying hardware and operating system, which can protect the code and data running therein from external attacks, including attacks from operating system, hardware and other application programs. This technology has been used in some fields to achieve the purpose described above, and some basic principles of this technology are known to those skilled in the art, but those skilled in the art will know how to use this technology in this context after reading this application, and will clearly know that this technology has novelty in combination with other features in the specific context.

[0019] The technical solution of the present application will be described in detail below with reference to the drawings and specific embodiments.

[0020] Example 1: A blockchain user transaction object recommendation method, such asFigure 1 As shown, comprising the following steps: S100, obtaining the on-chain data of the blockchain network transaction object and performing feature extraction and preprocessing to obtain a feature vector; S200, presetting an initial clustering number, obtaining model parameters of a corresponding Gaussian model, combining the Bayesian information criterion to obtain a corresponding overall profile coefficient, and then obtaining a group label of the user, and performing similarity calculation based on the group label to obtain a similarity score of the user; S300, constructing a user risk pre-training model, training the user risk pre-training model through historical transaction object data, obtaining a user risk model, and performing inference through the feature vector to obtain a risk value score of the user; S400, inputting the similarity score of the user and the risk value score of the user into a weighted scoring function to obtain a comprehensive recommendation table; S500, obtaining an actual transaction rate of the comprehensive recommendation table and feeding back to the on-chain data.

[0021] The specific flowchart of the method in this embodiment is as shown in Figure 3 .

[0022] (1) The on-chain data of the blockchain network transaction object is extracted and preprocessed to obtain a feature vector, wherein the preprocessing includes differential cleaning, format unification, missing value processing and vectorization encoding; The on-chain data of the blockchain network transaction object is extracted and preprocessed to obtain a feature vector, wherein the preprocessing includes differential cleaning, format unification, missing value processing and vectorization encoding; The data transparency feature of the public chain is used to obtain the on-chain data through a blockchain node or a third-party blockchain data interface, wherein the data includes continuous feature data and discrete feature data; The continuous feature data is filled by using the mean interpolation method and is standardized by using the Z-Score method to obtain a continuous feature vector; The formula of Z-Score is:

[0023] The discrete feature data is filled by using similar samples and is one-hot encoded to obtain a discrete feature vector; The continuous feature vector and the discrete feature vector are aggregated to obtain a feature vector.

[0024] (2) The feature vector is subjected to clustering analysis based on a Gaussian mixture model to obtain a similarity score of the user, including the following steps: Let the initial clustering number be , the corresponding Gaussian distribution initialization parameters are calculated according to the initial clustering number, wherein the Gaussian distribution initialization parameters include: a mean vector , a covariance matrix , mixing coefficient , satisfy ; using the EM algorithm to iteratively update the Gaussian distribution initialization parameters to obtain the log-likelihood value of each Gaussian distribution model; wherein, E step:

[0025] wherein is the posterior probability of the sample belonging to the cluster ; M step:

[0026] wherein, the log-likelihood value calculation formula is:

[0027] a preset difference threshold, until the model converges, and the model convergence determination condition is ; by the initial number of clusters corresponding Gaussian model to calculate the model parameters; wherein, the model parameters include: maximum likelihood value , total number of parameters , total number of samples ; by the Bayesian information criterion, based on the model parameters, to obtain the BIC value, which is represented as follows:

[0028] select the minimum value of the BIC value corresponding to the n initial cluster numbers as the candidate cluster number, and calculate the corresponding overall silhouette coefficient according to the candidate cluster number, which is represented as follows:

[0029] wherein, is the average distance of the sample to all other samples in the same cluster, is the average distance of the sample to all samples in other clusters; select the candidate cluster number with the maximum overall silhouette coefficient as the final cluster number k; calculate the posterior probability of the user belonging to each cluster according to the final cluster number and the feature vector, and select the cluster label with the highest posterior probability as the group label of the user; wherein, the group label is specifically:

[0030] Based on the user's same group tags, calculate the first similarity between the target user and the center of the same group, and the second similarity between the target user and the candidate transaction object; In this embodiment, the first similarity is cosine similarity, which is expressed as follows:

[0031] The second similarity is d-similarity, which is represented as follows:

[0032] The cosine similarity and d-similarity are weighted and fused to obtain the user's similarity score.

[0033] (3) Construct a user risk pre-training model. Train the user risk pre-training model using historical transaction data to obtain the user risk model and infer the user's risk score using feature vectors. This includes the following steps: Users are divided into normal users and tainted users. A user threshold is preset. The data in the feature vector is divided and calculated according to the division rules to obtain the first probability that a normal user is greater than the user threshold and the second probability that a tainted user is greater than the user threshold. The division rules are as follows:

[0034] Based on the partitioning rules, the binarized feature vector is obtained. ; In this embodiment, the user risk pre-training model includes a Bayesian model. Based on the inherent assumptions of the Bayesian model, the feature vectors are independent of each other. Specifically, given that a user is known to be a "normal user" or a "tainted user," the various features describing the user (including transaction amount, transaction frequency, and transaction type, etc.) do not affect each other; the value of one feature does not affect the probability of another feature. For example, when determining whether a user is a "tainted user," the features "whether the transaction amount exceeds the limit" and "whether the transaction occurred at night" are independent of each other, and there is no need to consider whether there is a correlation between them. This simplifies the computation of the Bayesian formula and allows for a more efficient acquisition of the user risk score.

[0035] By binarizing the feature vectors, the posterior probability is calculated using Bayes' theorem, and the user risk model is obtained through training, as shown below:

[0036] in, This represents the probability that each characteristic data point of a normal user is greater than a given constant. This represents the probability that each feature data of a tainted user is greater than a given constant. For joint probability, stain user prior probability estimated by historical proportion; According to the Bayesian model and the feature vector, the risk value score of the user is obtained, and is expressed as follows: .

[0037] (4) The similarity score and the risk value score of the user are input into a weighted scoring function to obtain a comprehensive recommendation table, including the following steps: According to the similarity score and the risk value score, a comprehensive recommendation score of the candidate transaction object is obtained by weighted fusion; A preset transaction threshold is set, the candidate transaction objects are sorted in descending order according to the comprehensive recommendation score, and the transaction objects greater than the transaction threshold are selected to generate a comprehensive recommendation table, wherein the comprehensive recommendation table includes user ID, comprehensive score, group label, matching degree and other key information; (5) The actual transaction rate of the recommendation table is fed back to the on-chain data set, and the weight in the comprehensive scoring function is dynamically adjusted.

[0038] In summary, the present application realizes the group label division and risk level evaluation of the user by clustering and classification analysis of the on-chain data, which can effectively reduce the potential fraud, money laundering and other security risks in the blockchain transaction platform. Then, combined with the group center similarity and the candidate object similarity, a comprehensive recommendation list is generated to provide high-quality transaction object recommendation for users, thereby improving the transaction efficiency and user satisfaction of the platform. Finally, the transaction feedback of the user is used to dynamically optimize the scoring model, enhancing the adaptability and intelligent level of the system, so that the platform forms a positive cycle in user interaction. The present application improves the security, transparency and credibility of the ecological environment in the blockchain transaction scene, and promotes the long-term healthy development and user stickiness of the platform.

[0039] Example 2: A blockchain user transaction object recommendation system, as shown in Figure 2 The feature extraction module 100, the clustering analysis module 200, the risk calculation module 300, the comprehensive recommendation module 400 and the output feedback module 500 are shown in the figure: The feature extraction module 100 obtains the on-chain data of the blockchain network transaction object and performs feature extraction and preprocessing to obtain a feature vector; The clustering analysis module 200 presets an initial clustering number, obtains the model parameters of the corresponding Gaussian model, combines the Bayesian information criterion to obtain the corresponding overall contour coefficient, and then obtains the group label of the user, and performs similarity calculation based on the group label to obtain the similarity score of the user; The risk calculation module 300 constructs a user risk pre-training model, trains the user risk pre-training model through historical transaction object data, obtains a user risk model, and performs inference through a feature vector to obtain a risk value score of the user; The comprehensive recommendation module 400 inputs the similarity score of the user and the risk value score of the user into a weighted score function to obtain a comprehensive recommendation table; The output feedback module 500 obtains an actual transaction rate of the comprehensive recommendation table and feeds back to on-chain data.

[0040] Various changes and modifications can be made within the spirit and scope of the application without departing from it.

[0041] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between the embodiments can be referred to each other.

[0042] Those skilled in the art should understand that embodiments of the present application can be provided as a method, device, or computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0043] The present application is described with reference to flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device produce a means for implementing the functions specified in the flowchart and / or block diagram. Figure 1 The function specified in one flow or multiple flows and / or blocks. Figure 1 The function specified in one flow or multiple flows and / or blocks.

[0044] These computer program instructions can also be stored in a computer readable storage medium that can guide the computer or other programmable data processing terminal device to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction means, which implements the functions specified in the flowchart and / or block diagram. Figure 1 The function specified in one flow or multiple flows and / or blocks. Figure 1 The function specified in one flow or multiple flows and / or blocks.

[0045] These computer program instructions can also be loaded into computer or other programmable data processing devices to cause a series of operational steps to be performed on the computer or other programmable devices to produce computer-implemented processes such that the instructions which execute on the computer or other programmable devices provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 Figure 1 The functions needs not be implemented as listed in any particular order. One of ordinary skill in the art will immediately appreciate that the functions can be implemented in any order or in any combination.

[0046] It is noted that: The phrase "one embodiment" or "an embodiment" as used herein does not necessarily refer to one and the same embodiment, but can

[0047] The above description of the embodiments is for the purpose of understanding and application of the present application. Those skilled in the art will readily perceive various modifications to the above-described embodiments, and apply the general principles of the present application to other embodiments and applications without departing from the spirit and scope of the present application. Thus, the present application is not to be limited to the embodiments described herein but is to be given the full scope defined by the appended claims and their equivalents.​

Claims

1. A blockchain user transaction object recommendation method, characterized in that, The method comprises the following steps: Obtaining chain data of a blockchain network transaction object and performing feature extraction and preprocessing to obtain a feature vector; Presetting an initial clustering number, obtaining model parameters of a corresponding Gaussian model, obtaining a corresponding overall silhouette coefficient based on a Bayesian information criterion, and then obtaining a group label of a user, and performing similarity calculation based on the group label to obtain a similarity score of the user; Constructing a user risk pre-training model, training the user risk pre-training model through historical transaction object data, obtaining a user risk model, and performing inference through the feature vector to obtain a risk value score of the user; Inputting the similarity score of the user and the risk value score of the user into a weighted scoring function to obtain a comprehensive recommendation table; Obtaining an actual transaction rate of the comprehensive recommendation table and feeding back to the chain data. 2.The method of claim 1, wherein, The method comprises the following steps: Using the data transparency feature of a public chain, obtaining chain data through a blockchain node or a third-party blockchain data interface, wherein the chain data comprises continuous feature data and discrete feature data; Filling the continuous feature data by using a mean value interpolation method and performing Z-Score standardization to obtain a continuous feature vector; Filling the discrete feature data by using a similar sample and performing one-hot encoding to obtain a discrete feature vector; Aggregating the continuous feature vector and the discrete feature vector and performing preprocessing to obtain a feature vector, wherein the preprocessing comprises one or more of differential cleaning, format unification, missing value processing, and vectorization encoding. 3.The method of claim 1, wherein, The similarity score of the user is obtained by the following steps: Calculating Gaussian distribution initialization parameters according to an initial clustering number, iteratively updating the Gaussian distribution initialization parameters, and obtaining a log-likelihood value of each Gaussian distribution model; Presetting a difference threshold, and according to a difference of a comparison result of adjacent log-likelihood values, the difference is less than the difference threshold; Calculating model parameters of each initial clustering number through a Gaussian model, and obtaining a BIC value based on the Bayesian information criterion and the model parameters; Selecting n initial clustering numbers corresponding to a minimum value of the BIC value as candidate clustering numbers, and calculating overall silhouette coefficients under each clustering number through the candidate clustering numbers; Obtaining a final clustering number according to the overall silhouette coefficients, and obtaining a similarity score of the user based on the final clustering number and the feature vector. 4.The method of claim 3, wherein, The similarity score of the user based on the final clustering number and the feature vector comprises the following steps: Obtaining a posterior probability of the user belonging to each cluster based on the final clustering number and the feature vector, and obtaining a group label of the current user based on a cluster label corresponding to the posterior probability; Obtaining a first similarity between a target user and a same-group center according to the same-group label of the user, and calculating a second similarity between the target user and a candidate transaction object; Weighted fusion of the first similarity and the second similarity to obtain a similarity score of a blockchain user. 5.The method of claim 1, wherein, The user risk pre-training model is constructed, the user risk pre-training model is trained through historical transaction object data, a user risk model is obtained, and a risk value score of the user is obtained through inference of the feature vector, and the method comprises the following steps: The user is divided into a normal user and a tainted user, a user threshold is preset, and the feature vector is divided and calculated to obtain a first probability that the normal user is greater than the user threshold and a second probability that the tainted user is greater than the user threshold; The user risk pre-training model is trained through a Bayesian formula to calculate a posterior probability according to the feature vector, the first probability and the second probability, and a user risk model is obtained, wherein the user risk pre-training model comprises a Bayesian model; The feature vector is inferred based on the user risk model to obtain a risk value score of the user. 6.The method of claim 1, wherein, The similarity score of the user and the risk value score of the user are input into a weighted scoring function to obtain a comprehensive recommendation table, and the method comprises the following steps: The similarity score of the user and the risk value score of the user are weighted and fused to obtain a comprehensive recommendation score of a candidate transaction object; A transaction threshold is preset, the candidate transaction object is sorted in descending order according to the comprehensive recommendation score, and a comprehensive recommendation table is obtained based on a transaction object that meets the transaction threshold, wherein the comprehensive recommendation table comprises a user ID, a comprehensive score, a group label and a matching degree. 7.The method of claim 1, wherein, The weight in the comprehensive scoring function is dynamically adjusted according to an actual transaction rate of the comprehensive recommendation table. 8.A blockchain user transaction object recommendation system, characterized in that, The method comprises a feature extraction module, a clustering analysis module, a risk calculation module, a comprehensive recommendation module and an output feedback module: The feature extraction module acquires chain data of a blockchain network transaction object and performs feature extraction and preprocessing to obtain a feature vector; The clustering analysis module presets an initial clustering number, acquires model parameters of a corresponding Gaussian model, obtains a corresponding overall contour coefficient based on a Bayesian information criterion, and further obtains a group label of the user, and performs similarity calculation based on the group label to obtain a similarity score of the user; The risk calculation module constructs a user risk pre-training model, trains the user risk pre-training model through historical transaction object data, obtains a user risk model and performs inference through the feature vector to obtain a risk value score of the user; The comprehensive recommendation module inputs the similarity score of the user and the risk value score of the user into a weighted scoring function to obtain a comprehensive recommendation table; The output feedback module acquires an actual transaction rate of the comprehensive recommendation table and feeds back to the chain data.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by the processor to implement the method of any one of claims 1 to 7. 10.A blockchain user transaction object recommendation apparatus, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 1 to 7.

Images