Blockchain data multidimensional space full life cycle supervision method, system, equipment and storage medium

Abstract

This application discloses a blockchain data multi-dimensional space full life cycle supervision method, system, device and storage medium, which belongs to the field of data processing technology. The embodiment of this application realizes the comprehensive and in-depth mining of supervision elements in blockchain data, and realizes accurate and full life cycle supervision of supervision targets through intelligent algorithms. In this way, it not only improves the comprehensiveness and timeliness of supervision, but also provides strong support and guarantee for the widespread application of blockchain technology.

Description

Technical Field

[0001] The present application belongs to the field of data processing technology, and specifically relates to a method, system, device and storage medium for supervising the entire life cycle of blockchain data multi-dimensional space. Background Art

[0002] With the rapid development of blockchain technology, its application in various fields is becoming more and more extensive. However, the supervision of blockchain data faces many challenges. Traditional supervision methods often find it difficult to fully and deeply explore the regulatory elements in blockchain data, resulting in poor supervision results. Especially in areas such as smart manufacturing, the multidimensionality and complexity of blockchain data make supervision more difficult. How to effectively obtain and process regulatory elements in blockchain data and achieve comprehensive and accurate supervision has become a technical problem that needs to be solved urgently.

[0003] To address this problem, existing technical solutions often focus on single-dimensional data mining or simple pattern matching, which is difficult to fully reflect the complexity and multidimensionality of blockchain data. At the same time, due to the temporal and correlation nature of blockchain data, traditional data mining algorithms often do not work well when processing such data. Therefore, a new technical solution is needed that can fully and deeply mine regulatory elements in blockchain data and achieve full life cycle supervision. Summary of the invention

[0004] The present application provides a method, system, device and storage medium for supervising the entire life cycle of a blockchain data multi-dimensional space, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0005] The embodiment of the present application provides a method for supervising the full life cycle of a multi-dimensional space of blockchain data, which is applied to a big data supervision system. The method comprises: obtaining the original supervision element text semantics of a first block supervision target, wherein the original supervision element text semantics comprises the original supervision element text semantics of a block supervision target index keyword and the original supervision element text semantics of block supervision target behavior data, wherein the first block supervision target is a preliminarily screened regulated blockchain data entity, and the first block supervision target is in the multi-dimensional space of blockchain data; obtaining the derived supervision element text semantics of the block supervision target index keyword of the first block supervision target through a supervision decision processing algorithm according to the original supervision element text semantics of the first block supervision target; obtaining the target supervision element text semantics of the first block supervision target according to the derived supervision element text semantics of the block supervision target index keyword of the first block supervision target and the original supervision element text semantics of the block supervision target behavior data of the first block supervision target, so as to perform full life cycle supervision of the first block supervision target in the multi-dimensional space of blockchain data according to the target supervision element text semantics; wherein the supervision decision processing algorithm is based on multi-dimensional space. The method is obtained by debugging a first long short-term memory algorithm based on a first training data, wherein the first training data includes the original regulatory element text semantics of the second block regulatory target and the debugging learning basis, wherein the debugging learning basis is the prior regulatory element text semantics of the second block regulatory target obtained by the debugged second long short-term memory algorithm based on the original regulatory element text semantics of the second block regulatory target, and the second block regulatory target is a regulated blockchain data entity; the debugged second long short-term memory algorithm is obtained by debugging the second long short-term memory algorithm based on multiple second training data configured with prior knowledge, wherein the second training data includes the original regulatory element text semantics of the regulatory participating entity and the regulated blockchain data entity, and the prior knowledge represents whether the regulatory participating entity in the second training data triggers a risk regulatory contract on the regulated blockchain data entity, the input information of the second long short-term memory algorithm includes the original regulatory element text semantics of each block regulatory target in the second training data, and the output is the derived regulatory element text semantics of each block regulatory target, and the debugging error information is determined based on the semantic commonality reinforcement coefficient between the derived regulatory element text semantics of the two block regulatory targets in each second training data.

[0006] An embodiment of the present application provides a big data supervision system, comprising at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor executes the above method.

[0007] The embodiment of the present application provides a big data supervision device, which is used to: obtain the original supervision element text semantics of the first block supervision target, the original supervision element text semantics includes the original supervision element text semantics of the block supervision target index keyword and the original supervision element text semantics of the block supervision target behavior data, the first block supervision target is a preliminarily screened regulated blockchain data entity, and the first block supervision target is in the blockchain data multi-dimensional space; according to the original supervision element text semantics of the first block supervision target, the derived supervision element text semantics of the block supervision target index keyword of the first block supervision target is obtained through a supervision decision processing algorithm; according to the derived supervision element text semantics of the block supervision target index keyword of the first block supervision target and the original supervision element text semantics of the block supervision target behavior data of the first block supervision target, the target supervision element text semantics of the first block supervision target is obtained, so as to perform full life cycle supervision of the first block supervision target in the blockchain data multi-dimensional space according to the target supervision element text semantics; wherein the supervision decision processing algorithm is based on a plurality of first training data pairs. The first training data includes the original regulatory element text semantics of the second block regulatory target and the debugging learning basis, the debugging learning basis is the prior regulatory element text semantics of the second block regulatory target obtained by the debugged second long short-term memory algorithm based on the original regulatory element text semantics of the second block regulatory target, and the second block regulatory target is the regulated blockchain data entity; the debugged second long short-term memory algorithm is debugged by the second long short-term memory algorithm based on multiple second training data configured with prior knowledge, the second training data includes the original regulatory element text semantics of the regulatory participating entity and the regulated blockchain data entity, the prior knowledge represents whether the regulatory participating entity in the second training data triggers the risk supervision contract on the regulated blockchain data entity, the input information of the second long short-term memory algorithm includes the original regulatory element text semantics of each block regulatory target in the second training data, and the output is the derived regulatory element text semantics of each block regulatory target, and the debugging error information is determined according to the semantic commonality reinforcement coefficient between the derived regulatory element text semantics of the two block regulatory targets in each second training data.

[0008] An embodiment of the present application provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0009] The embodiment of this application realizes the comprehensive and in-depth mining of regulatory elements in blockchain data, and realizes accurate and full-life cycle supervision of regulatory targets through intelligent algorithms. In this way, it not only improves the comprehensiveness and timeliness of supervision, but also provides strong support and guarantee for the widespread application of blockchain technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flowchart of a method for supervising the entire life cycle of multi-dimensional space of blockchain data provided in an embodiment of the present application.

[0011] Figure 2 A schematic diagram of the structure of a big data supervision system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0013] The terms "first", "second", etc. in this application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are usually a class, and the number of objects is not limited. For example, the first object can be one or more. In addition, in this application, "and / or" represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0014] Figure 1 A method for supervising the entire life cycle of multidimensional space of blockchain data is shown, which is applied to a big data supervision system. The method includes the following steps 110-130.

[0015] In today's digital age, blockchain technology, as a decentralized, tamper-proof data storage and processing method, is being widely used in various fields. However, with the rapid development of blockchain technology and the continuous expansion of application scenarios, how to effectively regulate blockchain data to ensure its compliance, security and traceability has become an urgent problem to be solved.

[0016] As an advanced data supervision tool, the big data supervision system provided in the embodiment of the present application can achieve effective supervision of the entire life cycle of blockchain data through comprehensive collection, intelligent analysis and real-time monitoring of blockchain data. The following is a specific application scenario example based on steps 110 to 130, which is used to introduce in detail how the big data supervision system performs full life cycle supervision in the multidimensional space of blockchain data.

[0017] Step 110: Obtain the original regulatory element text semantics of the first block regulatory target, wherein the original regulatory element text semantics include the original regulatory element text semantics of the block regulatory target index keyword and the original regulatory element text semantics of the block regulatory target behavior data. The first block regulatory target is a preliminarily screened regulated blockchain data entity, and the first block regulatory target is in the multi-dimensional space of blockchain data.

[0018] Step 120: Based on the original regulatory element text semantics of the first block regulatory target, the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target is obtained through a regulatory decision processing algorithm.

[0019] Step 130: Obtain the target regulatory element text semantics of the first block regulatory target based on the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target, so as to perform full life cycle supervision of the first block regulatory target in the multi-dimensional space of the blockchain data based on the target regulatory element text semantics.

[0020] Among them, the regulatory decision processing algorithm is obtained by debugging the first long short-term memory algorithm based on multiple first training data, the first training data includes the original regulatory element text semantics of the second block regulatory target and the debugging learning basis, the debugging learning basis is the prior regulatory element text semantics of the second block regulatory target obtained by the debugged second long short-term memory algorithm based on the original regulatory element text semantics of the second block regulatory target, and the second block regulatory target is the regulated blockchain data entity; the debugged second long short-term memory algorithm is debugged by the second long short-term memory algorithm based on multiple second training data configured with prior knowledge, the second training data includes the original regulatory element text semantics of the regulatory participating entity and the regulated blockchain data entity, the prior knowledge represents whether the regulatory participating entity in the second training data triggers the risk supervision contract on the regulated blockchain data entity, the input information of the second long short-term memory algorithm includes the original regulatory element text semantics of each block regulatory target in the second training data, and the output is the derived regulatory element text semantics of each block regulatory target, and the debugging error information is determined based on the semantic commonality reinforcement coefficient between the derived regulatory element text semantics of the two block regulatory targets in each second training data.

[0021] A specific application scenario example involved in step 110-step 130 involves cross-border payment blockchain supervision, which is an important application field of blockchain technology. In the traditional cross-border payment process, due to the complex factors such as financial institutions, payment systems, laws and regulations in different countries and regions, the payment process is cumbersome, costly and inefficient. The decentralized, tamper-proof and smart contract characteristics of blockchain technology provide a more convenient, secure and low-cost solution for cross-border payments. However, with the continuous promotion of cross-border payment blockchain applications, how to effectively supervise them has become an urgent problem to be solved.

[0022] In step 110, the primary task of the big data supervision system is to obtain the original supervision element text semantics of the first block supervision target. The first block supervision target in the embodiment of the present application refers to the initially screened supervised blockchain data entity, which is in the blockchain data multidimensional space. The original supervision element text semantics includes the original supervision element text semantics of the block supervision target index keyword and the original supervision element text semantics of the block supervision target behavior data.

[0023] Specifically, the big data supervision system will obtain key data such as transaction data, smart contract execution records, account information, etc. on the blockchain in real time through the data interface with the cross-border payment blockchain platform. These data will be stored in text form and contain rich regulatory element information, such as transaction amount, transaction parties, transaction time, transaction type, smart contract content, etc. The big data supervision system will perform semantic analysis on these original regulatory element texts and extract key information for subsequent processing.

[0024] In step 120, after obtaining the original regulatory element text semantics of the first block regulatory target, the big data regulatory system will obtain the derived regulatory element text semantics of the block regulatory target index keyword through the regulatory decision processing algorithm based on these text semantics. The regulatory decision processing algorithm is one of the core algorithms of the big data regulatory system. It is based on deep learning technology and can perform deep mining and intelligent analysis on the original regulatory element text to extract useful information hidden in the text.

[0025] In detail, the regulatory decision processing algorithm is obtained by debugging the first long short-term memory algorithm based on multiple first training data. The first training data includes the original regulatory element text semantics of the second block regulatory target and the debugging learning basis. The debugging learning basis is the prior regulatory element text semantics of the second block regulatory target obtained by the debugged second long short-term memory algorithm based on the original regulatory element text semantics of the second block regulatory target. The second block regulatory target here is also a regulated blockchain data entity, but unlike the first block regulatory target, it is used to train the regulatory decision processing algorithm.

[0026] Through the processing of regulatory decision-making processing algorithms, the big data regulatory system can extract more useful information from the original regulatory element text, such as the abnormality of transaction patterns, potential risks of smart contracts, etc. This information is expressed in the form of derived regulatory element text semantics, providing strong support for subsequent full life cycle supervision.

[0027] In step 130, after obtaining the derived regulatory element text semantics of the block regulatory target index keyword, the big data regulatory system will further combine the original regulatory element text semantics of the block regulatory target behavior data to obtain the target regulatory element text semantics of the first block regulatory target. The target regulatory element text semantics is important information for a comprehensive and accurate description of the first block regulatory target, and it contains all the key regulatory element information of the block regulatory target.

[0028] Based on the text semantics of the target supervision elements, the big data supervision system can conduct full life cycle supervision of the first block supervision target. Full life cycle supervision refers to the monitoring and management of the entire process from the generation, storage, processing, analysis to destruction of block supervision targets. In this application scenario, full life cycle supervision includes real-time monitoring and analysis of key data such as transaction data, smart contract execution records, account information, etc. on the cross-border payment blockchain.

[0029] The big data supervision system will discover potential supervision risks and problems through in-depth analysis and mining of the text semantics of the target supervision elements. For example, the system will monitor in real time whether the transaction amount is abnormal, whether there is a relationship between the two parties of the transaction, whether there are loopholes in the smart contract, etc. Once an abnormal situation or potential risk is found, the system will immediately issue an alarm and notify the relevant supervisors for further investigation and handling.

[0030] In addition, the big data supervision system will also conduct retrospective analysis on the historical data of block supervision targets to discover possible historical violations or risk points. Through in-depth mining and analysis of historical data, the system can reveal the laws and trends hidden behind the data and provide strong decision-making support for regulatory authorities.

[0031] It can be understood that in the big data supervision system, the supervision decision processing algorithm plays a vital role. In order to ensure the accuracy and effectiveness of the algorithm, it needs to be continuously debugged and optimized. In this application scenario, the supervision decision processing algorithm is obtained by debugging the first long short-term memory algorithm based on multiple first training data. The first long short-term memory algorithm is debugged based on the second long short-term memory algorithm. The second long short-term memory algorithm is a deep learning algorithm with powerful time series processing capabilities. In this application scenario, it is used to process the time series information of blockchain data. In order to obtain better processing results, the second long short-term memory algorithm needs to be debugged. The basis for debugging is multiple second training data configured with prior knowledge. The second training data includes the original supervision element text semantics of the supervision participating entity and the regulated blockchain data entity. The prior knowledge characterizes whether the supervision participating entity in the second training data triggers the risk supervision contract on the regulated blockchain data entity.

[0032] By debugging the second long short-term memory algorithm, more accurate and useful derived regulatory element text semantics can be obtained. These derived regulatory element text semantics provide strong support for subsequent regulatory decisions. At the same time, by further debugging and optimizing the first long short-term memory algorithm, a regulatory decision processing algorithm that is more suitable for cross-border payment blockchain regulatory scenarios can be obtained.

[0033] Through the implementation of the above steps, the big data supervision system can effectively supervise the cross-border payment blockchain throughout its life cycle in the multi-dimensional space of blockchain data. It can not only monitor and analyze key data on the blockchain in real time and discover potential regulatory risks and problems, but also conduct retrospective analysis of historical data to reveal the laws and trends hidden behind the data. This provides strong decision-making support for regulatory authorities and helps maintain the stability and healthy development of the cross-border payment market.

[0034] It is worth mentioning that step 120 is the core invention point of the embodiment of the present application. It describes in detail how to obtain the derived regulatory element text semantics of the index keyword of the block regulatory target through the regulatory decision processing algorithm based on the original regulatory element text semantics of the first block regulatory target.

[0035] First, in the supervision of blockchain, each block supervision target contains a large amount of original supervision element text semantics, which is the basis of supervision decision-making. For example, in a cross-border payment blockchain application, the original supervision element text semantics may include the account information of both parties to the transaction, the transaction amount, the transaction time, the transaction type, etc. This information exists in the form of text and contains a wealth of supervision elements.

[0036] In order to extract useful information from the text semantics of these original regulatory elements, the big data regulatory system needs to use a regulatory decision processing algorithm. This algorithm is based on deep learning technology, especially the long short-term memory algorithm (LSTM), which can perform deep mining and intelligent analysis of text data. During the algorithm training process, the big data regulatory system uses a large amount of first training data, which includes the original regulatory element text semantics and debugging learning basis of the second block regulatory target. Through continuous training and optimization, the regulatory decision processing algorithm can learn how to extract useful information from the original regulatory element text semantics.

[0037] When the regulatory decision processing algorithm is applied to the original regulatory element text semantics of the first block regulatory target, it will first perform semantic analysis on these texts to understand the meaning and contextual relationship. Then, the algorithm will extract key information and express this information in the form of derived regulatory element text semantics. These derived regulatory element text semantics are further refinements and sublimations of the original regulatory element text semantics, and they contain more useful information and regulatory elements.

[0038] For example, in blockchain applications for cross-border payments, the derived regulatory element text semantics may include the abnormality of transaction patterns, potential risks of smart contracts, and the relationship between the two parties to the transaction. This information is not in the original regulatory element text semantics, but is obtained through intelligent analysis of the regulatory decision processing algorithm. This information is very useful for regulators because it can help them better understand transaction activities on the blockchain and discover potential regulatory risks and problems.

[0039] It can be seen that step 120 uses the regulatory decision processing algorithm to perform in-depth mining and intelligent analysis on the original regulatory element text semantics of the first block regulatory target, and obtains the derived regulatory element text semantics containing more useful information and regulatory elements. This information provides strong support for subsequent regulatory decisions and helps maintain the stability and healthy development of blockchain applications.

[0040] In addition, step 130, as another core invention point of the embodiment of the present application, elaborates in detail how to combine the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data to obtain the target regulatory element text semantics of the block regulatory target, and perform full life cycle supervision of the block regulatory target in the multi-dimensional space of blockchain data based on the target regulatory element text semantics.

[0041] First, the big data supervision system has obtained the derived supervision factor text semantics of the block supervision target index keyword of the first block supervision target through the supervision decision processing algorithm. These derived supervision factor text semantics are in-depth mining and refinement of the original supervision factor text semantics, and contain more useful information and supervision factors, such as the abnormality of transaction patterns and the potential risks of smart contracts.

[0042] Secondly, the big data supervision system also needs to consider the original supervision element text semantics of the block supervision target behavior data of the first block supervision target. These data record all the behaviors of the block supervision target on the blockchain, such as transaction records, execution of smart contracts, etc. These data also exist in the form of text and contain rich supervision elements.

[0043] In order to obtain the target regulatory element text semantics of the first block regulatory target, the big data regulatory system needs to organically combine the above two text semantics. This means that the big data regulatory system needs to fuse the derived regulatory element text semantics with the original regulatory element text semantics to form a more comprehensive and accurate regulatory element text semantics. This fusion process may involve operations such as matching, merging, and deduplication of text semantics to ensure that the target regulatory element text semantics obtained is accurate and non-redundant.

[0044] After obtaining the text semantics of the target regulatory elements, the big data regulatory system can conduct full life cycle supervision of the first block regulatory target in the multi-dimensional space of blockchain data based on the semantics. Full life cycle supervision means that the big data regulatory system must monitor and manage the entire process of block regulatory targets from generation, storage, processing, analysis to destruction. In this process, the big data regulatory system will analyze the text semantics of the target regulatory elements in real time to discover potential regulatory risks and problems. Once an abnormal situation or potential risk is found, the big data regulatory system will immediately issue an alarm and notify relevant regulators for further investigation and processing.

[0045] Therefore, step 130 obtains a more comprehensive and accurate target regulatory element text semantics by combining the derived regulatory element text semantics with the original regulatory element text semantics, and performs full life cycle supervision of the first block regulatory target in the blockchain data multidimensional space based on the semantics. This technical solution provides strong support for the supervision of blockchain applications and helps maintain the stability and healthy development of blockchain applications.

[0046] To facilitate understanding of the technical solutions described in the embodiments of the present application, steps 110 to 130 are introduced below through an application scenario.

[0047] In the field of smart manufacturing, quality supervision is a key link to ensure that products meet standards and enhance brand reputation. With the introduction of blockchain technology, quality supervision of smart manufacturing can be carried out in a more transparent and traceable environment. However, how to effectively supervise smart manufacturing data on the blockchain has become an urgent problem to be solved. As an advanced data analysis tool, the big data supervision system can provide strong support for quality supervision of smart manufacturing.

[0048] For example, there is a smart manufacturing factory that uses blockchain technology to record various data in the production process to ensure the immutability and traceability of the data. In this scenario, the big data supervision system is used to supervise the smart manufacturing data on the blockchain, especially the quality-related data. The first block supervision target, that is, the initially screened blockchain data entity, may be a specific production batch, a key production link, or a specific smart device.

[0049] The big data supervision system first obtains the original supervision element text semantics of the first block supervision target. These original supervision element text semantics include the original supervision element text semantics of the block supervision target index keywords and the original supervision element text semantics of the block supervision target behavior data. Taking the production batch as an example, the original supervision element text semantics of the block supervision target index keywords may include "production batch number", "production date", "production line number", etc. These keywords are key information for identifying a specific production batch and are of great significance for subsequent supervision decisions. The original supervision element text semantics of the block supervision target behavior data may include various data of the production batch during the production process, such as raw material quality, production parameters, equipment status, inspection records, etc. These data exist in the form of text and contain rich supervision elements. They are an important basis for judging whether the quality of the production batch is qualified.

[0050] The big data supervision system obtains the derived supervision element text semantics of the block supervision target index keyword of the first block supervision target through the supervision decision processing algorithm based on the original supervision element text semantics of the first block supervision target. The supervision decision processing algorithm is the core of the big data supervision system. In this scenario, the supervision decision processing algorithm is obtained by debugging the first long short-term memory algorithm (LSTM) based on multiple first training data. The first training data includes the original supervision element text semantics of the second block supervision target and the debugging learning basis. The debugging learning basis is the prior supervision element text semantics of the second block supervision target obtained by the debugged second long short-term memory algorithm based on the original supervision element text semantics of the second block supervision target. Taking another production batch as the second block supervision target as an example, its original supervision element text semantics also includes index keywords such as "production batch number", "production date", "production line number" and various data in the production process. Through the debugged second long short-term memory algorithm, the prior supervision element text semantics of this production batch can be obtained, such as "qualified quality" and "raw material problems". These prior supervision element text semantics are obtained based on historical data and experience, and have important reference value for quality supervision of new production batches. After obtaining the original regulatory element text semantics of the first block regulatory target, the big data regulatory system inputs these text semantics into the regulatory decision processing algorithm. The algorithm will conduct in-depth mining and intelligent analysis on these text semantics, extract key information, and express this information in the form of derived regulatory element text semantics. These derived regulatory element text semantics are further refinements and sublimations of the original regulatory element text semantics, and they contain more useful information and regulatory elements. Taking the production batch as an example, the derived regulatory element text semantics may include "production batch quality stability analysis", "production parameter anomaly detection", "raw material quality risk assessment", etc. This information is not in the original regulatory element text semantics, but is obtained through the intelligent analysis of the regulatory decision processing algorithm. They are of great significance for judging whether the quality of the production batch is qualified and whether there are potential risks.

[0051] The big data supervision system obtains the target supervision element text semantics of the first block supervision target based on the derived supervision element text semantics of the block supervision target index keyword of the first block supervision target and the original supervision element text semantics of the block supervision target behavior data of the first block supervision target. The target supervision element text semantics is an important basis for the big data supervision system to conduct full life cycle supervision of the first block supervision target. It integrates the information of the derived supervision element text semantics and the original supervision element text semantics to form a more comprehensive and accurate supervision element text semantics. Taking the production batch as an example, the target supervision element text semantics may include "the production batch quality is qualified", "there are problems with the quality of raw materials", "need to further strengthen the monitoring of production parameters", etc. Based on the target supervision element text semantics, the big data supervision system can conduct full life cycle supervision of the first block supervision target in the multi-dimensional space of blockchain data. Full life cycle supervision means that the big data supervision system must monitor and manage the entire process of the block supervision target from generation, storage, processing, analysis to destruction. In this process, the big data supervision system will analyze the target supervision element text semantics in real time to discover potential supervision risks and problems. Taking the production batch as an example, the big data supervision system will monitor all data of the production batch on the blockchain in real time. Once data anomalies are found or risk information such as "quality problems" appears in the text semantics of the target regulatory element, the system will immediately issue an alarm and notify relevant supervisors for further investigation and processing. In this way, the big data supervision system can ensure that quality supervision in the intelligent manufacturing process is more accurate and efficient.

[0052] In this way, through the three steps of obtaining the original regulatory element text semantics, obtaining the derived regulatory element text semantics, and obtaining the target regulatory element text semantics and conducting full life cycle supervision, the big data supervision system can effectively supervise and analyze the smart manufacturing data on the blockchain. This not only improves the accuracy and efficiency of quality supervision, but also provides strong support for the sustainable development of smart manufacturing.

[0053] In summary, by applying the embodiments of the present application, by obtaining the original regulatory element text semantics of the first block regulatory target, including the original regulatory element text semantics of the block regulatory target index keywords and the original regulatory element text semantics of the block regulatory target behavior data, a comprehensive and detailed information basis is provided for the supervision of blockchain data entities. Furthermore, through the regulatory decision processing algorithm, the derived regulatory element text semantics of the block regulatory target index keywords of the first block regulatory target can be mined, which greatly enriches the dimensions of the regulatory elements and makes the supervision more in-depth and precise. Finally, by combining the derived regulatory element text semantics and the original regulatory element text semantics, the target regulatory element text semantics of the first block regulatory target is obtained, thereby realizing the full life cycle supervision of the first block regulatory target in the multi-dimensional space of blockchain data, and significantly improving the comprehensiveness and timeliness of supervision.

[0054] What is particularly important is that the regulatory decision processing algorithm of the embodiment of the present application is obtained through a carefully debugged first long short-term memory algorithm, in which the original regulatory element text semantics and the prior regulatory element text semantics of the second block regulatory target are fully utilized to ensure the accuracy and effectiveness of the algorithm. At the same time, prior knowledge is incorporated into the debugging process of the second long short-term memory algorithm, so that the algorithm can better understand and process regulatory data, further improving the intelligent level of supervision. Through technological innovation, the embodiment of the present application effectively solves technical problems such as lack of in-depth information mining, incomplete supervision, and poor timeliness in blockchain data supervision, and provides strong regulatory support for the widespread application of blockchain technology.

[0055] In some possible embodiments, the block supervision target behavior data includes at least one block supervision target behavior vector, and the method further includes: taking the first block supervision target as a latent space feature unit, obtaining a latent space feature relationship network corresponding to the first block supervision target, wherein the latent space feature relationship network includes latent space feature units of at least two states, and there are relationship transmission pointers between latent space feature units involved in the business, and a latent space feature unit in the latent space feature relationship network represents a regulated blockchain data entity, a regulatory participating entity, a block supervision target behavior vector of a regulated blockchain data entity, or a block supervision target behavior vector of a regulatory participating entity; taking the latent space feature unit corresponding to the first block supervision target as a key latent space feature unit, and determining at least one upstream and downstream latent space feature unit of the key latent space feature unit based on the latent space feature relationship network; and obtaining the regulatory element text semantics of each upstream and downstream latent space feature unit in the at least one upstream and downstream latent space feature unit.

[0056] Furthermore, the target regulatory element text semantics of the first block regulatory target is obtained based on the derived regulatory element text semantics of the block regulatory target index keywords of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target, including: obtaining the derived regulatory element text semantics of the first block regulatory target by integrating the derived regulatory element text semantics of the block regulatory target index keywords of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target; obtaining the target regulatory element text semantics of the first block regulatory target based on the derived regulatory element text semantics of the first block regulatory target and the regulatory element text semantics of each of the upstream and downstream latent space feature units.

[0057] In combination with the foregoing, in the two major application scenarios of cross-border payment blockchain supervision and intelligent manufacturing quality supervision, the technical solution described in the embodiments of this application demonstrates its unique advantages and wide application value.

[0058] 1. Cross-border payment blockchain supervision

[0059] In the field of cross-border payments, the introduction of blockchain technology has greatly improved the transparency and security of transactions, but it has also brought new challenges to the big data supervision system. How to effectively supervise blockchain data in cross-border payments and ensure the legitimacy and security of transactions has become an urgent problem to be solved by the big data supervision system.

[0060] First, the embodiment of the present application regards each block regulatory target in cross-border payment (such as a cross-border payment transaction) as a hidden space feature unit. This hidden space feature unit not only contains the basic information of the transaction, but also establishes connections with other related units (such as other transactions, payment participants, etc.) through the hidden space feature relationship network. In this hidden space feature relationship network, each hidden space feature unit represents a regulated blockchain data entity or regulatory participant entity, which are connected to each other through relationship conduction pointers, forming a complex and orderly network.

[0061] Secondly, when the big data supervision system needs to supervise a cross-border payment transaction, they can use the latent space feature unit corresponding to the transaction as the key latent space feature unit. Through the latent space feature relationship network, the big data supervision system can quickly determine the upstream and downstream latent space feature units of the transaction, that is, other transactions or payment participants that are directly or indirectly involved in the transaction. This step is crucial for fully understanding the transaction background and identifying potential risks.

[0062] Then, the big data supervision system needs to obtain the text semantics of the supervision elements of these upstream and downstream latent space feature units. These text semantics of supervision elements include detailed information of transactions, identity authentication of payment participants, historical records of transactions, etc., which are important bases for risk assessment and decision-making by the big data supervision system. Through the technical solution described in the embodiment of this application, the big data supervision system can obtain this information comprehensively and accurately, laying a solid foundation for subsequent supervision work.

[0063] After obtaining the regulatory element text semantics of the key latent space feature units and their upstream and downstream latent space feature units, the big data supervision system can integrate these text semantics to obtain the derived regulatory element text semantics of this cross-border payment transaction. This derived regulatory element text semantics not only contains the information of the transaction itself, but also incorporates the information of other transactions and payment participants related to it, providing a comprehensive and multi-dimensional regulatory perspective for the big data supervision system.

[0064] Finally, the big data supervision system can obtain the target supervision element text semantics of this transaction based on the derived supervision element text semantics of this cross-border payment transaction and the supervision element text semantics of each upstream and downstream latent space feature unit. This target supervision element text semantics is an important basis for the big data supervision system to conduct risk assessment, decision-making and supervision execution. Through the technical solution described in the embodiment of this application, the big data supervision system can achieve comprehensive and accurate supervision of cross-border payment transactions, effectively prevent financial risks, and ensure the security and stability of cross-border payments.

[0065] 2. Intelligent Manufacturing Quality Supervision

[0066] In the field of smart manufacturing, the application of blockchain technology provides a new idea for product quality supervision. However, how to effectively use blockchain data to supervise the quality of smart manufacturing products and ensure product compliance and safety is a major challenge facing the big data supervision system.

[0067] First, the embodiment of the present application regards each block supervision target (such as a manufacturing batch, a component, etc.) in intelligent manufacturing as a latent space feature unit. This latent space feature unit not only contains the basic information of the product, but also establishes connections with other related units (such as other manufacturing batches, component suppliers, production equipment, etc.) through the latent space feature relationship network. In this latent space feature relationship network, each latent space feature unit represents a regulated blockchain data entity or regulatory participating entity, which are connected to each other through relationship conduction pointers, forming a complex and orderly network.

[0068] Secondly, when the big data supervision system needs to supervise the quality of a certain intelligent manufacturing product, they can use the latent space feature unit corresponding to this product as the key latent space feature unit. Through the latent space feature relationship network, the big data supervision system can quickly determine the upstream and downstream latent space feature units of this product, that is, other manufacturing batches, parts suppliers, production equipment, etc. that are directly or indirectly involved in the business of this product. This step is crucial for fully understanding the product manufacturing background and identifying potential quality risks.

[0069] Then, the big data supervision system needs to obtain the text semantics of the supervision elements of these upstream and downstream latent space feature units. These text semantics of supervision elements include product manufacturing information, identity authentication of parts suppliers, historical records of production equipment, etc., which are important bases for the big data supervision system to conduct quality risk assessment and decision-making. Through the technical solution described in the embodiment of this application, the big data supervision system can obtain this information comprehensively and accurately, laying a solid foundation for subsequent supervision work.

[0070] After obtaining the regulatory element text semantics of the key latent space feature units and their upstream and downstream latent space feature units, the big data supervision system can integrate these text semantics to obtain the derived regulatory element text semantics of this intelligent manufacturing product. This derived regulatory element text semantics not only contains the information of the product itself, but also incorporates the information of other related manufacturing batches, parts suppliers and production equipment, providing a comprehensive and multi-dimensional regulatory perspective for the big data supervision system.

[0071] Finally, the big data supervision system can obtain the target supervision element text semantics of this product based on the derived supervision element text semantics of this intelligent manufacturing product and the supervision element text semantics of each upstream and downstream latent space feature unit. This target supervision element text semantics is an important basis for the big data supervision system to conduct quality risk assessment, decision-making and supervision execution. Through the technical solution described in the embodiment of this application, the big data supervision system can achieve comprehensive and accurate supervision of intelligent manufacturing products, effectively prevent quality risks, and ensure the compliance and safety of intelligent manufacturing products.

[0072] In summary, this embodiment has demonstrated its unique advantages and wide application value in the two major application scenarios of cross-border payment blockchain supervision and intelligent manufacturing quality supervision. By introducing the concepts of latent space feature units and latent space feature relationship networks, the embodiment of this application realizes comprehensive, in-depth mining and intelligent processing of blockchain data, and provides a comprehensive, multi-dimensional supervision perspective and accurate and efficient supervision tools for the big data supervision system. This not only improves the comprehensiveness and timeliness of supervision, but also provides strong support and guarantee for the widespread application of blockchain technology.

[0073] In some preferred embodiments, the method of determining at least one upstream and downstream latent space feature unit of the key latent space feature unit based on the latent space feature relationship network includes: taking the key latent space feature unit as the latent space feature activation unit, and according to each type of risk control feature conduction strategy in at least one type of risk control feature conduction strategy, determining in the latent space feature relationship network a local feature relationship network of the key latent space feature unit corresponding to each type of the risk control feature conduction strategy, wherein the at least one upstream and downstream latent space feature unit includes all latent space feature units other than the key latent space feature unit in the local feature relationship network corresponding to each type of the risk control feature conduction strategy; wherein each type of the risk control feature conduction strategy corresponds to at least two latent space feature units, and the state of each latent space feature unit in the at least two latent space feature units and the involvement situation between the latent space feature units are pre-configured.

[0074] Furthermore, in the two major application scenarios of cross-border payment blockchain supervision and intelligent manufacturing quality supervision, how to effectively determine the upstream and downstream latent space feature units of key latent space feature units is the key to achieving comprehensive and precise supervision.

[0075] 1. Application of blockchain in cross-border payment supervision

[0076] In the cross-border payment blockchain supervision, each block supervision target, such as a cross-border payment transaction, is regarded as a hidden space feature unit. This hidden space feature unit not only contains the basic information of the transaction, but also establishes connections with other related units through the hidden space feature relationship network. In order to achieve comprehensive supervision of cross-border payment transactions, the big data supervision system needs to determine the upstream and downstream hidden space feature units of the key hidden space feature unit (i.e. a specific cross-border payment transaction).

[0077] First, the big data supervision system uses the key latent space feature unit as the latent space feature activation unit, which is the starting point of the entire risk control feature transmission process. Then, according to at least one type of risk control feature transmission strategy set, the big data supervision system begins to search for other latent space feature units that have direct or indirect business involvement with the key latent space feature unit in the latent space feature relationship network.

[0078] Each type of risk control feature transmission strategy corresponds to a specific business involvement model. For example, one type of risk control feature transmission strategy may focus on the flow of funds in the transaction, while another type may focus on the identity authentication of the transaction participants. Therefore, in the latent space feature relationship network, each type of risk control feature transmission strategy will determine a local feature relationship network, which contains all the related latent space feature units with the key latent space feature unit in a specific business involvement model.

[0079] In addition to the key latent space feature units themselves, other latent space feature units in the local feature relationship network corresponding to various risk control feature transmission strategies are the upstream and downstream latent space feature units that the big data supervision system needs to determine. These upstream and downstream latent space feature units may include other cross-border payment transactions, payment participants, fund recipients, etc., which are directly or indirectly involved in the business with the key latent space feature units.

[0080] In this process, each type of risk control feature transmission strategy corresponds to at least two latent space feature units, and the states of these latent space feature units and the situations involved between them are pre-configured. This means that the big data supervision system can flexibly define and adjust the risk control feature transmission strategy according to actual needs to adapt to different supervision scenarios and needs.

[0081] In this way, the big data supervision system can comprehensively and accurately determine the upstream and downstream latent space feature units of key latent space feature units, thereby achieving comprehensive and accurate supervision of cross-border payment transactions. This not only improves the comprehensiveness and timeliness of supervision, but also provides a strong guarantee for the security and stability of cross-border payments.

[0082] 2. Application of Intelligent Manufacturing Quality Supervision

[0083] In the quality supervision of intelligent manufacturing, each block supervision target, such as a manufacturing batch, a component, etc., is also regarded as a latent space feature unit. Similar to the cross-border payment blockchain supervision, this latent space feature unit also contains the basic information of the product and establishes connections with other related units through the latent space feature relationship network. In order to achieve comprehensive quality supervision of intelligent manufacturing products, the big data supervision system also needs to determine the upstream and downstream latent space feature units of the key latent space feature unit (i.e. a specific manufacturing batch or component).

[0084] First, the big data supervision system also uses the key latent space feature unit as the latent space feature activation unit to start the entire risk control feature transmission process. Then, according to at least one type of risk control feature transmission strategy set, the big data supervision system searches for other latent space feature units that have direct or indirect business involvement with the key latent space feature unit in the latent space feature relationship network.

[0085] In smart manufacturing quality supervision, each type of risk control feature transmission strategy may focus on different quality risk factors. For example, one type of risk control feature transmission strategy may focus on the source and quality of raw materials, while another type may focus on the status and maintenance records of production equipment. Therefore, in the latent space feature relationship network, each type of risk control feature transmission strategy will determine a local feature relationship network, which contains all the related latent space feature units under specific quality risk factors with the key latent space feature units.

[0086] In addition to the key latent space feature units themselves, other latent space feature units in the local feature relationship network corresponding to various risk control feature transmission strategies are the upstream and downstream latent space feature units that the big data supervision system needs to determine. These upstream and downstream latent space feature units may include other manufacturing batches, parts suppliers, production equipment, etc., which are directly or indirectly involved in quality risk factors with key latent space feature units.

[0087] Similarly, in this process, each type of risk control feature transmission strategy corresponds to at least two latent space feature units, and the states of these latent space feature units and the involvement between them are also pre-configured. This means that the big data supervision system can flexibly define and adjust the risk control feature transmission strategy according to actual needs to adapt to different smart manufacturing products and supervision scenarios.

[0088] In this way, the big data supervision system can comprehensively and accurately determine the upstream and downstream latent space feature units of key latent space feature units, thereby achieving comprehensive and accurate quality supervision of intelligent manufacturing products. This not only improves the comprehensiveness and timeliness of quality supervision, but also provides a strong guarantee for the compliance and safety of intelligent manufacturing products.

[0089] In this way, by introducing the concepts of latent space feature units and latent space feature relationship networks, and combining the implementation method of risk control feature transmission strategy, the embodiment of the present application realizes comprehensive and in-depth mining and intelligent processing of blockchain data.

[0090] In some other preferred embodiments, the target regulatory element text semantics of the first block regulatory target is obtained based on the derived regulatory element text semantics of the first block regulatory target and the regulatory element text semantics of each upstream and downstream latent space feature unit, including: for each type of the risk control feature transmission strategy, the integrated regulatory element text semantics of the first block regulatory target corresponding to the type of risk control feature transmission strategy is obtained by integrating the regulatory element text semantics of each upstream and downstream latent space feature unit in the local feature relationship network corresponding to the type of risk control feature transmission strategy, and the derived regulatory element text semantics of the first block regulatory target; the target regulatory element text semantics of the first block regulatory target is obtained by integrating the integrated regulatory element text semantics of the first block regulatory target corresponding to each type of the risk control feature transmission strategy.

[0091] In actual implementation, how to accurately obtain the text semantics of the target regulatory elements of the first block's regulatory objectives is the key to achieving effective supervision.

[0092] 1. Application of blockchain in cross-border payment supervision

[0093] In cross-border payment blockchain supervision, each block supervision target, such as a cross-border payment transaction, has its own specific derivative supervision element text semantics. These derivative supervision element text semantics contain key data such as basic transaction information, identity information of both parties to the transaction, and transaction amount. However, in order to have a more comprehensive understanding of the risk status of this transaction, the big data supervision system also needs to obtain the supervision element text semantics of its upstream and downstream latent space feature units.

[0094] First, for each type of risk control feature transmission strategy, the big data supervision system needs to integrate the textual semantics of the regulatory elements of each upstream and downstream latent space feature unit in the local feature relationship network corresponding to this type of strategy. For example, if a type of risk control feature transmission strategy focuses on the flow of funds in a transaction, then the big data supervision system needs to integrate the textual semantics of the regulatory elements of all other transactions that have capital flows involved in this transaction. The textual semantics of the regulatory elements of these upstream and downstream transactions may include transaction time, transaction amount, account information of both parties to the transaction, etc.

[0095] During the integration process, the big data supervision system can use text semantic vectorization technology to convert the text semantics of regulatory elements into numerical feature vectors. For example, for an upstream transaction, the text semantics of its regulatory elements can be converted into a numerical vector containing features such as transaction time, transaction amount, and account information of both parties to the transaction. By integrating the numerical feature vectors of all upstream and downstream transactions, the big data supervision system can obtain an integrated regulatory element text semantic vector that comprehensively reflects the risk of the flow of funds in this transaction.

[0096] Secondly, the big data supervision system needs to integrate the integrated supervision element text semantics corresponding to the first block supervision target and various risk control feature transmission strategies. This means that in addition to the risk of capital flow, the big data supervision system also needs to consider other types of risks, such as the credit risk of both parties to the transaction, the regional risk involved in the transaction, etc. For each type of risk, the big data supervision system can obtain an integrated supervision element text semantic vector.

[0097] Finally, by further integrating and fusing these integrated regulatory element text semantic vectors, the big data regulatory system can obtain the target regulatory element text semantics of the first block regulatory target. This target regulatory element text semantics is a numerical feature vector that comprehensively reflects the various risks of this cross-border payment transaction, which can help the big data regulatory system to more accurately judge the risk status of this transaction and take corresponding regulatory measures.

[0098] 2. Application of Intelligent Manufacturing Quality Supervision

[0099] In smart manufacturing quality supervision, each block supervision target, such as a manufacturing batch or component, also has its specific derived supervision element text semantics. These derived supervision element text semantics contain key data such as basic product information, production batch, and production date. However, in order to have a more comprehensive understanding of the quality status of this manufacturing batch or component, the big data supervision system also needs to obtain the supervision element text semantics of its upstream and downstream latent space feature units.

[0100] First, for each type of risk control feature transmission strategy, the big data supervision system also needs to integrate the textual semantics of the regulatory elements of each upstream and downstream latent space feature unit in the local feature relationship network corresponding to this type of strategy. For example, if a type of risk control feature transmission strategy focuses on the quality risk of raw materials, then the big data supervision system needs to integrate the textual semantics of the regulatory elements of all other batches that provide raw materials for this manufacturing batch. The textual semantics of the regulatory elements of these upstream and downstream batches may include raw material batch numbers, raw material supplier information, raw material quality test results, etc.

[0101] Similarly, during the integration process, the big data supervision system can also use text semantic vectorization technology to convert the text semantics of supervision elements into numerical feature vectors. For example, for an upstream raw material batch, its supervision element text semantics can be converted into a numerical vector containing features such as raw material batch number, supplier information, and quality inspection results. By integrating the numerical feature vectors of all upstream and downstream raw material batches, the big data supervision system can obtain an integrated supervision element text semantic vector that comprehensively reflects the quality risk of raw materials in this manufacturing batch.

[0102] Secondly, the big data supervision system also needs to integrate the integrated supervision element text semantics corresponding to various risk control feature transmission strategies of the first block supervision target. This means that in addition to the raw material quality risk, the big data supervision system also needs to consider other types of risks, such as the status risk of production equipment and the stability risk of production process. For each type of risk, the big data supervision system can obtain an integrated supervision element text semantic vector.

[0103] Finally, by further integrating and fusing these integrated supervision element text semantic vectors, the big data supervision system can obtain the target supervision element text semantics of the first block supervision target. This target supervision element text semantics is a numerical feature vector that comprehensively reflects the various quality risks of this manufacturing batch or component, which can help the big data supervision system to more accurately judge the quality status of this manufacturing batch or component and take corresponding supervision measures.

[0104] For example, in cross-border payment blockchain supervision, the text semantics of the target regulatory element can be represented by a numerical feature vector, which contains numerical features reflecting various risks of cross-border payment transactions. The following is an example of a specific numerical feature vector: [transaction amount: 10000, transaction timestamp: 1672531200, transaction country risk rating: [2, 1], transaction type risk coefficient: 0.5, fund flow abnormality score: 0.2, ...].

[0105] In this numerical feature vector, the transaction amount indicates the amount of the cross-border payment transaction; the transaction timestamp indicates the time when the transaction occurred; the country risk rating of the two parties to the transaction is a list containing two elements, which respectively indicate the risk rating of the countries where the two parties to the transaction are located. The larger the value, the higher the risk; the transaction type risk coefficient indicates the risk level of this type of transaction; the abnormal score of the flow of funds indicates the abnormal degree of the flow of funds in the transaction. The larger the value, the higher the possibility of abnormality. These numerical features together constitute the text semantics of the target regulatory element that reflects the risk status of cross-border payment transactions.

[0106] 2. Numerical feature vectors in intelligent manufacturing quality supervision

[0107] In smart manufacturing quality supervision, the text semantics of the target supervision element can also be represented by a numerical feature vector, which contains numerical features that reflect various quality risks of manufacturing batches or parts. The following is a specific example of a numerical feature vector: [Raw material batch number: "ABC123", Raw material supplier risk rating: 3, Raw material quality test results: [0.95, 0.98, 0.96], Production equipment status score: 0.85, Production process stability score: 0.9, Product quality test pass rate: 0.97, ...].

[0108] In this numerical feature vector, the raw material batch number indicates the raw material batch used in the manufacturing batch; the raw material supplier risk rating indicates the risk level of the raw material supplier, and the larger the value, the higher the risk; the raw material quality test result is a list containing multiple elements, which respectively indicate the quality test results of different raw materials, and the closer the value is to 1, the better the quality; the production equipment status score indicates the state of the production equipment, and the larger the value, the better the state; the production process stability score indicates the stability of the production process, and the larger the value, the more stable it is; the product quality inspection pass rate indicates the quality inspection pass rate of the manufacturing batch or parts, and the larger the value, the better the quality. These numerical features together constitute the text semantics of the target regulatory element that reflects the quality risk status of the manufacturing batch or parts.

[0109] Through specific numerical feature vector examples, we can more clearly understand the application of target regulatory element text semantics in cross-border payment blockchain supervision and smart manufacturing quality supervision. These numerical feature vectors not only contain key regulatory element information, but also enable big data supervision systems to conduct risk analysis and judgment more conveniently through numerical methods.

[0110] In other examples, the local feature relationship network corresponding to each type of the risk control feature transmission strategy includes x layers of upstream and downstream latent space feature units of the key latent space feature unit, wherein x is an integer greater than or equal to 1, and the first layer of upstream and downstream latent space feature units are upstream and downstream latent space feature units having real-time relationship transmission pointers with the key latent space feature units; for each type of the risk control feature transmission strategy, the integrated regulatory element text semantics of the first block supervision target corresponding to the type of risk control feature transmission strategy is obtained based on the following steps: the regulatory element text semantics of the x-th layer of upstream and downstream latent space feature units in the local feature relationship network corresponding to the type of risk control feature transmission strategy is taken as the integrated regulatory element text semantics of the x-th layer of upstream and downstream latent space feature units, the key latent space feature unit is taken as the 0th layer of upstream and downstream latent space feature units, and according to the layer decreasing rule, the first block supervision target corresponds to the integrated regulatory element text semantics of the x-th layer of upstream and downstream latent space feature units. The loop implements the following steps until the integrated regulatory element text semantics of the key latent space feature unit corresponding to the risk control feature transmission strategy of this type is obtained: for each upstream and downstream latent space feature unit u in the y-1th layer, determine each upstream and downstream latent space feature unit v in the upstream and downstream latent space feature units of the yth layer that has a relationship transmission pointer with the upstream and downstream latent space feature unit u, and obtain the integrated regulatory element text semantics of the upstream and downstream latent space feature unit u by integrating the regulatory element text semantics of the upstream and downstream latent space feature unit u and the regulatory element text semantics of each upstream and downstream latent space feature unit v; wherein y is an integer greater than or equal to 1 and less than or equal to x, and when the upstream and downstream latent space feature unit u is a key latent space feature unit, the regulatory element text semantics of the upstream and downstream latent space feature unit u is the derived regulatory element text semantics of the key latent space feature unit.

[0111] It can be understood that in the two major scenarios of cross-border payment blockchain supervision and smart manufacturing quality supervision, how to accurately obtain the integrated supervision element text semantics of the first block supervision target corresponding to various risk control feature transmission strategies is the key to achieving effective supervision.

[0112] 1. Application of blockchain in cross-border payment supervision

[0113] In cross-border payment blockchain supervision, each block supervision target, such as a cross-border payment transaction, has its specific derivative supervision element text semantics. However, in order to have a more comprehensive understanding of the risk status of this transaction, the big data supervision system also needs to obtain the supervision element text semantics of its upstream and downstream latent space feature units and consider the upstream and downstream relationships at different levels.

[0114] First, for each type of risk control feature transmission strategy, the big data supervision system needs to determine the key latent space feature units in the local feature relationship network corresponding to this type of strategy, which are usually entities directly related to cross-border payment transactions, such as transaction parties, transaction banks, etc. The derived supervision element text semantics of the key latent space feature unit includes the basic information of the entity, historical transaction records and other key data.

[0115] Next, the big data supervision system uses the key latent space feature units as the 0th layer upstream and downstream latent space feature units, and determines the first layer upstream and downstream latent space feature units, that is, the upstream and downstream latent space feature units with real-time relationship transmission pointers with the key latent space feature units. These units may include entities that have other transaction relationships with the transaction parties, other customers of the transaction bank, etc. The text semantics of the supervision elements of the first layer upstream and downstream latent space feature units include the basic information of these entities, their relationship with the key latent space feature units, etc.

[0116] Then, according to the layer-decreasing rule, the big data supervision system cyclically implements the following steps until the integrated supervision element text semantics of the key latent space feature unit corresponding to the risk control feature transmission strategy of this type is obtained: for each upstream and downstream latent space feature unit u in the y-1th layer, determine the upstream and downstream latent space feature units v in the yth layer that have relationship transmission pointers with the upstream and downstream latent space feature unit u. These units v may include entities that have other transaction relationships with u, other business partners of u, etc. By integrating the supervision element text semantics of the upstream and downstream latent space feature unit u and the supervision element text semantics of each upstream and downstream latent space feature unit v, the big data supervision system can obtain the integrated supervision element text semantics of the upstream and downstream latent space feature unit u.

[0117] In this process, the big data supervision system can use text semantic vectorization technology to convert the text semantics of regulatory elements into numerical feature vectors. For example, for an upstream transaction (as the upstream and downstream latent space feature unit v), its regulatory element text semantics can be converted into a numerical vector containing transaction time, transaction amount, account information of both parties to the transaction, and other features. By integrating the numerical feature vectors of all upstream and downstream transactions, the big data supervision system can obtain an integrated regulatory element text semantic vector that comprehensively reflects the risks at this level.

[0118] Finally, when the upstream and downstream latent space feature unit u is a key latent space feature unit, its regulatory element text semantics is the derived regulatory element text semantics of the key latent space feature unit. Through continuous iteration and integration, the big data supervision system can eventually obtain the integrated regulatory element text semantics of the key latent space feature unit corresponding to this type of risk control feature transmission strategy.

[0119] Taking a specific cross-border payment transaction as an example, the key latent space feature unit is one of the two parties in the transaction (such as the buyer), and its derived regulatory element text semantics includes the buyer's basic information, historical transaction records, etc. By determining the upstream and downstream latent space feature units at different levels and integrating their regulatory element text semantics, the big data supervision system can obtain an integrated regulatory element text semantic vector that fully reflects the risk status of the transaction. This vector may include numerical features of multiple dimensions such as the buyer's transaction behavior characteristics, the risk status of the buyer's transaction partners, and the regional risks involved in the transaction.

[0120] 2. Application of Intelligent Manufacturing Quality Supervision

[0121] In smart manufacturing quality supervision, each block supervision target, such as a manufacturing batch or component, also has its specific derived supervision element text semantics. However, in order to more comprehensively understand the quality status of this manufacturing batch or component, the big data supervision system also needs to obtain the supervision element text semantics of its upstream and downstream latent space feature units and consider the upstream and downstream relationships at different levels.

[0122] First, for each type of risk control feature transmission strategy, the big data supervision system needs to determine the key latent space feature units in the local feature relationship network corresponding to this type of strategy, which are usually entities directly related to manufacturing batches or parts, such as production equipment, raw material suppliers, etc. The derived supervision element text semantics of the key latent space feature unit contains key data such as the basic information of the entity, historical production or supply records, etc.

[0123] Next, the big data supervision system uses the key latent space feature units as the 0th layer upstream and downstream latent space feature units, and determines the first layer upstream and downstream latent space feature units, that is, the upstream and downstream latent space feature units with real-time relationship transmission pointers with the key latent space feature units. These units may include entities with other production relationships with production equipment, other customers of raw material suppliers, etc. The text semantics of the supervision elements of the first layer upstream and downstream latent space feature units include the basic information of these entities, their relationship with the key latent space feature units, etc.

[0124] Then, according to the layer-decreasing rule, the big data supervision system cyclically implements the following steps until the integrated supervision element text semantics of the key latent space feature unit corresponding to the risk control feature transmission strategy of this type is obtained: for each upstream and downstream latent space feature unit u in the y-1th layer, determine the upstream and downstream latent space feature units v in the yth layer that have relationship transmission pointers with the upstream and downstream latent space feature unit u. These units v may include entities that have other production relationships with u, other suppliers of u, etc. By integrating the supervision element text semantics of the upstream and downstream latent space feature unit u and the supervision element text semantics of each upstream and downstream latent space feature unit v, the big data supervision system can obtain the integrated supervision element text semantics of the upstream and downstream latent space feature unit u.

[0125] In this process, the big data supervision system can also use text semantic vectorization technology to convert the text semantics of regulatory elements into numerical feature vectors. For example, for an upstream raw material supplier (as the upstream and downstream latent space feature unit v), its regulatory element text semantics can be converted into a numerical vector containing supplier information, raw material quality test results, historical supply records and other features. By integrating the numerical feature vectors of all upstream and downstream raw material suppliers, the big data supervision system can obtain an integrated regulatory element text semantic vector that comprehensively reflects the quality risks at this level.

[0126] Finally, when the upstream and downstream latent space feature unit u is a key latent space feature unit, its regulatory element text semantics is the derived regulatory element text semantics of the key latent space feature unit. Through continuous iteration and integration, the big data supervision system can eventually obtain the integrated regulatory element text semantics of the key latent space feature unit corresponding to this type of risk control feature transmission strategy.

[0127] Taking a specific manufacturing batch as an example, the key latent space feature unit is the production equipment, and its derived regulatory element text semantics includes the basic information of the equipment, historical production records, etc. By determining the upstream and downstream latent space feature units at different levels and integrating their regulatory element text semantics, the big data supervision system can obtain an integrated regulatory element text semantic vector that fully reflects the quality status of the manufacturing batch. This vector may include numerical features of multiple dimensions such as the status characteristics of the production equipment, the quality status of the raw materials, and the stability of the production process.

[0128] In some other preferred embodiments, the integrated regulatory element text semantics of the upstream and downstream latent space feature unit u is obtained by integrating the regulatory element text semantics of the upstream and downstream latent space feature unit u and the regulatory element text semantics of each of the upstream and downstream latent space feature units v, including: determining the reinforcement coefficient corresponding to each of the upstream and downstream latent space feature units v based on the semantic correlation between the regulatory element text semantics of the upstream and downstream latent space feature unit u and the regulatory element text semantics of each of the upstream and downstream latent space feature units v; integrating the regulatory element text semantics of each of the upstream and downstream latent space feature units v according to the reinforcement coefficient corresponding to each of the upstream and downstream latent space feature units v, integrating the integrated result with the regulatory element text semantics of the upstream and downstream latent space feature unit u, and obtaining the integrated regulatory element text semantics of the upstream and downstream latent space feature unit u.

[0129] In cross-border payment blockchain supervision, each block supervision target, such as a cross-border payment transaction, involves multiple upstream and downstream hidden space feature units, including the transaction parties, transaction banks, intermediary institutions, etc. These units each have unique regulatory element text semantics, such as basic information, historical transaction records, credit ratings, etc. In order to comprehensively assess the risk status of a transaction, the big data supervision system needs to integrate the regulatory element text semantics of these upstream and downstream hidden space feature units.

[0130] First, the big data supervision system needs to determine the text semantics of the supervision elements of each upstream and downstream latent space feature unit u. This usually involves extracting relevant information from the blockchain and performing necessary text processing and semantic analysis. For example, for both parties to a transaction, the big data supervision system can extract their identity information, historical transaction records, credit ratings, etc., and convert them into text semantic vectors.

[0131] Next, the big data supervision system needs to analyze the semantic association between the upstream and downstream latent space feature units u and their upstream and downstream latent space feature units v. This can be achieved by calculating the similarity between text semantic vectors, such as using cosine similarity, Euclidean distance and other methods. The semantic association reflects the degree of association between different units in terms of supervision elements and is an important basis for determining the reinforcement coefficient.

[0132] Then, based on the semantic relevance, the big data supervision system assigns a reinforcement coefficient to each upstream and downstream latent space feature unit v. The size of the reinforcement coefficient reflects the degree of influence of the unit on the risk status of the upstream and downstream latent space feature unit u. For example, if a transaction bank has a long-term cooperative relationship with both parties to the transaction and has a good historical transaction record, then the transaction bank may have a greater impact on the risk status of both parties to the transaction, and therefore will be assigned a higher reinforcement coefficient.

[0133] Next, the big data supervision system integrates the text semantics of the supervision elements of each upstream and downstream latent space feature unit v according to the reinforcement coefficient. This can be achieved through weighted summation, weighted average and other methods. The integrated result reflects the comprehensive impact of each upstream and downstream latent space feature unit on the risk status of the upstream and downstream latent space feature unit u.

[0134] Finally, the big data supervision system integrates the integrated results with the text semantics of the supervision elements of the upstream and downstream latent space feature units u to obtain the integrated text semantics of the supervision elements of the upstream and downstream latent space feature units u. This integrated text semantic vector contains the supervision element information of the upstream and downstream latent space feature units u and all its upstream and downstream latent space feature units, providing a comprehensive and in-depth basis for transaction risk assessment for the big data supervision system.

[0135] Taking a specific cross-border payment transaction as an example, the big data supervision system is evaluating the risk status of a cross-border payment transaction initiated by the buyer. First, the big data supervision system extracts the text semantics of the buyer's regulatory elements, including identity information, historical transaction records, credit ratings, etc. Then, the big data supervision system analyzes the semantic association between the buyer and its upstream and downstream latent space feature units such as transaction banks, sellers, and intermediaries, and assigns reinforcement coefficients to them. For example, if the buyer's transaction bank has a good historical transaction record and has a long-term cooperative relationship with the buyer, then the transaction bank will be assigned a higher reinforcement coefficient. Next, the big data supervision system integrates the text semantics of the regulatory elements of each upstream and downstream latent space feature unit according to the reinforcement coefficient to obtain a text semantic vector that comprehensively reflects the transaction risk status. Finally, the big data supervision system integrates this integrated text semantic vector with the buyer's regulatory element text semantics to obtain the integrated regulatory element text semantics of the buyer and all its upstream and downstream latent space feature units, providing a comprehensive and in-depth basis for transaction risk assessment for the big data supervision system.

[0136] In smart manufacturing quality supervision, each block supervision target, such as a manufacturing batch or component, also involves multiple upstream and downstream latent space feature units, including production equipment, raw material suppliers, production processes, etc. These units also have unique supervision element text semantics, such as equipment status, raw material quality, production process parameters, etc. In order to comprehensively evaluate the quality status of a manufacturing batch or component, the big data supervision system also needs to integrate the supervision element text semantics of these upstream and downstream latent space feature units.

[0137] First, the big data supervision system needs to determine the text semantics of the supervision elements of each upstream and downstream latent space feature unit u. This usually involves extracting relevant information from systems such as the manufacturing execution system (MES) and enterprise resource planning (ERP), and performing necessary text processing and semantic analysis. For example, for production equipment, the big data supervision system can extract its equipment status, historical production records, maintenance records, etc., and convert them into text semantic vectors.

[0138] Next, the big data supervision system also needs to analyze the semantic correlation between the upstream and downstream latent space feature units u and their upstream and downstream latent space feature units v. This can also be achieved by calculating the similarity between text semantic vectors. The semantic correlation reflects the degree of correlation between different units in terms of supervision elements, and is also an important basis for determining the reinforcement coefficient.

[0139] Then, based on the semantic relevance, the big data supervision system assigns a reinforcement coefficient to each upstream and downstream latent space feature unit v. The size of the reinforcement coefficient reflects the degree of influence of the unit on the quality status of the upstream and downstream latent space feature unit u. For example, if the quality of raw materials provided by a raw material supplier has been stable and meets the standards, then the supplier may have a greater impact on the quality status of the manufacturing batch or parts, and therefore will be assigned a higher reinforcement coefficient.

[0140] Next, the big data supervision system also integrates the text semantics of the supervision elements of each upstream and downstream latent space feature unit v according to the reinforcement coefficient. This can also be achieved through weighted summation, weighted average and other methods. The integrated result reflects the comprehensive impact of each upstream and downstream latent space feature unit on the quality status of the upstream and downstream latent space feature unit u.

[0141] Finally, the big data supervision system integrates the integrated results with the text semantics of the supervision elements of the upstream and downstream latent space feature units u to obtain the integrated text semantics of the supervision elements of the upstream and downstream latent space feature units u. This integrated text semantic vector contains the supervision element information of the upstream and downstream latent space feature units u and all its upstream and downstream latent space feature units, providing a comprehensive and in-depth basis for manufacturing quality assessment for the big data supervision system.

[0142] Taking a specific manufacturing batch as an example, the big data supervision system is evaluating the quality status of a manufacturing batch produced by a certain production equipment. First, the big data supervision system extracts the text semantics of the supervision elements of the production equipment, including equipment status, historical production records, maintenance records, etc. Then, the big data supervision system analyzes the semantic association between the production equipment and its upstream and downstream latent space feature units such as raw material suppliers, production processes, and quality inspections, and assigns reinforcement coefficients to them. For example, if the quality of raw materials provided by the raw material supplier has been stable and meets the standards, then the supplier will be assigned a higher reinforcement coefficient. Next, the big data supervision system integrates the text semantics of the supervision elements of each upstream and downstream latent space feature unit according to the reinforcement coefficient to obtain a text semantic vector that comprehensively reflects the quality status of the manufacturing batch. Finally, the big data supervision system integrates this integrated text semantic vector with the text semantics of the supervision elements of the production equipment to obtain the integrated text semantics of the supervision elements of the production equipment and all its upstream and downstream latent space feature units, providing a comprehensive and in-depth basis for manufacturing quality evaluation for the big data supervision system.

[0143] In another possible embodiment, the target regulatory element text semantics of the first block regulatory target are obtained by integrating the integrated regulatory element text semantics corresponding to each type of the risk control feature transmission strategy of the first block regulatory target, including: determining the reinforcement coefficient corresponding to each type of the risk control feature transmission strategy according to the semantic correlation between the integrated regulatory element text semantics of each type of the risk control feature transmission strategy; integrating the integrated regulatory element text semantics of each type of the risk control feature transmission strategy according to the reinforcement coefficient corresponding to each type of the risk control feature transmission strategy to obtain the target regulatory element text semantics of the first block regulatory target.

[0144] In the complex environment of cross-border payment blockchain supervision, each block supervision target, such as a cross-border payment transaction, may involve multiple risk control feature transmission strategies, which cover multiple aspects such as identity authentication of both parties to the transaction, rationality analysis of the transaction amount, and security assessment of the transaction path. Each risk control feature transmission strategy corresponds to a unique set of regulatory element text semantics, which contain rich regulatory information, such as the historical transaction records, credit ratings, transaction amounts and frequencies of both parties to the transaction.

[0145] In order to comprehensively and deeply assess the risk status of a cross-border payment transaction, the big data supervision system needs to integrate the text semantics of the regulatory elements scattered in various risk control feature transmission strategies to form a unified and comprehensive target regulatory element text semantics. This process requires not only that the big data supervision system can accurately extract and parse the text semantics of the regulatory elements in each risk control feature transmission strategy, but also that it can understand and grasp the internal connections and mutual influences between these text semantics.

[0146] First, the big data supervision system needs to determine the reinforcement coefficients corresponding to each risk control feature transmission strategy based on the semantic correlation between the text semantics of the integrated supervision elements of each risk control feature transmission strategy. The key to this step is to accurately calculate and evaluate the semantic correlation between each risk control feature transmission strategy, that is, the correlation and mutual influence between them. For example, if a risk control feature transmission strategy focuses on the identity authentication of both parties to the transaction, while another strategy focuses on the rationality analysis of the transaction amount, then the semantic correlation between the two strategies may be relatively low; but if both strategies involve the historical transaction records or credit ratings of both parties to the transaction, then the semantic correlation between them may be higher.

[0147] Secondly, the big data supervision system needs to integrate the integrated supervision element text semantics of various risk control feature transmission strategies according to the corresponding reinforcement coefficients of various risk control feature transmission strategies. The purpose of this step is to organically integrate and merge the text semantics of the supervision elements scattered in various risk control feature transmission strategies to form a unified and comprehensive target supervision element text semantics. In the integration process, the big data supervision system needs to fully consider the semantic correlation and mutual influence between various risk control feature transmission strategies to ensure the accuracy and effectiveness of the integration results. For example, for risk control feature transmission strategies with high semantic correlation, the big data supervision system can give higher reinforcement coefficients to fully reflect their importance and influence in the text semantics of the target supervision element; while for strategies with low semantic correlation, lower reinforcement coefficients can be given to avoid their interference and misleading of the results during the integration process.

[0148] Finally, through the integration and processing of the above steps, the big data supervision system can obtain the text semantics of the target supervision element of the first block supervision target. This target supervision element text semantics not only contains the supervision element information in each risk control feature transmission strategy, but also reflects the inherent connection and mutual influence between these elements. It provides a comprehensive and in-depth supervision perspective for the big data supervision system, enabling it to more accurately assess the risk status of cross-border payment transactions and formulate corresponding supervision measures and strategies.

[0149] Taking a specific cross-border payment transaction as an example, the big data supervision system can first extract and analyze the text semantics of the regulatory elements of multiple risk control feature transmission strategies involved in the transaction, such as identity authentication, transaction amount analysis, and transaction path evaluation. Then, according to the semantic correlation between these text semantics, the reinforcement coefficients corresponding to each strategy are determined. Next, these reinforcement coefficients are used to integrate the text semantics of the regulatory elements of each strategy to obtain a unified and comprehensive target regulatory element text semantics. Finally, the big data supervision system can evaluate the risk status of the cross-border payment transaction based on this target regulatory element text semantics, and formulate corresponding regulatory measures and strategies.

[0150] In the field of smart manufacturing quality supervision, each block supervision target, such as a manufacturing batch or component, may also involve multiple risk control feature transmission strategies. These strategies cover multiple aspects such as quality control of raw materials, monitoring of production processes, and testing of product quality. Each risk control feature transmission strategy corresponds to a unique set of regulatory element text semantics, which contain rich manufacturing quality information, such as quality certification of raw materials, parameter settings of production processes, and product quality test reports.

[0151] In order to comprehensively and deeply evaluate the quality status of a manufacturing batch or component, the big data supervision system also needs to integrate the text semantics of the supervision elements scattered in various risk control feature transmission strategies to form a unified and comprehensive target supervision element text semantics. This process is similar to the application in cross-border payment blockchain supervision, and also requires the big data supervision system to accurately extract and parse the text semantics of the supervision elements in each risk control feature transmission strategy, and understand and grasp the internal connection and mutual influence between these text semantics.

[0152] First, the big data supervision system needs to determine the reinforcement coefficients corresponding to various risk control feature conduction strategies based on the semantic correlation between the text semantics of the integrated supervision elements of various risk control feature conduction strategies. In the quality supervision of intelligent manufacturing, the semantic correlation between different risk control feature conduction strategies may be more complex and diverse. For example, the quality control strategy of raw materials may have a high semantic correlation with the monitoring strategy of the production process, because they are both directly related to the quality of the final product; while the product quality detection strategy may have a low semantic correlation with the former two, because it is more of a quality assessment after the product is produced.

[0153] Secondly, the big data supervision system needs to integrate the text semantics of the integrated supervision elements of various risk control feature transmission strategies according to the corresponding reinforcement coefficients of each risk control feature transmission strategy. In the integration process, the big data supervision system needs to fully consider the characteristics and requirements of intelligent manufacturing quality supervision to ensure the accuracy and effectiveness of the integration results. For example, for risk control feature transmission strategies that are directly related to the quality of the final product, the big data supervision system can give a higher reinforcement coefficient; while for strategies that are indirectly related to product quality, a lower reinforcement coefficient can be given.

[0154] Finally, through the integration and processing of the above steps, the big data supervision system can obtain the target supervision element text semantics of the first block supervision target. This target supervision element text semantics not only contains the supervision element information in each risk control feature transmission strategy, but also reflects the inherent connection and mutual influence between these elements. It provides a comprehensive and in-depth supervision perspective for the big data supervision system, enabling it to more accurately evaluate the quality status of smart manufacturing batches or parts, and formulate corresponding supervision measures and strategies.

[0155] Taking a specific manufacturing batch as an example, the big data supervision system can first extract and analyze the text semantics of the regulatory elements of multiple risk control feature transmission strategies involved in the batch, such as raw material quality control, production process monitoring, and product quality testing. Then, based on the semantic correlation between these text semantics, the reinforcement coefficients corresponding to each strategy are determined. Next, these reinforcement coefficients are used to integrate the text semantics of the regulatory elements of each strategy to obtain a unified and comprehensive target regulatory element text semantics. Finally, the big data supervision system can evaluate the quality status of the manufacturing batch based on this target regulatory element text semantics and formulate corresponding regulatory measures and strategies.

[0156] In an independent embodiment, the target regulatory element text semantics of the first block regulatory target are obtained based on the derived regulatory element text semantics of the block regulatory target index keywords of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target, including: obtaining the derived regulatory element text semantics of the first block regulatory target by integrating the derived regulatory element text semantics of the block regulatory target index keywords of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target; extracting the target regulatory element text semantics of the first block regulatory target through a deep loop model based on the derived regulatory element text semantics of the first block regulatory target; wherein the deep loop model is the second long short-term memory algorithm after debugging, or a long short-term memory algorithm obtained by secondary debugging of the debugged second long short-term memory algorithm.

[0157] Furthermore, the first training data also includes the original regulatory element text semantics of the third block regulatory target, and the third block regulatory target corresponding to the first training data is a regulatory participating entity that has a regulatory contract trigger tag for the second block regulatory target or does not have a regulatory contract trigger tag for the second block regulatory target; wherein the regulatory decision processing algorithm and the deep loop model are obtained based on the multiple first training data by repeatedly implementing debugging processing until the debugging compliance requirements are met, and the debugging processing includes the following steps: the original regulatory element text semantics of the second block regulatory target in each of the first training data are respectively entered into the first long short-term memory algorithm to obtain the first regulatory element text semantics of the block regulatory target index keyword of each of the second block regulatory targets; for each of the second block regulatory targets, the second regulatory element of the second block regulatory target is obtained by integrating the first regulatory element text semantics of the block regulatory target index keyword of the second block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the second block regulatory target. Text semantics, according to the second regulatory element text semantics of the second block regulatory target, through the debugged second long short-term memory algorithm, obtain the third regulatory element text semantics of the second block regulatory target; for each third block regulatory target corresponding to each of the first training data, according to the original regulatory element text semantics of the third block regulatory target, through the debugged second long short-term memory algorithm, obtain the fourth regulatory element text semantics of the third block regulatory target; according to the difference between the first regulatory element text semantics corresponding to each of the second block regulatory targets and the debugging learning basis, determine the first debugging error information; according to the commonality weight between the third regulatory element text semantics of the second block regulatory target corresponding to each of the first training data and the fourth regulatory element text semantics of the third block regulatory target, determine the second debugging error information; according to the first debugging error information and the second debugging error information, determine the global debugging error; according to the global debugging error, improve the algorithm variables of the first long short-term memory algorithm and the debugged second long short-term memory algorithm.

[0158] Furthermore, the plurality of first training data include a plurality of training data sets, each of the training data sets including an original regulatory element text semantics of a third block regulatory target, an original regulatory element text semantics of a second block regulatory target having a regulatory contract trigger tag relative to the third block regulatory target, and an original regulatory element text semantics of a second block regulatory target having no regulatory contract trigger tag relative to the third block regulatory target; the determining of the second debugging error information based on the commonality weight between the third regulatory element text semantics of the second block regulatory target and the fourth regulatory element text semantics of the third block regulatory target corresponding to each of the first training data includes: determining the third regulatory element text semantics of the second block regulatory target in each active training data; The first commonality weight between the present semantics and the fourth regulatory element text semantics of the third block regulatory target, and the second commonality weight between the third regulatory element text semantics of the second block regulatory target and the fourth regulatory element text semantics of the third block regulatory target in each negative training data; wherein the positive training data is the third block regulatory target in the training data set and the second block regulatory target with a regulatory contract trigger tag relative to the third block regulatory target, and the negative training data is the third block regulatory target in the training data set and the second block regulatory target with no regulatory contract trigger tag relative to the third block regulatory target; the second debugging error information is determined according to the difference between the first commonality weight and the second commonality weight corresponding to each of the training data sets.

[0159] In the complex scenario of intelligent manufacturing quality supervision, each block supervision target, such as a manufacturing batch or component, contains rich supervision information. In order to fully and deeply grasp the quality status of these supervision targets, the big data supervision system needs to extract and parse the key supervision element text semantics in these targets. This process requires not only that the big data supervision system can accurately grasp the attribute characteristics of the supervision target itself, but also that it can understand and analyze the behavior data of the supervision target in the intelligent manufacturing process. To this end, an innovative technical solution is proposed to obtain the target supervision element text semantics by integrating the index keyword-derived supervision element text semantics of the block supervision target and the original supervision element text semantics of the behavior data.

[0160] First of all, it should be made clear that each block supervision target has its own unique index keywords, which are important bases for the big data supervision system to identify and classify the target during the supervision process. At the same time, these index keywords also contain the text semantics of the derivative supervision elements related to the target, such as the production date of the manufacturing batch, the specifications and models of the parts, etc. In addition, the behavior data of the block supervision target in the intelligent manufacturing process is also the information that the big data supervision system needs to focus on. These data record the actual performance of the target in various links such as production, testing, and transportation, and are an important basis for evaluating the quality status of the target.

[0161] In order to comprehensively extract and parse the text semantics of the regulatory elements of the block regulatory targets, the big data regulatory system first integrates the text semantics of the regulatory elements derived from the index keywords of the block regulatory targets and the original text semantics of the regulatory elements of the behavioral data. The purpose of this step is to organically integrate the text semantics of the regulatory elements scattered from different sources to form a more comprehensive and in-depth text semantics of the derived regulatory elements. In this process, the big data regulatory system needs to fully consider the inherent connection and mutual influence between the text semantics of the regulatory elements derived from the index keywords and the text semantics of the original regulatory elements of the behavioral data to ensure the accuracy and effectiveness of the integration results.

[0162] Next, the big data supervision system will use the deep loop model to further extract and process the text semantics of the integrated derivative supervision elements. The deep loop model is an advanced machine learning algorithm that can master the inherent laws and characteristics of data by learning a large amount of training data. In this application scenario, the big data supervision system chooses the debugged second long short-term memory algorithm (LSTM) as the specific implementation of the deep loop model. The LSTM algorithm is a special type of recurrent neural network (RNN) that can extract and parse complex text semantic information by capturing long-term dependencies in sequence data.

[0163] In order to obtain more accurate and effective text semantics of target regulatory elements, the big data regulatory system also needs to further debug and optimize the LSTM algorithm. Specifically, the big data regulatory system will use a large amount of first training data to repeatedly train and debug the LSTM algorithm until it meets specific debugging requirements. These first training data contain the original regulatory element text semantics of multiple block regulatory targets and the regulatory contract trigger label information between them. By entering these training data into the LSTM algorithm, the big data regulatory system can allow the algorithm to learn how to accurately extract and parse the regulatory element text semantics of block regulatory targets.

[0164] During the debugging process, the big data supervision system also needs to pay attention to two important debugging error information: the first debugging error information and the second debugging error information. The first debugging error information is determined by comparing the difference between the text semantics of the first supervision element extracted by the LSTM algorithm and the debugging learning basis, which reflects the accuracy and reliability of the algorithm in extracting the text semantics of the supervision element. The second debugging error information is determined by comparing the commonality weights between the text semantics of the third supervision element and the text semantics of the fourth supervision element of different block supervision targets, which reflects the stability and consistency of the algorithm when processing different supervision targets.

[0165] In order to determine the second debugging error information, the big data supervision system needs to further analyze the positive training data and negative training data in the training data set. Positive training data refers to the data set of those third block supervision targets and the second block supervision targets that have a supervision contract trigger tag relative to the third block supervision target, while negative training data refers to the data set of those third block supervision targets and the second block supervision targets that do not have a supervision contract trigger tag relative to the third block supervision target. By comparing the commonality weights between the third supervision element text semantics of the second block supervision target and the fourth supervision element text semantics of the third block supervision target in these two types of training data, the big data supervision system can determine the second debugging error information.

[0166] Finally, the big data supervision system will improve the algorithm variables of the LSTM algorithm based on the global debugging error to improve the accuracy and stability of the algorithm. The global debugging error is a comprehensive reflection of the first debugging error information and the second debugging error information, which reflects the overall performance of the algorithm when processing all training data. By continuously adjusting and optimizing the algorithm variables of the LSTM algorithm, the big data supervision system can make the algorithm more adaptable to the actual needs of intelligent manufacturing quality supervision and provide more accurate and effective supervision support for the big data supervision system.

[0167] This design derives the text semantics of regulatory elements and the original text semantics of regulatory elements of behavioral data by integrating the index keywords of the block regulatory target, and further extracts and processes them using the debugged LSTM algorithm to obtain comprehensive and in-depth text semantics of target regulatory elements. This not only improves the accuracy and efficiency of intelligent manufacturing quality supervision, but also provides more dimensional regulatory analysis basis for the big data supervision system.

[0168] In other independent embodiments, the method also includes: obtaining the target regulatory element text semantics of the target regulatory participating entity; determining whether to use the first block regulatory target as the initial screening risk regulatory contract triggering entity of the target regulatory participating entity based on the semantic commonality reinforcement coefficient between the target regulatory element text semantics of the first block regulatory target and the target regulatory element text semantics of the target regulatory participating entity, wherein the target regulatory element text semantics of the target regulatory participating entity is obtained by mining the deep loop model based on the original regulatory element text semantics of the target regulatory participating entity.

[0169] In the current digital age, cross-border payment blockchain supervision and smart manufacturing quality supervision are two crucial application scenarios. They are not only related to the smooth progress of economic activities, but also directly affect product quality and consumer rights. In order to supervise more effectively, the embodiment of this application proposes an innovative technical solution, which combines the text semantics of the target supervision elements of the target supervision participating entity and the text semantics of the target supervision elements of the first block supervision target, and mines and analyzes them through a deep cycle model.

[0170] First, in the field of cross-border payments, the introduction of blockchain technology has brought unprecedented transparency and security to the payment process. However, this has also brought new challenges to regulators. How to effectively supervise cross-border payment activities on the blockchain and ensure the legality and compliance of capital flows is an important task facing regulators.

[0171] In order to solve this problem, a technical solution based on the text semantics of target regulatory participating entities and target regulatory elements is proposed. Specifically, the big data regulatory system first obtains the original regulatory element text semantics of the target regulatory participating entities, which contains important contents such as the basic information, transaction history, and risk rating of the entity. Then, the big data regulatory system uses a deep loop model to mine and analyze these original text semantics and extract the target regulatory element text semantics of the target regulatory participating entities. These target regulatory element text semantics are more refined and representative, and can more accurately reflect the characteristics and risk status of the entity.

[0172] Next, the big data supervision system compares and analyzes the text semantics of the target supervision elements of the target supervision participating entity with the text semantics of the target supervision elements of the first block supervision target. The first block supervision target may be a specific cross-border payment transaction or all transactions related to a specific entity. The big data supervision system evaluates the degree of association and risk level between them by calculating the semantic commonality reinforcement coefficient between the two. If the semantic commonality reinforcement coefficient is high, it means that there is a strong correlation and risk commonality between the target supervision participating entity and the first block supervision target. In this case, the big data supervision system will use the first block supervision target as the initial screening risk supervision contract triggering entity of the target supervision participating entity.

[0173] Such a design can not only improve the accuracy and efficiency of supervision, but also timely detect and warn potential risks and violations. Through the mining and analysis of the deep cycle model, the regulatory authorities can have a deeper understanding of the characteristics and risk status of the target regulatory entities, so as to formulate more accurate and effective regulatory strategies.

[0174] Next, in the field of intelligent manufacturing, product quality is a key factor affecting the survival and development of enterprises. However, due to the complexity and diversity of the intelligent manufacturing process, product quality supervision faces many challenges. How to ensure that the quality of intelligent manufacturing products meets standards and requirements is a common concern of regulators and enterprises.

[0175] Similarly, the big data supervision system also applies the technical solutions based on the text semantics of the target supervision participant entity and the target supervision element to the quality supervision of intelligent manufacturing. The big data supervision system first obtains the original text semantics of the target supervision participant entity (which may be an intelligent manufacturing enterprise, supplier, testing agency, etc.), which contains the basic information of the entity, production process, quality inspection report, etc. Then, the big data supervision system uses the deep loop model to mine and analyze these original text semantics and extract the target supervision element text semantics of the target supervision participant entity.

[0176] In the quality supervision of intelligent manufacturing, the first block supervision target may be a batch of products, a production link or a specific quality problem. The big data supervision system compares and analyzes the text semantics of the target supervision element of the target supervision participant entity with the text semantics of the target supervision element of the first block supervision target. By calculating the semantic commonality reinforcement coefficient between the two, the big data supervision system can evaluate the degree of association and risk level between the target supervision participant entity and the first block supervision target. If the semantic commonality reinforcement coefficient is high, it means that there is a strong correlation and risk commonality between the target supervision participant entity and the first block supervision target, then the regulatory agency needs to conduct a more in-depth investigation and supervision of the entity.

[0177] This brings new breakthroughs and innovations to smart manufacturing quality supervision. It can help regulators to understand the characteristics and risk status of the target regulatory entities more comprehensively and deeply, so as to timely discover and solve potential quality problems. At the same time, it can also improve the self-regulatory ability and product quality awareness of enterprises, and promote enterprises to continuously improve their quality management system and production process control.

[0178] Embodiment 1: obtaining the original regulatory element text semantics of the first block regulatory target, wherein the original regulatory element text semantics include the original regulatory element text semantics of the block regulatory target index keyword and the original regulatory element text semantics of the block regulatory target behavior data, wherein the first block regulatory target is a preliminarily screened regulated blockchain data entity, and the first block regulatory target is in the blockchain data multi-dimensional space; obtaining the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target through a regulatory decision processing algorithm according to the original regulatory element text semantics of the first block regulatory target; obtaining the target regulatory element text semantics of the first block regulatory target according to the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target, so as to perform full life cycle supervision of the first block regulatory target in the blockchain data multi-dimensional space according to the target regulatory element text semantics; wherein the regulatory decision processing algorithm is to debug a first long short-term memory algorithm according to a plurality of first training data The first training data includes the original regulatory element text semantics of the second block regulatory target and the debugging learning basis, the debugging learning basis is the prior regulatory element text semantics of the second block regulatory target obtained by the debugged second long short-term memory algorithm based on the original regulatory element text semantics of the second block regulatory target, the second block regulatory target is the regulated blockchain data entity; the debugged second long short-term memory algorithm is debugged by the second long short-term memory algorithm based on multiple second training data configured with prior knowledge, the second training data includes the original regulatory element text semantics of the regulatory participating entity and the regulated blockchain data entity, the prior knowledge represents whether the regulatory participating entity in the second training data triggers the risk supervision contract on the regulated blockchain data entity, the input information of the second long short-term memory algorithm includes the original regulatory element text semantics of each block regulatory target in the second training data, and the output is the derived regulatory element text semantics of each block regulatory target, and the debugging error information is determined according to the semantic commonality reinforcement coefficient between the derived regulatory element text semantics of the two block regulatory targets in each second training data.

[0179] Embodiment 2: The block supervision target behavior data includes at least one block supervision target behavior vector; the method also includes: taking the first block supervision target as a latent space feature unit, obtaining the latent space feature relationship network corresponding to the first block supervision target, wherein the latent space feature relationship network includes latent space feature units of at least two states, and there are relationship transmission pointers between latent space feature units involved in the business, and a latent space feature unit in the latent space feature relationship network represents a regulated blockchain data entity, a regulatory participating entity, a block supervision target behavior vector of a regulated blockchain data entity, or a block supervision target behavior vector of a regulatory participating entity; taking the latent space feature unit corresponding to the first block supervision target as a key latent space feature unit, and determining at least one upstream and downstream latent space feature of the key latent space feature unit according to the latent space feature relationship network. The regulatory element text semantics of each upstream and downstream latent space feature unit in the at least one upstream and downstream latent space feature unit are obtained; the target regulatory element text semantics of the first block regulatory target are obtained based on the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target, including: obtaining the derived regulatory element text semantics of the first block regulatory target by integrating the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target; obtaining the target regulatory element text semantics of the first block regulatory target based on the derived regulatory element text semantics of the first block regulatory target and the regulatory element text semantics of each upstream and downstream latent space feature unit.

[0180] Embodiment 3, the method of determining at least one upstream and downstream latent space feature unit of the key latent space feature unit based on the latent space feature relationship network includes: taking the key latent space feature unit as the latent space feature activation unit, and according to each type of risk control feature transmission strategy in at least one type of risk control feature transmission strategy set, determining in the latent space feature relationship network a local feature relationship network of the key latent space feature unit corresponding to each type of the risk control feature transmission strategy, wherein the at least one upstream and downstream latent space feature unit includes all latent space feature units other than the key latent space feature unit in the local feature relationship network corresponding to each type of the risk control feature transmission strategy; wherein each type of the risk control feature transmission strategy corresponds to at least two latent space feature units, and the state of each latent space feature unit in the at least two latent space feature units and the involvement situation between the latent space feature units are pre-configured.

[0181] Example 4: The target regulatory element text semantics of the first block regulatory target are obtained based on the derived regulatory element text semantics of the first block regulatory target and the regulatory element text semantics of each upstream and downstream latent space feature unit, including: for each type of risk control feature transmission strategy, the integrated regulatory element text semantics of the first block regulatory target corresponding to the type of risk control feature transmission strategy is obtained by integrating the regulatory element text semantics of each upstream and downstream latent space feature unit in the local feature relationship network corresponding to the type of risk control feature transmission strategy, and the derived regulatory element text semantics of the first block regulatory target; the target regulatory element text semantics of the first block regulatory target is obtained by integrating the integrated regulatory element text semantics of the first block regulatory target corresponding to each type of risk control feature transmission strategy.

[0182] Embodiment 5: The local feature relationship network corresponding to each type of the risk control feature transmission strategy includes x layers of upstream and downstream latent space feature units of the key latent space feature unit, wherein x is an integer greater than or equal to 1, and the first layer of upstream and downstream latent space feature units are upstream and downstream latent space feature units that have real-time relationship transmission pointers with the key latent space feature unit; for each type of the risk control feature transmission strategy, the integrated regulatory element text semantics of the first block supervision target corresponding to the risk control feature transmission strategy of this type is obtained based on the following steps: the regulatory element text semantics of the x-th layer of upstream and downstream latent space feature units in the local feature relationship network corresponding to this type of risk control feature transmission strategy is used as the integrated regulatory element text semantics of the x-th layer of upstream and downstream latent space feature units, and the key latent space feature unit is used as the 0th layer of upstream and downstream latent space feature units, and according to the layer decreasing rule, it is cyclically obtained. Implement the following steps until the integrated regulatory element text semantics of the key latent space feature unit corresponding to the risk control feature transmission strategy of this type is obtained: for each upstream and downstream latent space feature unit u in the y-1th layer, determine each upstream and downstream latent space feature unit v in the upstream and downstream latent space feature units of the yth layer that has a relationship transmission pointer with the upstream and downstream latent space feature unit u, and obtain the integrated regulatory element text semantics of the upstream and downstream latent space feature unit u by integrating the regulatory element text semantics of the upstream and downstream latent space feature unit u and the regulatory element text semantics of each upstream and downstream latent space feature unit v; wherein y is an integer greater than or equal to 1 and less than or equal to x, and when the upstream and downstream latent space feature unit u is a key latent space feature unit, the regulatory element text semantics of the upstream and downstream latent space feature unit u is the derived regulatory element text semantics of the key latent space feature unit.

[0183] Example 6, the integrated regulatory element text semantics of the upstream and downstream latent space feature unit u is obtained by integrating the regulatory element text semantics of the upstream and downstream latent space feature unit u and the regulatory element text semantics of each upstream and downstream latent space feature unit v, including: determining the reinforcement coefficient corresponding to each upstream and downstream latent space feature unit v based on the semantic correlation between the regulatory element text semantics of the upstream and downstream latent space feature unit u and the regulatory element text semantics of each upstream and downstream latent space feature unit v; integrating the regulatory element text semantics of each upstream and downstream latent space feature unit v according to the reinforcement coefficient corresponding to each upstream and downstream latent space feature unit v, integrating the integrated result with the regulatory element text semantics of the upstream and downstream latent space feature unit u, and obtaining the integrated regulatory element text semantics of the upstream and downstream latent space feature unit u.

[0184] Embodiment 7, the method of obtaining the target regulatory element text semantics of the first block regulatory target by integrating the integrated regulatory element text semantics corresponding to each type of the risk control feature transmission strategy of the first block regulatory target, including: determining the reinforcement coefficient corresponding to each type of the risk control feature transmission strategy according to the semantic correlation between the integrated regulatory element text semantics of the various types of the risk control feature transmission strategy; integrating the integrated regulatory element text semantics of each type of the risk control feature transmission strategy according to the reinforcement coefficient corresponding to each type of the risk control feature transmission strategy to obtain the target regulatory element text semantics of the first block regulatory target; the derivative regulatory element text semantics based on the block regulatory target index keyword of the first block regulatory target, and the block regulatory element of the first block regulatory target; The original regulatory element text semantics of the target behavior data are used to obtain the target regulatory element text semantics of the first block regulatory target, including: obtaining the derived regulatory element text semantics of the first block regulatory target by integrating the derived regulatory element text semantics of the block regulatory target index keywords of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target; according to the derived regulatory element text semantics of the first block regulatory target, the target regulatory element text semantics of the first block regulatory target are extracted through a deep loop model; wherein the deep loop model is the second long short-term memory algorithm after debugging, or a long short-term memory algorithm obtained after secondary debugging of the second long short-term memory algorithm after debugging.

[0185] The following is an exemplary introduction to the above 7 embodiments.

[0186] 1. Embodiment 1

[0187] 1) Obtain the original regulatory element text semantics

[0188] In the big data supervision system, for the initially screened regulated blockchain data entity (i.e., the first block supervision target), its original supervision element text semantics must be obtained. This first block supervision target is in the multidimensional space of blockchain data. The original supervision element text semantics consists of two parts: the original supervision element text semantics of the block supervision target index keyword and the original supervision element text semantics of the block supervision target behavior data.

[0189] For example, if the first block supervision target is a production batch, the original supervision element text semantics of the block supervision target index keyword may include "production batch number: 12345", "production date: 2023-01-01", "production line number: 001", etc. The original supervision element text semantics of the block supervision target behavior data may contain various data of the production batch during the production process, such as raw material quality (for example, the raw material quality index value is: purity 95%), production parameters (such as temperature setting 200°C, pressure setting 10MPa), equipment status (equipment operation time is 500 hours, equipment wear rate is 5%), inspection records (initial inspection pass rate is 90%), etc. These original supervision element text semantics provide a comprehensive information basis for subsequent supervision.

[0190] 2) Obtain the text semantics of derived regulatory elements

[0191] According to the original regulatory element text semantics of the first block regulatory target, the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target is obtained through the regulatory decision processing algorithm. The regulatory decision processing algorithm here is obtained by debugging the first long short-term memory algorithm (LSTM) based on multiple first training data.

[0192] An example formula for the Long Short-Term Memory (LSTM) algorithm is as follows:

[0193] Forget gate: (f_t=sigma(W_fcdot[h_{t-1},x_t]+b_f)), where (W_f) is the weight matrix of the forget gate, (h_{t-1}) is the hidden state at the previous moment, (x_t) is the input at the current moment, (b_f) is the bias term, and (sigma) is the sigmoid function.

[0194] Input gate: (i_t=sigma(W_icdot[h_{t-1},x_t]+b_i)), (tilde{C}_t=tanh(W_Ccdot[h_{t-1},x_t]+b_C)), where (W_i), (W_C) are weight matrices, and (b_i), (b_C) are bias terms.

[0195] Cell state update: (C_t=f_t*C_{t-1}+i_t*tilde{C}_t).

[0196] Output gate: (o_t=sigma(W_ocdot[h_{t-1},x_t]+b_o)), (h_t=o_t*tanh(C_t)).

[0197] In this embodiment, taking the production batch as an example, the first training data includes the original regulatory element text semantics and debugging learning basis of the second block regulatory target (also a regulated blockchain data entity, such as another production batch). For the second block regulatory target, its original regulatory element text semantics also includes index keywords (such as production batch number, production date, production line number, etc.) and production process data. Through the debugged second long short-term memory algorithm, the prior regulatory element text semantics of this second block regulatory target can be obtained, such as "qualified quality", "there are raw material problems", etc. For example, the input of the first long short-term memory algorithm is the original regulatory element text semantics of the first block regulatory target (such as the various index keywords and production data of the production batch mentioned above), and after the algorithm processing (similar to the neuron activation and information transmission process in the neural network), the output is the derived regulatory element text semantics. For example, it is possible to obtain "production batch quality stability analysis: relatively stable but with a fluctuating trend", "production parameter abnormality detection: temperature parameters occasionally exceed the normal range of 1-2°C", "raw material quality risk assessment: purity close to the lower limit is risky" and other derived regulatory element text semantics.

[0198] 3) Obtain the text semantics of the target regulatory element and conduct full life cycle supervision

[0199] The target regulatory element text semantics of the first block regulatory target are obtained based on the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target. Taking the production batch as an example, the target regulatory element text semantics may include "the quality of the production batch is qualified but the purity of the raw materials needs to be paid attention to", "there are problems with the quality of the raw materials but they can be improved through subsequent processes", "the monitoring of production parameters (especially temperature parameters) needs to be further strengthened", etc.

[0200] Based on the text semantics of this target regulatory element, the big data regulatory system conducts full life cycle supervision of the first block regulatory target in the multidimensional space of blockchain data. This means that the entire process from the data generation of the production batch on the blockchain (such as recording the initial data when the production batch is created), to storage (data is stored in a specific node of the blockchain), processing (such as updating and analyzing data in different production links), analysis (such as analyzing data according to quality monitoring algorithms) to destruction (if the data no longer needs to be deleted according to the prescribed process) is monitored and managed. In this process, for example, when the temperature parameters exceed the normal range many times during the production process, the system will analyze and discover potential regulatory risks in real time based on the information in the text semantics of the target regulatory element "need to further strengthen the monitoring of production parameters (especially temperature parameters)", and then issue an alarm to notify relevant regulators to handle it.

[0201] 2. Embodiment 2

[0202] 1) Hidden space feature relationship network related operations

[0203] In this embodiment, the block supervision target behavior data includes at least one block supervision target behavior vector. First, take the first block supervision target as a latent space feature unit, and obtain the latent space feature relationship network corresponding to the first block supervision target. This latent space feature relationship network includes latent space feature units of at least two states, and there are relationship conduction pointers between latent space feature units involved in the business. For example, in the scenario of a smart manufacturing factory, the latent space feature unit in the latent space feature relationship network may represent a regulated blockchain data entity (such as a production equipment), a regulatory participating entity (such as a quality inspector), a block supervision target behavior vector of a regulated blockchain data entity (such as the energy consumption data vector of the production equipment), or a block supervision target behavior vector of a regulatory participating entity (such as the detection frequency vector of the quality inspector).

[0204] The latent space feature unit corresponding to the first block supervision target is taken as the key latent space feature unit, and at least one upstream and downstream latent space feature unit of the key latent space feature unit is determined based on the latent space feature relationship network. For example, if the key latent space feature unit is a certain production equipment, then its upstream latent space feature unit may be the raw material supplier (because the raw material supply affects the production of the equipment), and the downstream latent space feature unit may be the product packaging department (because the products produced by the equipment will enter the packaging link).

[0205] Obtain the text semantics of the regulatory elements of each upstream and downstream latent space feature unit in at least one upstream and downstream latent space feature unit. For example, the text semantics of the regulatory elements of the upstream raw material supplier include "raw material supply stability: high, raw material quality consistency: good", and the text semantics of the regulatory elements of the downstream product packaging department include "packaging efficiency: 80%, packaging quality pass rate: 95%".

[0206] 2) Improved method for obtaining text semantics of target regulatory elements

[0207] The derived regulatory element text semantics of the first block regulatory target are obtained by integrating the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target. For example, for the first block regulatory target of production equipment, the derived regulatory element text semantics such as "equipment operation efficiency analysis: 85%" and the original behavior data of the equipment (such as equipment energy consumption value, etc.) are integrated to obtain a more comprehensive derived regulatory element text semantics, such as "equipment operation efficiency analysis: 85%, energy consumption is within the normal range but close to the upper limit".

[0208] Here, we can use the weighted summation formula for integration. For example, the derived regulatory factor text semantics is (A), the original behavior data is (B), and the integrated derived regulatory factor text semantics is (C=alphaA+(1-alpha)B), where (alpha) is the weight coefficient ((0 <alpha<1))。

[0209] According to the derived regulatory element text semantics of the regulatory target of the first block and the regulatory element text semantics of each upstream and downstream latent space feature unit, the target regulatory element text semantics of the regulatory target of the first block is obtained. For example, based on the derived regulatory element text semantics of production equipment and the regulatory element text semantics of raw material suppliers and product packaging departments, the target regulatory element text semantics may be "the operating efficiency of production equipment can be improved, which is not greatly affected by the stability of raw material supply, but may affect the downstream packaging efficiency, and energy consumption needs to be paid attention to."

[0210] 3. Embodiment 3

[0211] 1) Specific method for determining the upstream and downstream latent space feature units

[0212] Taking the key latent space feature unit as the latent space feature activation unit, according to each type of risk control feature transmission strategy in at least one set risk control feature transmission strategy, a local feature relationship network of the key latent space feature unit corresponding to each type of risk control feature transmission strategy is determined in the latent space feature relationship network. For example, two types of risk control feature transmission strategies are set, one is a transmission strategy based on the production process sequence, and the other is a transmission strategy based on cost association.

[0213] For the risk control feature transmission strategy based on the production process sequence, if the key latent space feature unit is a certain device in the production process, then according to the production process sequence, its upstream latent space feature unit may be the device of the previous process, and the downstream latent space feature unit may be the device of the next process. These upstream and downstream latent space feature units and the key latent space feature unit constitute a local feature relationship network based on the production process sequence.

[0214] For the risk control feature transmission strategy based on cost association, if the key latent space feature unit is a high-cost production equipment, then its upstream latent space feature unit may be a supplier of high-priced parts, and the downstream latent space feature unit may be a subsequent processing department that has a greater impact on product costs. These upstream and downstream latent space feature units and the key latent space feature units constitute a local feature relationship network based on cost association. Among them, each type of risk control feature transmission strategy corresponds to at least two latent space feature units, and the states of these latent space feature units and the involvement situations between the latent space feature units are pre-configured. For example, in the risk control feature transmission strategy based on the production process sequence, the operating status of each production equipment (normal, faulty, etc.) and the sequence relationship between them are pre-configured (the output of the previous process equipment is the input of the next process equipment).

[0215] When determining the upstream and downstream latent space feature units, an algorithm based on graph theory can be used. For example, the latent space feature relationship network is regarded as a directed graph (G=(V, E)), where (V) is a vertex set (representing latent space feature units) and (E) is an edge set (representing relationship conduction pointers). For the risk control feature conduction strategy based on the production process sequence, starting from the key latent space feature unit (v_0), search for adjacent vertices along the direction of the edge (representing the production process sequence), which are the upstream and downstream latent space feature units. The depth-first search (DFS) algorithm or the breadth-first search (BFS) algorithm can be used. The recursive formula of the depth-first search is: (DFS (v)), mark (v) as visited; for each unvisited adjacent vertex (u) of (v), (DFS (u)). The breadth-first search uses a queue data structure, first enqueues the starting vertex (v_0), then loops to take out the head vertex (v), marks (v) as visited, and enqueues the unvisited adjacent vertices of (v) until the queue is empty.

[0216] 4. Embodiment 4

[0217] 1) The process of obtaining the text semantics of integrated regulatory elements

[0218] For each type of risk control feature transmission strategy, the integrated regulatory element text semantics of the first block regulatory target corresponding to this type of risk control feature transmission strategy is obtained by integrating the regulatory element text semantics of each upstream and downstream latent space feature unit in the local feature relationship network corresponding to this type of risk control feature transmission strategy and the derived regulatory element text semantics of the first block regulatory target.

[0219] For example, for the risk control feature transmission strategy based on the production process sequence, for example, the text semantics of the regulatory element of the upstream equipment is "equipment operation status: normal, production efficiency: 90%", the text semantics of the regulatory element of the downstream equipment is "equipment operation status: normal, production efficiency: 85%", and the derived regulatory element text semantics of the first block regulatory target (such as intermediate production equipment) is "equipment operation efficiency: 88%, with a risk of minor failures". The weighted average formula can be used for integration. Suppose the text semantics of the regulatory element of the upstream equipment is (A), the text semantics of the regulatory element of the downstream equipment is (B), and the derived regulatory element text semantics of the first block regulatory target is (C). The integrated regulatory element text semantics are:

[0220] (D=omega_1A+omega_2B+omega_3C);

[0221] Among them, (omega_1+omega_2+omega_3=1); (omega_1), (omega_2), (omega_3) are weight coefficients determined based on factors such as the importance of the equipment to the overall production.

[0222] Then, by integrating the integrated regulatory element text semantics corresponding to various risk control feature transmission strategies of the first block regulatory target, the target regulatory element text semantics of the first block regulatory target is obtained. For example, the integrated regulatory element text semantics of the risk control feature transmission strategy based on the production process sequence and the integrated regulatory element text semantics of the risk control feature transmission strategy based on cost association are integrated (algorithms such as weighted average can also be used) to obtain the final target regulatory element text semantics, such as "the overall status of production equipment: efficiency is acceptable but there is a risk of minor failures, and the cost is less affected by the price fluctuations of parts suppliers, but it may affect the cost-effectiveness of downstream processing departments."

[0223] V. Embodiment 5

[0224] 1) Semantic acquisition of integrated regulatory elements based on the number of layers

[0225] The local feature relationship network corresponding to each type of risk control feature transmission strategy includes (x) layers of upstream and downstream latent space feature units of the key latent space feature unit, where (x) is an integer greater than or equal to 1, and the first layer of upstream and downstream latent space feature units are upstream and downstream latent space feature units that have real-time relationship transmission pointers with the key latent space feature units.

[0226] For each type of risk control feature transmission strategy, the first block supervision target corresponds to the integrated supervision element text semantics of this type of risk control feature transmission strategy based on the following steps. For example (x=2), take the risk control feature transmission strategy based on the production process sequence as an example.

[0227] The regulatory element text semantics of the second-layer upstream and downstream latent space feature units in the local feature relationship network corresponding to this type of risk control feature transmission strategy are used as the integrated regulatory element text semantics of the second-layer upstream and downstream latent space feature units. For example, the regulatory element text semantics of the second-layer upstream equipment is "equipment maintenance cycle: once a month, equipment aging degree: 10%", and the regulatory element text semantics of the second-layer downstream equipment is "equipment idle rate: 15%, equipment load capacity: 80%".

[0228] Take the key latent space feature unit as the upstream and downstream latent space feature unit of the 0th layer, and implement the following steps repeatedly according to the layer-decreasing rule until the key latent space feature unit is obtained corresponding to the integrated regulatory element text semantics of this type of risk control feature transmission strategy.

[0229] For each upstream and downstream latent space feature unit (u) in the (y-1)th layer (starting from (y=1)), determine the upstream and downstream latent space feature units (v) in the upstream and downstream latent space feature units of the (y)th layer that have relationship conduction pointers with the upstream and downstream latent space feature unit (u), and obtain the integrated regulatory element text semantics of the upstream and downstream latent space feature unit (u) by integrating the regulatory element text semantics of the upstream and downstream latent space feature unit (u) and the regulatory element text semantics of each upstream and downstream latent space feature unit (v).

[0230] For example, the integration is performed by weighted summation. Let the textual semantics of the regulatory elements of the upstream and downstream latent space feature units (u) be (A), and the textual semantics of the regulatory elements of each upstream and downstream latent space feature unit (v) be (B_i) ((i=1, 2, cdots, n)), and the corresponding weight be (omega_i). Then the integrated textual semantics of the regulatory elements is (C=A+sum_{i=1}^{n}omega_iB_i). For example, when (y=1), the derived textual semantics of the regulatory elements of the 0th layer (key latent space feature unit, such as intermediate production equipment) is "equipment operation efficiency: 88%, with a risk of minor failure", and the textual semantics of the regulatory elements of the 1st layer upstream equipment is "equipment operation status: normal, production efficiency: 90%". The integrated textual semantics of the regulatory elements of the 1st layer upstream equipment is obtained by integration (such as weighted summation, weights are assigned according to the degree of influence of the equipment on the intermediate equipment), and then similar integration operations are performed with the integrated textual semantics of the regulatory elements of the 2nd layer upstream equipment, and the integrated textual semantics of the regulatory elements of the key latent space feature units corresponding to this type of risk control feature transmission strategy are gradually obtained.

[0231] VI. Example 6

[0232] In this embodiment, the focus is on integrating the textual semantics of the regulatory elements of the upstream and downstream latent space feature units through a specific algorithm, thereby obtaining the integrated textual semantics of the regulatory elements of the upstream and downstream latent space feature units (u). This process mainly determines the reinforcement coefficient based on the semantic association between the upstream and downstream latent space feature units (u) and each upstream and downstream latent space feature unit (v), thereby realizing the integration of the textual semantics of the regulatory elements.

[0233] 1. Determining semantic relevance and reinforcement coefficient

[0234] 1) Calculate semantic relevance

[0235] Calculate the semantic association between the text semantics of the regulatory elements of the upstream and downstream latent space feature units (u) and the text semantics of the regulatory elements of each upstream and downstream latent space feature unit (v). Here, a word vector model is used, such as the Word2Vec model. For example, the text semantics of the regulatory elements of the upstream and downstream latent space feature units (u) is (A), and the text semantics of the regulatory elements of each upstream and downstream latent space feature unit (v) is:

[0236] (B_i) ((i=1, 2, cdots, n)).

[0237] First, convert the text into word vectors. Let the word vector corresponding to (A) be (vec{a}) and the word vector corresponding to (B_i) be (vec{b}_i), then the semantic relevance

[0238] (s_i

[0239] =cos(vec{a},vec{b}_i)

[0240] =frac{vec{a}cdotvec{b}_i}{vertvec{a}vertvertvec{b}_ivert}).

[0241] For example, if (vec{a}=(1, 2, 3)), (vec{b}_1=(2, 3, 4)), then:

[0242] (vertvec{a}vert=sqrt{1^2+2^2+3^2}=sqrt{14}),

[0243] (vertvec{b}_1vert=sqrt{2^2+3^2+4^2}=sqrt{29}),

[0244] (vec{a}cdotvec{b}_1=1times2+2times3+3times4=20),

[0245] So (s_1=frac{20}{sqrt{14}timessqrt{29}}approx0.89).

[0246] 2) Determine the reinforcement coefficient

[0247] The reinforcement coefficients corresponding to each upstream and downstream latent space feature unit (v) are determined according to the calculated semantic relevance. For example, a mapping function can be set such that when (s_igeq0.8), the reinforcement coefficient (r_i=0.8); when (0.5leqs_i<0.8), (r_i=0.5); when (s_i<0.5), (r_i=0.3).

[0248] 2. Integrating regulatory elements into text semantics

[0249] 1) Semantic integration of regulatory elements of (v)

[0250] According to the corresponding reinforcement coefficients (r_i) of each upstream and downstream latent space feature unit (v), the text semantics of the regulatory elements of each upstream and downstream latent space feature unit (v) are integrated (such as weighted summation). Let the result of the integration of (B_i) be (B_{sum}), then (B_{sum}=sum_{i=1}^{n}r_iB_i). For example, there are three upstream and downstream latent space feature units (v_1), (v_2), and (v_3), and their text semantics of regulatory elements are (B_1=5), (B_2=3), and (B_3=4), respectively, and the corresponding reinforcement coefficients are (r_1=0.8), (r_2=0.5), and (r_3=0.3), then:

[0251] (B_{sum}=0.8times5+0.5times3+0.3times4=4+1.5+1.2=6.7).

[0252] 2) Semantic integration with the regulatory element text of (u)

[0253] The integrated result (B_{sum}) is integrated with the regulatory element text semantics (A) of the upstream and downstream latent space feature units (u) to obtain the integrated regulatory element text semantics of the upstream and downstream latent space feature units (u). Adaptive addition or weighted summation can be used here. For example, if addition is used, the integrated result is (C=A+B_{sum}). For example, if (A=3), then (C=3+6.7=9.7).

[0254] VII. Embodiment 7

[0255] This embodiment mainly involves obtaining the target regulatory element text semantics of the first block regulatory target by integrating the integrated regulatory element text semantics corresponding to various risk control feature transmission strategies of the first block regulatory target, and also introduces another method for obtaining the target regulatory element text semantics based on a deep loop model.

[0256] 1. Obtaining the text semantics of target regulatory elements based on the integration of various risk control feature transmission strategies

[0257] 1) Determine the reinforcement coefficient

[0258] According to the semantic relevance between the text semantics of the integrated regulatory elements of various risk control feature transmission strategies, the reinforcement coefficients corresponding to each risk control feature transmission strategy are determined. For example, the semantic relevance is calculated using cosine similarity. Suppose the text semantics of the integrated regulatory elements of the risk control feature transmission strategy based on the production process sequence is (P), and the text semantics of the integrated regulatory elements of the risk control feature transmission strategy based on cost association is (Q). Convert (P) and (Q) into word vectors, set as (vec{p}) and (vec{q}), then the semantic relevance is:

[0259] (s

[0260] =cos(vec{p},vec{q})

[0261] =frac{vec{p}cdotvec{q}}{vertvec{p}vertvertvec{q}vert}).

[0262] The reinforcement coefficient is determined according to the semantic association. For example, when (sgeq0.7), the reinforcement coefficient of the risk control feature transmission strategy based on the production process sequence is (omega_1=0.7), and the reinforcement coefficient of the risk control feature transmission strategy based on cost association is (omega_2=0.3).

[0263] 2) Integrate regulatory elements text semantics

[0264] According to the reinforcement coefficients corresponding to each risk control feature transmission strategy, the integrated regulatory element text semantics of each risk control feature transmission strategy are integrated (such as weighted summation) to obtain the target regulatory element text semantics of the first block regulatory target. Assume that the target regulatory element text semantics is (R), then:

[0265] (R=omega_1P+omega_2Q).

[0266] For example, if (P=10), (Q=8), then (R=0.7times10+0.3times8=7+2.4=9.4).

[0267] 2. Obtaining the text semantics of target regulatory elements based on the deep recurrent model

[0268] 1) Integration to obtain the textual semantics of derived regulatory elements

[0269] According to the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target, the derived regulatory element text semantics of the first block regulatory target is obtained. For example, by using the weighted summation method, assuming that the derived regulatory element text semantics of the block regulatory target index keyword is (D), and the original regulatory element text semantics of the block regulatory target behavior data is (E), then the derived regulatory element text semantics (F=alphaD+(1-alpha)E), where (alpha) is the weight coefficient ((0 <alpha<1))。

[0270] 2) Extraction through deep recurrent models

[0271] Based on the text semantics of the derived regulatory elements of the regulatory objectives of the first block, the text semantics of the target regulatory elements of the regulatory objectives of the first block are extracted through a deep loop model (such as a second long short-term memory algorithm after debugging, or a long short-term memory algorithm obtained after secondary debugging of the debugged second long short-term memory algorithm).

[0272] Taking the long short-term memory algorithm (LSTM) as an example, its exemplary formula is as follows:

[0273] Forget gate: (f_t=sigma(W_fcdot[h_{t-1},x_t]+b_f)), where (W_f) is the weight matrix of the forget gate, (h_{t-1}) is the hidden state at the previous moment, (x_t) is the input at the current moment, (b_f) is the bias term, and (sigma) is the sigmoid function.

[0274] Input gate: (i_t=sigma(W_icdot[h_{t-1},x_t]+b_i)), (tilde{C}_t=tanh(W_Ccdot[h_{t-1},x_t]+b_C)), where (W_i), (W_C) are weight matrices, and (b_i), (b_C) are bias terms.

[0275] Cell state update: (C_t=f_t*C_{t-1}+i_t*tilde{C}_t).

[0276] Output gate: (o_t=sigma(W_ocdot[h_{t-1},x_t]+b_o)), (h_t=o_t*tanh(C_t)).

[0277] The derived regulatory element text semantics (F) is appropriately encoded as input (x_t), and after calculation by the LSTM algorithm, the target regulatory element text semantics is finally obtained.

[0278] Further, Figure 2 A schematic diagram of the structure of a big data supervision system 200 provided in an embodiment of the present application. Figure 2 The big data supervision system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0279] Alternatively, if Figure 2 As shown, the big data supervision system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present application.

[0280] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0281] Alternatively, if Figure 2 As shown, the big data supervision system 200 may also include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices. Specifically, it may send information or data to other devices, or receive information or data sent by other devices.

[0282] Optionally, the big data supervision system 200 can implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or a device deployed with a storage engine in each method of the embodiment of the present application. For the sake of brevity, it will not be repeated here. It should be understood that the processor of the embodiment of the present application may be an integrated circuit chip with signal processing capabilities. It can be understood that the memory in the embodiment of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that the memory of the system and method described herein is intended to include, but is not limited to, a suitable type of memory.

[0283] Based on the above, a readable storage medium is provided, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0284] Based on the same inventive concept, a big data supervision device is also provided, which is used to: obtain the original supervision element text semantics of the first block supervision target, the original supervision element text semantics includes the original supervision element text semantics of the block supervision target index keyword and the original supervision element text semantics of the block supervision target behavior data, the first block supervision target is a preliminarily screened regulated blockchain data entity, and the first block supervision target is in the blockchain data multi-dimensional space; according to the original supervision element text semantics of the first block supervision target, the derived supervision element text semantics of the block supervision target index keyword of the first block supervision target is obtained through a supervision decision processing algorithm; according to the derived supervision element text semantics of the block supervision target index keyword of the first block supervision target, and the original supervision element text semantics of the block supervision target behavior data of the first block supervision target, the target supervision element text semantics of the first block supervision target is obtained, so as to perform full life cycle supervision of the first block supervision target in the blockchain data multi-dimensional space according to the target supervision element text semantics; wherein the supervision decision processing algorithm is based on multiple first training The first training data is obtained by debugging a first long short-term memory algorithm, the first training data includes the original regulatory element text semantics of the second block regulatory target and the debugging learning basis, the debugging learning basis is based on the original regulatory element text semantics of the second block regulatory target, and the prior regulatory element text semantics of the second block regulatory target is obtained by the debugged second long short-term memory algorithm, the second block regulatory target is a regulated blockchain data entity; the debugged second long short-term memory algorithm is obtained by debugging the second long short-term memory algorithm based on multiple second training data configured with prior knowledge, the second training data includes the original regulatory element text semantics of the regulatory participating entity and the regulated blockchain data entity, the prior knowledge represents whether the regulatory participating entity in the second training data triggers a risk supervision contract on the regulated blockchain data entity, the input information of the second long short-term memory algorithm includes the original regulatory element text semantics of each block regulatory target in the second training data, and the output is the derived regulatory element text semantics of each block regulatory target, and the debugging error information is determined based on the semantic commonality reinforcement coefficient between the derived regulatory element text semantics of two block regulatory targets in each second training data.

[0285] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the present application, all of which are within the protection of the present application.

Claims

1. A method for supervising the full life cycle of blockchain data in multidimensional space, characterized in that: The method uses a big data supervision system, and the method includes: Obtaining the original regulatory element text semantics of the first block regulatory target, wherein the original regulatory element text semantics includes the original regulatory element text semantics of the block regulatory target index keyword and the original regulatory element text semantics of the block regulatory target behavior data, wherein the first block regulatory target is a preliminarily screened regulated blockchain data entity, and the first block regulatory target is in the blockchain data multidimensional space; wherein both the block regulatory target behavior data and the regulated blockchain data entity are intelligent manufacturing data; According to the original regulatory element text semantics of the first block regulatory objective, the derived regulatory element text semantics of the block regulatory objective index keyword of the first block regulatory objective are obtained through a regulatory decision processing algorithm; According to the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target, the target regulatory element text semantics of the first block regulatory target are obtained, so as to perform full life cycle supervision of the first block regulatory target in the multi-dimensional space of the blockchain data according to the target regulatory element text semantics; Among them, the regulatory decision processing algorithm is obtained by debugging the first long short-term memory algorithm based on multiple first training data, the first training data includes the original regulatory element text semantics of the second block regulatory target and the debugging learning basis, the debugging learning basis is the prior regulatory element text semantics of the second block regulatory target obtained by the debugged second long short-term memory algorithm based on the original regulatory element text semantics of the second block regulatory target, and the second block regulatory target is the regulated blockchain data entity; the debugged second long short-term memory algorithm is debugged by the second long short-term memory algorithm based on multiple second training data configured with prior knowledge, the second training data includes the original regulatory element text semantics of the regulatory participating entity and the regulated blockchain data entity, the prior knowledge represents whether the regulatory participating entity in the second training data triggers the risk supervision contract on the regulated blockchain data entity, the input information of the second long short-term memory algorithm includes the original regulatory element text semantics of each block regulatory target in the second training data, and the output is the derived regulatory element text semantics of each block regulatory target, and the debugging error information is determined based on the semantic commonality reinforcement coefficient between the derived regulatory element text semantics of the two block regulatory targets in each second training data.

2. The method according to claim 1, characterized in that The block supervision target behavior data includes at least one block supervision target behavior vector; the method further includes: Taking the first block supervision target as a latent space feature unit, obtaining a latent space feature relationship network corresponding to the first block supervision target, wherein the latent space feature relationship network includes latent space feature units of at least two states, and there are relationship transmission pointers between latent space feature units involved in business, and a latent space feature unit in the latent space feature relationship network represents a regulated blockchain data entity, a regulatory participating entity, a block supervision target behavior vector of a regulated blockchain data entity, or a block supervision target behavior vector of a regulatory participating entity; Taking the latent space feature unit corresponding to the first block supervision target as the key latent space feature unit, and determining at least one upstream and downstream latent space feature unit of the key latent space feature unit according to the latent space feature relationship network; Obtaining the supervision element text semantics of each upstream and downstream latent space feature unit in the at least one upstream and downstream latent space feature unit; The step of obtaining the target regulatory element text semantics of the first block regulatory target based on the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target includes: Obtaining the derived regulatory element text semantics of the first block regulatory target by integrating the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target; According to the derived regulatory element text semantics of the first block regulatory target and the regulatory element text semantics of each of the upstream and downstream latent space feature units, the target regulatory element text semantics of the first block regulatory target is obtained.

3. The method according to claim 2, characterized in that The determining, based on the latent space feature relationship network, at least one upstream and downstream latent space feature unit of the key latent space feature unit comprises: Taking the key latent space feature unit as the latent space feature activation unit, according to each type of risk control feature transmission strategy in at least one type of risk control feature transmission strategy, a local feature relationship network of the key latent space feature unit corresponding to each type of risk control feature transmission strategy is determined in the latent space feature relationship network, wherein the at least one upstream and downstream latent space feature unit includes each latent space feature unit except the key latent space feature unit in the local feature relationship network corresponding to each type of risk control feature transmission strategy; Among them, each type of risk control feature transmission strategy corresponds to at least two latent space feature units, and the state of each latent space feature unit in the at least two latent space feature units and the involvement situation between the latent space feature units are pre-configured.

4. The method according to claim 3, characterized in that The step of obtaining the target regulatory element text semantics of the first block regulatory target based on the derived regulatory element text semantics of the first block regulatory target and the regulatory element text semantics of each of the upstream and downstream latent space feature units includes: For each type of risk control feature transmission strategy, the integrated regulatory element text semantics of the first block regulatory target corresponding to the risk control feature transmission strategy is obtained by integrating the regulatory element text semantics of each upstream and downstream latent space feature unit in the local feature relationship network corresponding to the risk control feature transmission strategy and the derived regulatory element text semantics of the first block regulatory target; By integrating the integrated regulatory element text semantics of the first block regulatory target corresponding to each type of risk control feature transmission strategy, the target regulatory element text semantics of the first block regulatory target is obtained.

5. The method according to claim 4, characterized in that The local feature relationship network corresponding to each type of risk control feature transmission strategy includes x layers of upstream and downstream latent space feature units of the key latent space feature unit, wherein x is an integer greater than or equal to 1, and the first layer of upstream and downstream latent space feature units are upstream and downstream latent space feature units that have real-time relationship transmission pointers with the key latent space feature unit; For each type of risk control feature transmission strategy, the integrated regulatory element text semantics of the first block regulatory target corresponding to the risk control feature transmission strategy of this type is obtained based on the following steps: The text semantics of the regulatory elements of the x-th layer of upstream and downstream latent space feature units in the local feature relationship network corresponding to this type of risk control feature transmission strategy are used as the integrated text semantics of the regulatory elements of the x-th layer of upstream and downstream latent space feature units, and the key latent space feature units are used as the 0-th layer of upstream and downstream latent space feature units. According to the layer-decreasing rule, the following steps are implemented repeatedly until the integrated text semantics of the regulatory elements of the key latent space feature units corresponding to this type of risk control feature transmission strategy are obtained: For each upstream and downstream latent space feature unit u of the y-1th layer, determine each upstream and downstream latent space feature unit v in the upstream and downstream latent space feature units of the yth layer that has a relational conduction pointer with the upstream and downstream latent space feature unit u, and obtain the integrated regulatory element text semantics of the upstream and downstream latent space feature unit u by integrating the regulatory element text semantics of the upstream and downstream latent space feature unit u and the regulatory element text semantics of each upstream and downstream latent space feature unit v; Among them, y is an integer greater than or equal to 1 and less than or equal to x. When the upstream and downstream latent space feature unit u is a key latent space feature unit, the regulatory element text semantics of the upstream and downstream latent space feature unit u is the derived regulatory element text semantics of the key latent space feature unit.

6. The method according to claim 5, characterized in that The integrated supervision element text semantics of the upstream and downstream latent space feature unit u is obtained by integrating the supervision element text semantics of the upstream and downstream latent space feature unit u and the supervision element text semantics of each upstream and downstream latent space feature unit v, including: Based on the semantic association between the text semantics of the regulatory elements of the upstream and downstream latent space feature units u and the text semantics of the regulatory elements of each of the upstream and downstream latent space feature units v, determining the reinforcement coefficient corresponding to each of the upstream and downstream latent space feature units v; According to the reinforcement coefficients corresponding to each of the upstream and downstream latent space feature units v, the regulatory element text semantics of each of the upstream and downstream latent space feature units v are integrated, and the integrated results are integrated with the regulatory element text semantics of the upstream and downstream latent space feature units u to obtain the integrated regulatory element text semantics of the upstream and downstream latent space feature units u.

7. The method according to claim 4, characterized in that The target regulatory element text semantics of the first block regulatory target are obtained by integrating the integrated regulatory element text semantics corresponding to each of the risk control feature transmission strategies of the first block regulatory target, including: Determine the reinforcement coefficient corresponding to each of the risk control feature transmission strategies according to the semantic relevance between the semantics of the integrated supervision element texts of the risk control feature transmission strategies; According to the reinforcement coefficients corresponding to each of the risk control feature transmission strategies, the integrated supervision element text semantics of each of the risk control feature transmission strategies are integrated to obtain the target supervision element text semantics of the first block supervision target; The step of obtaining the target regulatory element text semantics of the first block regulatory target based on the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target includes: Obtaining the derived regulatory element text semantics of the first block regulatory target by integrating the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target; According to the derived regulatory element text semantics of the first block regulatory target, the target regulatory element text semantics of the first block regulatory target are extracted through a deep loop model; The deep loop model is the second long short-term memory algorithm after debugging, or a long short-term memory algorithm obtained by debugging the second long short-term memory algorithm after debugging for the second time.

8. A big data supervision system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 7.

9. A big data monitoring device, characterized in that: The big data monitoring device is used for: Obtaining the original regulatory element text semantics of the first block regulatory target, wherein the original regulatory element text semantics includes the original regulatory element text semantics of the block regulatory target index keyword and the original regulatory element text semantics of the block regulatory target behavior data, wherein the first block regulatory target is a preliminarily screened regulated blockchain data entity, and the first block regulatory target is in the blockchain data multidimensional space; wherein both the block regulatory target behavior data and the regulated blockchain data entity are intelligent manufacturing data; According to the original regulatory element text semantics of the first block regulatory objective, the derived regulatory element text semantics of the block regulatory objective index keyword of the first block regulatory objective are obtained through a regulatory decision processing algorithm; According to the derived regulatory element text semantics of the block regulatory target index keyword of the first block regulatory target and the original regulatory element text semantics of the block regulatory target behavior data of the first block regulatory target, the target regulatory element text semantics of the first block regulatory target are obtained, so as to perform full life cycle supervision of the first block regulatory target in the multi-dimensional space of the blockchain data according to the target regulatory element text semantics; Among them, the regulatory decision processing algorithm is obtained by debugging the first long short-term memory algorithm based on multiple first training data, the first training data includes the original regulatory element text semantics of the second block regulatory target and the debugging learning basis, the debugging learning basis is the prior regulatory element text semantics of the second block regulatory target obtained by the debugged second long short-term memory algorithm based on the original regulatory element text semantics of the second block regulatory target, and the second block regulatory target is the regulated blockchain data entity; the debugged second long short-term memory algorithm is debugged by the second long short-term memory algorithm based on multiple second training data configured with prior knowledge, the second training data includes the original regulatory element text semantics of the regulatory participating entity and the regulated blockchain data entity, the prior knowledge represents whether the regulatory participating entity in the second training data triggers the risk supervision contract on the regulated blockchain data entity, the input information of the second long short-term memory algorithm includes the original regulatory element text semantics of each block regulatory target in the second training data, and the output is the derived regulatory element text semantics of each block regulatory target, and the debugging error information is determined based on the semantic commonality reinforcement coefficient between the derived regulatory element text semantics of the two block regulatory targets in each second training data.

10. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images