Multi-role branch chain node state linkage management method and system

Abstract

The invention discloses a multi-role branch chain node state linkage management method and system, and belongs to the field of node collaborative management, and the management method comprises the following specific steps: S101, setting role nodes, adding attribute information of each role node, setting a node basic identification rule, and generating a branch chain according to an intervention approach; the method ensures that different role node numbering systems are independent and not confused, fundamentally avoids the problems of serial number repetition, vacancy or logic fracture, avoids branch structure chaos after role node upgrading, ensures that the hierarchical relationship of nodes in branches before and after conversion is coherent, avoids misadjustment of irrelevant role nodes during exception handling, and reduces system resource consumption. The state adjustment efficiency is improved, manual operation and maintenance links are reduced, manual operation errors are avoided, and meanwhile the response speed of node management and state adjustment is improved.

Description

Technical Field

[0001] This invention relates to the field of node collaborative management, and in particular to a method and system for managing the status linkage of multi-role branch chain nodes. Background Technology

[0002] With the deepening of digital transformation, distributed collaboration systems are increasingly widely used in areas such as user hierarchical management, business process control, and supply chain collaboration. Multi-role branching architectures have become one of the mainstream architectures for such systems due to their ability to adapt to the hierarchical collaboration requirements of "core nodes - intermediate nodes - edge nodes." In this architecture, different role nodes undertake differentiated functions (such as management, execution, and interaction). The branching chain exhibits a complex form of multiple parallel branches and cross-branch connections as business expands. Node states also exhibit dynamic fluctuations due to changes in operating load, network environment, and business needs, placing higher demands on the system's refined management. Traditional management methods often focus on single aspects (such as role definition, sequence number allocation, or status monitoring), lacking a coordinated design for the entire "role-branch-status" process: the branch sequence number system is not updated synchronously when roles change, and the status of associated nodes is not adjusted when nodes are abnormal, leading to fragmented system management. This fragmented management not only increases manual maintenance costs but also easily causes data inconsistencies and process bottlenecks due to loopholes in the connection between links, making it difficult to meet the efficiency, stability, and consistency requirements of multi-role branching systems. Against this backdrop, there is an urgent need to build a multi-role branch chain node status linkage management method and system that covers the entire lifecycle of nodes and realizes "dynamic role adaptation, orderly branch management, and status linkage response" to solve the management pain points of the current multi-role branch chain system and support the stable operation of distributed collaboration scenarios.

[0003] Existing multi-role branch chain node status linkage management methods and systems are prone to problems such as confusion in the numbering system of nodes of different roles. At the same time, after role nodes are upgraded, the branch structure becomes chaotic, and it is impossible to guarantee the continuity of the node hierarchy relationship within the branch before and after the conversion. When handling anomalies, irrelevant role nodes may be mistakenly adjusted, increasing system resource consumption, reducing the efficiency of status adjustment, increasing the manual operation and maintenance process, and making it prone to human operation errors. In addition, the response time for node management and status adjustment is relatively long. To address these issues, we propose a multi-role branch chain node status linkage management method and system. Summary of the Invention

[0004] The purpose of this invention is to address the deficiencies in the existing technology by proposing a multi-role branch chain node state linkage management method and system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for managing the status linkage of multi-role branch chain nodes, the specific steps of which are as follows: S101: Set up role nodes, add attribute information for each role node, set basic node identification rules, and generate branch chains based on the intervention path; S102: Collect interaction data of role nodes, identify the relationship between each role node, dynamically adjust the sorting and conversion thresholds, and pre-allocate resources to each role node. S103: Real-time collection of node data to update status labels, dynamic adjustment of node status weights according to business scenarios, and synchronous adjustment of sequence numbers according to the propagation path when status changes; S104: Predict the abnormal propagation path and impact range of the branch chain, preprocess risk nodes, and dynamically allocate processing priorities according to the importance of nodes at each time point. S105: Records all process operation information and logs it locally. Then, it periodically uploads the log data to the blockchain for storage and performs closed-loop updates based on the log data.

[0006] As a further aspect of the present invention, the specific steps for generating the branch chain according to the intervention path in S101 are as follows: P1.1: Determine the range of role node types based on the business scenario, and define the function of each role node. At the same time, the specific hierarchical position of each role node in the branch chain is initially marked with numbers from 1 to 9. Then, based on the automatic role nodes, a reverse sequence number is generated from the lowest position of the corresponding level of the branch chain as a signal to trigger the starting point of the branch line, with "1" corresponding to the lowest sequence number of the branch chain and "9" corresponding to the highest level of the branch chain. P1.2: Determine the intervention method of the role node. If it is manual intervention, after the user fills in the node code, the role node will be treated as a branch of the existing main branch and an initial sequence number will be assigned to it. If there is no manual intervention, the role node will be automatically registered and generated, and the role node will be treated as a brand new independent branch and an initial sequence number will be assigned to it. P1.3: Collect all branch chain types, determine the list of branches to be included in the consensus network, calculate the minimum bandwidth requirement of each role node based on the information of each role node in the branch chain, divide the independent consensus resource pool from the total resources based on the minimum bandwidth, computing power and storage requirements of each role node, and then set the consensus node screening criteria. P1.4: Calculate the corresponding qualification score for each branch chain role node based on its stability, computing power level, and online time percentage. If the qualification score of the role node is lower than the consensus node screening standard, the role node will be removed from the branch list. Then, the role nodes that meet the consensus node screening standard will be sorted from high to low according to their qualification scores. At the same time, 1 to 2 role nodes with the highest qualification scores will be selected as the topology center, and the remaining role nodes will be the consensus nodes. P1.5: The coordinating node establishes a main link with all consensus nodes. Each consensus node connects with other nodes according to the minimum number of connections to establish a slave link in order to build a complete consensus network. The nodes in the consensus network communicate with each other in point-to-point to synchronize data in real time. At the same time, the link connection status between each node is monitored in real time. If any link is interrupted, the node automatically selects the unconnected node to make up the connection. P1.6: Real-time rule consistency checks are performed on each branch chain. At the same time, each consensus node periodically sends the rule consistency score and node data of the corresponding branch chain to the consensus network. If the rule consistency score of each branch chain is 100%, the cross-branch rule is determined to be consistent, and the basic data of all branches is saved synchronously. If the rule consistency score of any branch chain is <100%, the consensus network sends a unified rule template to the corresponding branch chain and automatically corrects the identifier or sequence number of the conflicting nodes.

[0007] As a further aspect of the present invention, the specific calculation formula for the reverse numbering in P1.1 is as follows: In the formula, Represents the reverse number of the automatic role node; The total number of nodes in the branch chain level where the representative role node is located, of which ; Represents the ascending order position of the automatic role node within the branch chain hierarchy; The specific calculation formula for the pass / fail score mentioned in P1.4 is as follows: In the formula, The qualification score for the representative role node; Represents the node stability score; Represents the computing power score of the node; The percentage of time a node is online; , as well as These represent the weighting coefficients of each indicator.

[0008] As a further embodiment of the present invention, the initial sequence number mentioned in P1.2 is uniformly set to C1, where C1 is the lowest bit identifier of the branch line, and is not limited to the C role. It can be adjusted as the role changes.

[0009] As a further aspect of the present invention, the specific steps of identifying the relationship between each role node and dynamically adjusting the sorting and conversion thresholds in S102 are as follows: P2.1: Collect the interaction data between each role node, and record the node identifier, collection timestamp, interaction index and branch chain number of each interaction data. After collection, delete the duplicate records of the same node and the same index at the same timestamp, and only keep the first collection value. P2.2: Check the completeness of the interaction data of each role node. If the missing interaction data of a role node is detected, the mean filling method of the same role node in the same branch is used to extract the corresponding interaction data of the same role node in the same branch and fill it with the mean. The interaction data of each type are arranged in ascending order. Then, the data values ​​at the 25th position and the 75th position are taken as the lower quartile Q1 and the upper quartile Q3 of the interaction data of that type. P2.3: Obtain the interquartile range (IQR) of each type of interaction data through Q3-Q1. Then, based on Q1, Q3 and IQR, establish the outlier range [Q1-1.5×IQR, Q3+1.5×IQR] for each type of interaction data. If any data in each type of interaction data exceeds the corresponding outlier range, mark it as an outlier and delete it, and replace it with the median. Then, use Min-Max standardization to map each interaction index of each interaction data to the interval [0, 1]. P2.4: Based on the current business scenario and the standardized interaction data of each role node, and based on preset rules, the relationships between each role node are divided into strong dependency, weak dependency, collaborative relationship and independent relationship. Then, each role node is used as a vertex and the relationship type is used as an edge to construct the node relationship graph in each branch chain, and each relationship type is marked with different colors. P2.5: Based on the standardized interaction indicators, calculate the indicator weight of each role node, and then obtain the information entropy of each interaction indicator according to the indicator weight of each role node. Subsequently, calculate the weight of each indicator based on the information entropy, and combine the weights of each interaction indicator to calculate the pairwise correlation strength between each node by weighted summation. P2.6: Divide the branch chain hierarchy into core layer, middle layer and edge layer, and assign a standardized hierarchy coefficient to each layer. Then, calculate the corresponding ranking score based on the comprehensive association strength and hierarchy coefficient of each role node. Subsequently, arrange the role nodes in the same branch chain in descending order according to the ranking score. P2.7: Sort the overall association strength of each role node in each branch chain in ascending order, and select the value Q3_S at the 75th position. Then calculate the average association strength of each branch chain. Based on Q3_S and the corresponding average association strength, determine the basic conversion threshold of each branch chain. Then adjust the basic conversion threshold according to the branch activity of each branch chain to obtain the final role conversion threshold. Collect the interaction data of each role node in each branch chain in real time to update the role node position and role conversion threshold periodically.

[0010] As a further aspect of the present invention, the interactive indicators described in P2.1 specifically include invitation interaction, response interaction, collaborative operation, and state linkage; Among them, the invitation interaction is the number of times each node actively invites other nodes per unit time, and its collection frequency is 10 minutes / time; The response interaction is the average time (in seconds) for each node to return a confirmation message after receiving an invitation from another node, and the collection frequency is 5 minutes / time; Collaborative operation is the number of branch chain tasks jointly completed by each node and other nodes within a unit of time, with a collection frequency of 1 hour / time; Status linkage is the percentage (in %) of other nodes that synchronously change their status within 10 minutes after a status change occurs at each node, with a collection frequency of 15 minutes / time; The strong dependency determination rule described in P2.4 is: the normalized value of invitation frequency ≥ 0.7 and the normalized value of response latency ≤ 0.3; Weak dependency determination rules: 0.4 ≤ normalized value of invitation frequency < 0.7, and 0.3 < normalized value of response latency ≤ 0.6; Collaborative relationship determination rules: The standardized value of the collaborative task completion amount is ≥0.6, and the standardized value of the state follow-up rate is ≥0.5; Independent relationship determination rule: the standardized value of invitation frequency < 0.4, and the standardized value of collaborative task completion amount < 0.4; If interactive data satisfies both types of rules, the type with the higher average metric is selected.

[0011] As a further aspect of the present invention, the specific steps for pre-allocating resources to each role node as described in S102 are as follows: P3.1: Extract the feature data of each role node from the interaction data and associations recorded in the node relationship graph of each branch chain, and label the sample labels of each role node. If the role node changes roles within a preset period, it is labeled as 1, otherwise it is labeled as 0. At the same time, feature-label sample pairs are established, and each sample pair is divided into training set and validation set according to the proportion. P3.2: Based on the logistic regression architecture, a corresponding conversion prediction model is established, and the cross-entropy function is used as the loss function of the conversion prediction model. Then, the training set is divided into multiple batches and input into the conversion prediction model in sequence. The conversion prediction model receives each batch of training set and processes the training set data of each batch layer by layer through the forward propagation algorithm, and outputs the predicted conversion probability of the corresponding role node undergoing role conversion. P3.3: Calculate the loss value between the predicted conversion probability and the corresponding role node sample label using the loss function, and input the obtained loss value from the output layer of the conversion prediction model. The loss value is then passed layer by layer through the backpropagation algorithm. At the same time, the parameters of the conversion prediction model are iteratively updated based on the gradient descent method. The conversion prediction model is trained and updated repeatedly until the change value of the conversion prediction model converges to the preset range after multiple rounds of training. The training stops, and then the validation set is input into the conversion prediction model to evaluate the discrimination ability of the trained conversion prediction model. P3.4: If the discrimination ability of the conversion prediction model is lower than the preset requirement, the conversion prediction model will be retrained and validated. Otherwise, the feature data of the role node to be evaluated will be input into the conversion prediction model, and the conversion probability of each role node will be calculated. If the conversion probability of a role node is higher than the preset business requirement threshold, the node will be determined as a high-risk conversion node and added to the resource pre-allocation range. At the same time, resources will be dynamically allocated according to the conversion probability. P3.5: If the conversion probability of a role node is higher than the preset business requirement threshold and passes manual confirmation or automatic verification, the role node conversion is triggered, the node role node status identifier is updated, and the pre-allocated resources are allocated to the role node. At the same time, the sequence number of the same role node in the branch is adjusted synchronously.

[0012] As a further aspect of the present invention, the specific calculation formula for the predicted conversion probability described on page 3.2 is as follows: In the formula, Represents the feature data The predicted probability of a node undergoing a role change; The intercept term represents the transformation prediction model; ~ Coefficients representing various types of characteristic data; ~ Feature data representing different types of role nodes; Represents the natural constant; The specific calculation formula for the cross-entropy function described on page 3.2 is as follows: In the formula, Represents the cross-entropy loss value, where These are model parameters; Represents the total number of samples in the training set; Representing the The true labels of each training sample; Representing the The predicted conversion probability of each training sample; This represents the natural logarithm function.

[0013] The multi-role branch chain node status linkage management system includes a rule setting module, an initial construction module, a data collection and processing module, a correlation analysis module, a dynamic sorting module, a transformation prediction module, a resource allocation module, a transformation adjustment module, a monitoring and propagation module, an optimization scheduling module, and a log tracing module. The rule setting module is used to classify role node types and functions, and allows users to manually set identification rules. The initial construction module is used to generate an initial branch chain structure according to the node intervention method, and to perform node sequence number allocation and cross-branch data synchronization; The data acquisition and processing module is used to collect various types of interactive data during the node interaction process and to perform cleaning and standardization processing. The association analysis module analyzes the association strength between role nodes in each branch chain based on the preprocessed interaction data, and constructs the corresponding node relationship graph. The dynamic sorting module is used to dynamically adjust the position of a node in the branch chain by combining the association strength and the node hierarchy attribute. The conversion prediction module is used to predict the conversion probability of each role node based on the historical node conversion situation; The resource allocation module is used to pre-allocate resources to high-risk conversion nodes based on the predicted conversion probability. The conversion adjustment module is used to perform role conversion based on preset conversion conditions and the predicted conversion probability of role nodes, and to simultaneously adjust the sequence numbers of other nodes in the branch chain. The monitoring and propagation module is used to monitor the node status in real time, dynamically allocate status weights, and filter propagation paths based on changes in the status of role nodes. The optimization scheduling module is used to predict the optimal propagation path of anomalies, assign priorities based on the importance and weight values ​​of role nodes, and perform anomaly handling. The log tracing module is used to record the operation information of the entire branch chain and store the operation information on the chain. At the same time, the log data is used as historical samples for closed-loop updates.

[0014] As a further aspect of the present invention, the optimization scheduling module predicts the optimal propagation path of anomalies and allocates priorities based on the importance and weight values ​​of role nodes, with the following specific steps: P4.1: Collect the relationship graph of each node, and calculate the edge weight of each edge in the relationship graph of each node according to the association strength and state weight of the corresponding role node, and establish the corresponding edge weight matrix. Calculate the attention coefficient of each role node to the corresponding neighbor node through the attention mechanism. P4.2: Based on the attention coefficient of the role node, the probability of each role node being affected by the abnormal role node at the current time is calculated by the sigmoid function. The role nodes with an influence probability higher than 0.5 are counted and used as the abnormal influence set at the current time. P4.3: Starting from the source role node of the anomaly, traverse the propagation path of all role nodes in the set of anomaly impacts and record each propagation path. Then calculate the path impact degree of each propagation path and select the propagation path with the smallest path propagation impact degree as the optimal anomaly propagation path. P4.4: Based on the probability of being affected, the importance coefficient, and the shortest path distance between the role node and the anomaly source role node, calculate the risk level score of each role node in the anomaly impact set. If the risk level score is ≥0.6, it is judged as a high-risk node; if 0.3≤risk level score<0.6, it is judged as a medium-risk node; if the risk level score<0.3, it is judged as a low-risk node. Differentiated preprocessing operations are performed for each role node's risk level. P4.5: Based on the risk level and real-time load of each role node with pending status changes, calculate its corresponding status change priority. If multiple role nodes have the same status change priority, then sort them in order of risk level, node importance and real-time load. All role nodes with pending status changes are arranged in descending order of status change priority. Then the scheduler processes node requests one by one according to the queue order.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention establishes node identification rules by setting role nodes and assigning attributes, and creates a branch chain structure based on node intervention methods. The hierarchy is determined by numbering nodes in reverse order from 1 to 9. All branch chains are then collected, and a consensus resource pool is divided based on bandwidth, computing power, and storage requirements. Qualified nodes are selected as consensus nodes, and some nodes serve as topology centers, forming a master-slave consensus network. This enables real-time data synchronization and rule consistency detection, automatically correcting conflicts. Subsequently, role node interaction data is collected and preprocessed to construct a node relationship graph, categorizing relationships into strong dependency, weak dependency, collaboration, and independence. Association strength is calculated, hierarchy is defined, and ranking scores and conversion thresholds are obtained. Thresholds are dynamically adjusted based on activity levels, and node ranking and role conversion conditions are updated periodically. High-risk nodes are identified and resources are pre-allocated. Once the conversion probability exceeds the threshold and is confirmed, role conversion, sequence number adjustment, and resource allocation are triggered. Subsequently, the abnormal propagation path and impact range of each abnormal node are predicted, the risk level is calculated, and differentiated preprocessing is implemented. Based on the risk level, importance, and real-time load allocation priority, the processing is carried out one by one through the scheduler to ensure that the numbering system of different role nodes is independent and not confused, thereby avoiding the problems of duplicate, missing, or logically broken sequence numbers from the root, avoiding the confusion of branch structure after role node upgrade, ensuring the continuity of node hierarchy within the branch before and after conversion, avoiding the mis-adjustment of irrelevant role nodes during anomaly handling, reducing system resource consumption, improving state adjustment efficiency, reducing manual operation and maintenance links, avoiding human operation errors, and improving the response speed of node management and state adjustment. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0017] Figure 1 This is a flowchart of the multi-role branch chain node status linkage management method proposed in this invention; Figure 2 This is a system block diagram of the multi-role branch chain node status linkage management system proposed in this invention; Figure 3 This is a basic user hierarchy diagram of the multi-role branch chain node status linkage management system proposed in this invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Example 1

[0019] Reference Figure 1 , 3This embodiment discloses a method for managing the status linkage of multi-role branch chain nodes. The specific steps of this management method are as follows: Set up role nodes, add attribute information for each role node, set basic node identification rules, and generate branch chains based on the intervention path.

[0020] Specifically, the range of role node types is determined based on the business scenario, and the function of each role node is defined. Each role node's specific hierarchical position in the branch chain is initially identified using numbers from 1 to 9. Then, based on the automatic role node identification, a reverse-order number is generated from the lowest digit of the corresponding level in the branch chain, based on the total number of nodes at that level. This number serves as the signal to trigger the branch line's starting point, where "1" corresponds to the lowest sequence number in the branch chain and "9" corresponds to the highest level. The intervention method of the role node is then determined. If it is manual intervention, the role node is designated as an existing main branch after the user enters the node code. The system automatically registers and generates role nodes for each branch, treating them as new, independent branches and assigning them initial sequence numbers. Without manual intervention, it collects all branch chain types, determines the list of branches to be included in the consensus network, and calculates the minimum bandwidth requirements for each role node based on its information. Based on the minimum bandwidth, computing power, and storage requirements of each role node, it allocates an independent consensus resource pool from the total resources. Then, it sets consensus node selection criteria, calculating the impact of each branch chain's role nodes on the network's stability, computing power level, and online time percentage. For each role node, if its qualification score is lower than the consensus node selection standard, it will be removed from the branch list. Then, all role nodes that meet the consensus node selection standard will be sorted from highest to lowest qualification score. One to two groups of role nodes with the highest qualification scores will be selected as topology centers, and the remaining role nodes will serve as consensus nodes. Coordinating nodes will establish main links with all consensus nodes. Each consensus node will connect to other nodes using the minimum number of connections required to establish slave links, thus building a complete consensus network. Nodes in the consensus network will communicate point-to-point to synchronize data in real time and monitor the link connections between nodes. If any link is interrupted, the node will automatically select an unconnected node to fill the connection gap. Real-time rule consistency checks will be performed on each branch chain. Each consensus node will periodically send its corresponding branch chain's rule consistency score and node data to the consensus network. If the rule consistency score of all branch chains is 100%, cross-branch rule consistency will be determined, and the basic data of all current branches will be saved simultaneously. If any branch chain's rule consistency score is <100%, the consensus network will send a unified rule template to the corresponding branch chain to automatically correct the identifiers or sequence numbers of conflicting nodes.

[0021] In addition, it should be noted that the specific formula for calculating the reverse numbering is as follows: In the formula, Represents the reverse number of the automatic role node; The total number of nodes in the branch chain level where the representative role node is located, of which ; Represents the ascending order position of the automatic role node within the branch chain hierarchy; The specific formula for calculating the pass / fail score is as follows: In the formula, The qualification score for the representative role node; Represents the node stability score; Represents the computing power score of the node; The percentage of time a node is online; , as well as These represent the weighting coefficients of each indicator; The initial sequence number is uniformly set to C1, where C1 is the lowest bit identifier of the branch line. It is not limited to the C role and can be adjusted later as the role changes.

[0022] Collect interaction data of role nodes, identify the relationships between role nodes, dynamically adjust the sorting and conversion thresholds, and pre-allocate resources to each role node.

[0023] Specifically, the system collects interaction data between various role nodes and records the node identifier, collection timestamp, interaction metric, and branch chain number for each interaction. After collection, duplicate records of the same node and metric at the same timestamp are deleted, retaining only the initial collection value. The system checks the completeness of interaction data for each role node. If missing interaction data is detected, the system uses the mean-filling method for nodes of the same role within the same branch to extract the corresponding interaction data and fill it with the mean. All types of interaction data are arranged in ascending order, and the data values ​​at the 25th and 75th percentiles are taken as the next four values ​​for that type of interaction data. Quantile Q1 and upper quartile Q3 are used to obtain the interquartile range (IQR) of each type of interaction data by subtracting Q3 from Q1. Then, based on Q1, Q3, and IQR, outlier ranges [Q1 - 1.5 × IQR, Q3 + 1.5 × IQR] are established for each type of interaction data. If any data in any type of interaction data exceeds the corresponding outlier range, it is marked as an outlier and deleted, replaced by the median. Min-Max standardization is then used to map each interaction indicator of each interaction data to the [0, 1] interval. Based on the current business scenario and the standardized interaction data for each role node, the relationships between each role node are classified into strong and weak categories according to preset rules. Dependency, weak dependency, collaborative relationship, and independent relationship are identified. Then, each role node is designated as a vertex, and the relationship type as an edge, constructing a node relationship graph within each branch chain. Different colors are used to identify each relationship type. Based on standardized interaction metrics, the metric weight of each role node is calculated. Then, based on the metric weight of each role node, the information entropy of each interaction metric is obtained. Subsequently, the weight of each metric is calculated based on the information entropy. Combining the weights of each interaction metric, a weighted sum is used to calculate the pairwise association strength between each node. The branch chain hierarchy is divided into core layer, middle layer, and edge layer, and a standardized hierarchy coefficient is assigned to each layer. Finally, based on the relationship between each role node... By combining the overall association strength and the hierarchical coefficient, the corresponding ranking score is calculated. Then, the role nodes within the same branch chain are sorted in descending order according to their ranking scores. The overall association strength of each role node within each branch chain is sorted from smallest to largest, and the value Q3_S at the 75th position is selected. Then, the average association strength of each branch chain is calculated. Based on Q3_S and the corresponding average association strength, the basic conversion threshold of each branch chain is determined. Then, the basic conversion threshold is adjusted according to the branch activity of each branch chain to obtain the final role conversion threshold. The interaction data of each role node within each branch chain is collected in real time to periodically update the role node position and role conversion threshold.

[0024] Specifically, feature data of each role node is extracted from the interaction data and relationships recorded in the node relationship graph of each branch chain, and sample labels are labeled for each role node. If a role node undergoes a role change within a preset period, it is labeled as 1; otherwise, it is labeled as 0. Feature-label sample pairs are established, and each sample pair is proportionally divided into training and validation sets. A corresponding conversion prediction model is built based on a logistic regression architecture, and the cross-entropy function is used as the loss function of the conversion prediction model. The training set is then divided into multiple batches and sequentially input into the conversion prediction model. The conversion prediction model receives each batch of training set data and processes the data layer by layer through the forward propagation algorithm, outputting the predicted conversion probability of the corresponding role node undergoing a role change. The loss value between the predicted conversion probability and the corresponding role node sample label is calculated using the loss function, and the obtained loss value is input from the output layer of the conversion prediction model and passed layer by layer through the backpropagation algorithm, while iteratively updating based on gradient descent. The conversion prediction model parameters are changed, and the conversion prediction model is trained and updated repeatedly until the change in the loss value of the conversion prediction model converges to a preset range after multiple rounds of training. Training stops then. The validation set is then input into the conversion prediction model to evaluate the discrimination ability of the trained conversion prediction model. If the discrimination ability of the conversion prediction model is lower than the preset requirement, the conversion prediction model is retrained and validated. Otherwise, the feature data of the role node to be evaluated is input into the conversion prediction model, and the conversion probability of each role node is calculated. If the conversion probability of a role node is higher than the preset business requirement threshold, the node is determined to be a high-risk conversion node and added to the resource pre-allocation range. At the same time, resources are dynamically allocated according to the conversion probability. If the conversion probability of a role node is higher than the preset business requirement threshold and passes manual confirmation or automatic verification, the role node conversion is triggered, the node role node status identifier is updated, and the pre-allocated resources are allocated to the role node. At the same time, the sequence number of the same role node in the branch is adjusted synchronously.

[0025] refer to Figure 3 As can be seen, the sequence number from 1 to 9 depends on the ranking of the branch. Different roles in different branches of the same node correspond to different sequence numbers, and countless branches with the same sequence number can share the same sequence number.

[0026] It should be further explained that the interaction metrics specifically include invitation interaction, response interaction, collaborative operation, and status linkage; Among them, the invitation interaction is the number of times each node actively invites other nodes per unit time, and its collection frequency is 10 minutes / time; The response interaction is the average time (in seconds) for each node to return a confirmation message after receiving an invitation from another node, and the collection frequency is 5 minutes / time; Collaborative operation is the number of branch chain tasks jointly completed by each node and other nodes within a unit of time, with a collection frequency of 1 hour / time; Status linkage is the percentage (in %) of other nodes that synchronously change their status within 10 minutes after a status change occurs at each node, with a collection frequency of 15 minutes / time; Strong dependency determination rule: the normalized value of invitation frequency is ≥0.7, and the normalized value of response latency is ≤0.3; Weak dependency determination rules: 0.4 ≤ normalized value of invitation frequency < 0.7, and 0.3 < normalized value of response latency ≤ 0.6; Collaborative relationship determination rules: The standardized value of the collaborative task completion amount is ≥0.6, and the standardized value of the state follow-up rate is ≥0.5; Independent relationship determination rule: the standardized value of invitation frequency < 0.4, and the standardized value of collaborative task completion amount < 0.4; If interactive data satisfies both types of rules, the type with the higher average index value shall be selected. The specific formula for calculating the predicted transition probability is as follows: In the formula, Represents the feature data The predicted probability of a node undergoing a role change; The intercept term represents the transformation prediction model; ~ Coefficients representing various types of characteristic data; ~ Feature data representing different types of role nodes; Represents the natural constant; The specific formula for calculating the cross-entropy function is as follows: In the formula, Represents the cross-entropy loss value, where These are model parameters; Represents the total number of samples in the training set; Representing the The true labels of each training sample; Representing the The predicted conversion probability of each training sample; This represents the natural logarithm function.

[0027] The system collects node data in real time to update status labels, dynamically adjusts node status weights based on business scenarios, and adjusts sequence numbers synchronously according to the propagation path when a status changes.

[0028] Predict the abnormal propagation path and impact range of the branch chain, preprocess risk nodes, and dynamically allocate processing priorities according to the importance of nodes at each time point.

[0029] Record all process operation information locally via logs, then periodically upload the log data to the blockchain for storage, and perform closed-loop updates based on the log data. Example 2

[0030] Reference Figure 2 This embodiment discloses a multi-role branch chain node status linkage management system, including a rule setting module, an initial construction module, a data collection and processing module, a correlation analysis module, a dynamic sorting module, a conversion prediction module, a resource allocation module, a conversion adjustment module, a monitoring and propagation module, an optimization scheduling module, and a log tracing module; The rule setting module is used to classify the types and functions of role nodes and to manually set identification rules by users; the initial construction module is used to generate the initial branch chain structure according to the node intervention method, and to assign node numbers and synchronize cross-branch data; the collection and processing module is used to collect various interaction data during the node interaction process and to clean and standardize them.

[0031] The association analysis module analyzes the association strength between role nodes in each branch chain based on the preprocessed interaction data and constructs the corresponding node relationship graph; the dynamic sorting module is used to dynamically adjust the position of nodes in the branch chain by combining association strength and node hierarchy attributes; the conversion prediction module is used to predict the role conversion probability of each role node based on historical node conversion.

[0032] The resource allocation module is used to pre-allocate resources to high-risk conversion nodes based on the predicted conversion probability; the conversion adjustment module is used to execute role conversion based on preset conversion conditions and the predicted conversion probability of role nodes, and simultaneously adjust the sequence numbers of other nodes in the branch chain.

[0033] The monitoring and propagation module is used to monitor node status in real time, dynamically allocate status weights, and filter propagation paths based on changes in the status of role nodes; the optimization and scheduling module is used to predict the optimal propagation path for anomalies, allocate priorities based on the importance and weight values ​​of role nodes, and handle anomalies.

[0034] Specifically, the process involves collecting the relationship graphs of each node, calculating the edge weights of each edge in the graph based on the association strength and state weights of the corresponding role nodes, and establishing a corresponding edge weight matrix. An attention coefficient for each role node on its corresponding neighbor nodes is calculated using an attention mechanism. Based on these attention coefficients, the probability of each role node being influenced by an abnormal role node at the current time is calculated using the sigmoid function. Role nodes with an influence probability higher than 0.5 are identified and included as the abnormal influence set at the current time. Starting from the source of the abnormality, the propagation paths of all role nodes in the abnormal influence set are traversed, and each propagation path is recorded. Then, the path influence degree of each propagation path is calculated, and the propagation path with the lowest path propagation influence degree is selected as the optimal abnormal propagation path, based on the probability of being influenced and its importance. The risk level score of each role node in the anomaly impact set is calculated using the coefficient and the shortest path distance between the role node and the anomaly source role node. If the risk level score is ≥0.6, it is judged as a high-risk node; if 0.3≤risk level score<0.6, it is judged as a medium-risk node; and if the risk level score<0.3, it is judged as a low-risk node. Differentiated preprocessing operations are performed for each role node's risk level. Based on the risk level and real-time load of each role node with pending state changes, its corresponding state change priority is calculated. If multiple role nodes have the same state change priority, they are sorted in order of risk level, node importance, and real-time load. All role nodes with pending state changes are then arranged in descending order of state change priority. The scheduler then processes node requests one by one according to the queue order.

[0035] The log tracing module is used to record the operation information of the entire branch chain and store the operation information on the chain. At the same time, the log data is used as historical samples for closed-loop updates.

Claims

1. A method for managing the status linkage of multi-role branch chain nodes, characterized in that, The specific steps of this management method are as follows: S101: Set up role nodes, add attribute information for each role node, set basic node identification rules, and generate branch chains based on the intervention path; S102: Collect interaction data of role nodes, identify the relationship between each role node, dynamically adjust the sorting and conversion thresholds, and pre-allocate resources to each role node. S103: Real-time collection of node data to update status labels, dynamic adjustment of node status weights according to business scenarios, and synchronous adjustment of sequence numbers according to the propagation path when status changes; S104: Predict the abnormal propagation path and impact range of the branch chain, preprocess risk nodes, and dynamically allocate processing priorities according to the importance of nodes at each time point. S105: Records all process operation information and logs it locally. Then, it periodically uploads the log data to the blockchain for storage and performs closed-loop updates based on the log data.

2. The multi-role branch chain node status linkage management method according to claim 1, characterized in that, The specific steps for generating a branch chain based on the intervention path, as described in S101, are as follows: P1.1: Determine the range of role node types based on the business scenario and define the function of each role node. At the same time, each role node is initially marked with numbers from 1 to 9 at its specific level in the branch chain. Then, based on the automatic role nodes, a reverse sequence number is generated from the lowest position of the corresponding level of the branch chain as a signal to trigger the starting point of the branch line, with "1" corresponding to the lowest sequence number of the branch chain and "9" corresponding to the highest level of the branch chain. P1.2: Determine the intervention method of the role node. If it is manual intervention, after the user fills in the node code, the role node will be treated as a branch of the existing main branch and an initial sequence number will be assigned to it. If there is no manual intervention, the role node will be automatically registered and generated, and the role node will be treated as a brand new independent branch and an initial sequence number will be assigned to it. P1.3: Collect all branch chain types, determine the list of branches to be included in the consensus network, calculate the minimum bandwidth requirement of each role node based on the information of each role node in the branch chain, divide the independent consensus resource pool from the total resources based on the minimum bandwidth, computing power and storage requirements of each role node, and then set the consensus node screening criteria. P1.4: Calculate the corresponding qualification score for each branch chain role node based on its stability, computing power level, and online time percentage. If the qualification score of the role node is lower than the consensus node screening standard, the role node will be removed from the branch list. Then, the role nodes that meet the consensus node screening standard will be sorted from high to low according to their qualification scores. At the same time, 1 to 2 role nodes with the highest qualification scores will be selected as the topology center, and the remaining role nodes will be the consensus nodes. P1.5: The coordinating node establishes a main link with all consensus nodes. Each consensus node connects with other nodes according to the minimum number of connections to establish a slave link in order to build a complete consensus network. The nodes in the consensus network communicate with each other in point-to-point to synchronize data in real time. At the same time, the link connection status between each node is monitored in real time. If any link is interrupted, the node automatically selects the unconnected node to make up the connection. P1.6: Real-time rule consistency checks are performed on each branch chain. At the same time, each consensus node periodically sends the rule consistency score and node data of the corresponding branch chain to the consensus network. If the rule consistency score of each branch chain is 100%, the cross-branch rule is determined to be consistent, and the basic data of all branches is saved synchronously. If the rule consistency score of any branch chain is <100%, the consensus network sends a unified rule template to the corresponding branch chain and automatically corrects the identifier or sequence number of the conflicting nodes.

3. The multi-role branch chain node status linkage management method according to claim 2, characterized in that, The specific calculation formula for the reverse numbering mentioned in P1.1 is as follows: In the formula, Represents the reverse number of the automatic role node; The total number of nodes in the branch chain level where the representative role node is located, of which ; Represents the ascending order position of the automatic role node within the branch chain hierarchy; The specific calculation formula for the pass / fail score mentioned in P1.4 is as follows: In the formula, The qualification score for the representative role node; Represents the node stability score; Represents the computing power score of the node; The percentage of time a node is online; , as well as These represent the weighting coefficients of each indicator.

4. The multi-role branch chain node status linkage management method according to claim 2, characterized in that, The specific steps for identifying the relationships between various role nodes and dynamically adjusting the sorting and conversion thresholds described in S102 are as follows: P2.1: Collect the interaction data between each role node, and record the node identifier, collection timestamp, interaction index and branch chain number of each interaction data. After collection, delete the duplicate records of the same node and the same index at the same timestamp, and only keep the first collection value. P2.2: Check the completeness of the interaction data of each role node. If the missing interaction data of a role node is detected, the mean filling method of the same role node in the same branch is used to extract the corresponding interaction data of the same role node in the same branch and fill it with the mean. The interaction data of each type are arranged in ascending order. Then, the data values ​​at the 25th position and the 75th position are taken as the lower quartile Q1 and the upper quartile Q3 of the interaction data of that type. P2.3: Obtain the interquartile range (IQR) of each type of interaction data through Q3-Q1. Then, based on Q1, Q3 and IQR, establish the outlier range [Q1-1.5×IQR, Q3+1.5×IQR] for each type of interaction data. If any data in each type of interaction data exceeds the corresponding outlier range, mark it as an outlier and delete it, and replace it with the median. Then, use Min-Max standardization to map each interaction index of each interaction data to the interval [0, 1]. P2.4: Based on the current business scenario and the standardized interaction data of each role node, and based on preset rules, the relationships between each role node are divided into strong dependency, weak dependency, collaborative relationship and independent relationship. Then, each role node is used as a vertex and the relationship type is used as an edge to construct the node relationship graph in each branch chain, and each relationship type is marked with different colors. P2.5: Based on the standardized interaction indicators, calculate the indicator weight of each role node, and then obtain the information entropy of each interaction indicator according to the indicator weight of each role node. Subsequently, calculate the weight of each indicator based on the information entropy, and combine the weights of each interaction indicator to calculate the pairwise correlation strength between each node by weighted summation. P2.6: Divide the branch chain hierarchy into core layer, middle layer and edge layer, and assign a standardized hierarchy coefficient to each layer. Then, calculate the corresponding ranking score based on the comprehensive association strength and hierarchy coefficient of each role node. Subsequently, arrange the role nodes in the same branch chain in descending order according to the ranking score. P2.7: Sort the overall association strength of each role node in each branch chain in ascending order, and select the value Q3_S at the 75th position. Then calculate the average association strength of each branch chain. Based on Q3_S and the corresponding average association strength, determine the basic conversion threshold of each branch chain. Then adjust the basic conversion threshold according to the branch activity of each branch chain to obtain the final role conversion threshold. Collect the interaction data of each role node in each branch chain in real time to update the role node position and role conversion threshold periodically.

5. The multi-role branch chain node status linkage management method according to claim 4, characterized in that, The specific steps for pre-allocating resources to each role node as described in S102 are as follows: P3.1: Extract the feature data of each role node from the interaction data and associations recorded in the node relationship graph of each branch chain, and label the sample labels of each role node. If the role node changes roles within a preset period, it is labeled as 1, otherwise it is labeled as 0. At the same time, feature-label sample pairs are established, and each sample pair is divided into training set and validation set according to the proportion. P3.2: Based on the logistic regression architecture, a corresponding conversion prediction model is established, and the cross-entropy function is used as the loss function of the conversion prediction model. Then, the training set is divided into multiple batches and input into the conversion prediction model in sequence. The conversion prediction model receives each batch of training set and processes the training set data of each batch layer by layer through the forward propagation algorithm, and outputs the predicted conversion probability of the corresponding role node undergoing role conversion. P3.3: Calculate the loss value between the predicted conversion probability and the corresponding role node sample label using the loss function, and input the obtained loss value from the output layer of the conversion prediction model. The loss value is then passed layer by layer through the backpropagation algorithm. At the same time, the parameters of the conversion prediction model are iteratively updated based on the gradient descent method. The conversion prediction model is trained and updated repeatedly until the change value of the conversion prediction model converges to the preset range after multiple rounds of training. The training stops, and then the validation set is input into the conversion prediction model to evaluate the discrimination ability of the trained conversion prediction model. P3.4: If the discrimination ability of the conversion prediction model is lower than the preset requirement, the conversion prediction model will be retrained and validated. Otherwise, the feature data of the role node to be evaluated will be input into the conversion prediction model, and the conversion probability of each role node will be calculated. If the conversion probability of a role node is higher than the preset business requirement threshold, the node will be determined as a high-risk conversion node and added to the resource pre-allocation range. At the same time, resources will be dynamically allocated according to the conversion probability. P3.5: If the conversion probability of a role node is higher than the preset business requirement threshold and passes manual confirmation or automatic verification, the role node conversion is triggered, the node role node status identifier is updated, and the pre-allocated resources are allocated to the role node. At the same time, the sequence number of the same role node in the branch is adjusted synchronously.

6. The multi-role branch chain node status linkage management method according to claim 5, characterized in that, The specific calculation formula for the predicted transition probability described in P3.2 is as follows: In the formula, Represents the feature data The predicted probability of a node undergoing a role change; The intercept term represents the transformation prediction model; ~ Coefficients representing various types of characteristic data; ~ Feature data representing different types of role nodes; Represents the natural constant; The specific calculation formula for the cross-entropy function described on page 3.2 is as follows: In the formula, Represents the cross-entropy loss value, where These are model parameters; Represents the total number of samples in the training set; Representing the The true labels of each training sample; Representing the The predicted conversion probability of each training sample; This represents the natural logarithm function.

7. A multi-role branch chain node state linkage management system, used to implement the multi-role branch chain node state linkage management method according to any one of claims 1-6, characterized in that, It includes a rule setting module, an initial construction module, a data collection and processing module, a correlation analysis module, a dynamic sorting module, a transformation prediction module, a resource allocation module, a transformation adjustment module, a monitoring and propagation module, an optimization and scheduling module, and a log tracing module; The rule setting module is used to classify role node types and functions, and allows users to manually set identification rules. The initial construction module is used to generate an initial branch chain structure according to the node intervention method, and to perform node sequence number allocation and cross-branch data synchronization; The data acquisition and processing module is used to collect various types of interactive data during the node interaction process and to perform cleaning and standardization processing. The association analysis module analyzes the association strength between role nodes in each branch chain based on the preprocessed interaction data, and constructs the corresponding node relationship graph. The dynamic sorting module is used to dynamically adjust the position of a node in the branch chain by combining the association strength and the node hierarchy attribute. The conversion prediction module is used to predict the conversion probability of each role node based on the historical node conversion situation; The resource allocation module is used to pre-allocate resources to high-risk conversion nodes based on the predicted conversion probability. The conversion adjustment module is used to perform role conversion based on preset conversion conditions and the predicted conversion probability of role nodes, and to simultaneously adjust the sequence numbers of other nodes in the branch chain. The monitoring and propagation module is used to monitor the node status in real time, dynamically allocate status weights, and filter propagation paths based on changes in the status of role nodes. The optimization scheduling module is used to predict the optimal propagation path of anomalies, assign priorities based on the importance and weight values ​​of role nodes, and perform anomaly handling. The log tracing module is used to record the operation information of the entire branch chain and store the operation information on the chain. At the same time, the log data is used as historical samples for closed-loop updates.

8. The multi-role branch chain node status linkage management system according to claim 7, characterized in that, The optimization scheduling module predicts the optimal propagation path of anomalies and assigns priorities based on the importance and weight values ​​of role nodes. The specific steps are as follows: P4.1: Collect the relationship graph of each node, and calculate the edge weight of each edge in the relationship graph of each node according to the association strength and state weight of the corresponding role node, and establish the corresponding edge weight matrix. Calculate the attention coefficient of each role node to the corresponding neighbor node through the attention mechanism. P4.2: Based on the attention coefficient of the role node, the probability of each role node being affected by the abnormal role node at the current time is calculated by the sigmoid function. The role nodes with an influence probability higher than 0.5 are counted and used as the abnormal influence set at the current time. P4.3: Starting from the source role node of the anomaly, traverse the propagation path of all role nodes in the set of anomaly impacts and record each propagation path. Then calculate the path impact degree of each propagation path and select the propagation path with the smallest path propagation impact degree as the optimal anomaly propagation path. P4.4: Based on the probability of being affected, the importance coefficient, and the shortest path distance between the role node and the anomaly source role node, calculate the risk level score of each role node in the anomaly impact set. If the risk level score is ≥0.6, it is judged as a high-risk node; if 0.3≤risk level score<0.6, it is judged as a medium-risk node; if the risk level score<0.3, it is judged as a low-risk node. Differentiated preprocessing operations are performed for each role node's risk level. P4.5: Based on the risk level and real-time load of each role node with pending status changes, calculate its corresponding status change priority. If multiple role nodes have the same status change priority, then sort them in order of risk level, node importance and real-time load. All role nodes with pending status changes are arranged in descending order of status change priority. Then the scheduler processes node requests one by one according to the queue order.

Images