A server-based distributed block number diversion method

Abstract

The invention discloses a distributed block digital diversion method based on a server, which belongs to the field of block digital data management. The specific steps of the digital diversion method are as follows: S1: collecting digital node data and extracting data features; S2: statistical data and identifying and classifying new source data; S3: optimizing the layout of digital data storage and accurate calculation; S4: regularly statistically updating the data of each block and calculating the total number of outflow values; the invention can accurately divert data and reduce the complexity of maintenance, improve the overall digital management and sustainability of the system, effectively improve the efficiency and speed of digital data diversion, improve digital value in a targeted manner, reduce the digital value cost of users, exclude duplicate data and invalid data, optimize the digital active utilization rate of data, save the cost of wasted data resources, and can better exert the flexibility, scalability and value of digitalization.

Description

Technical Field

[0001] The present invention relates to the field of block digital data management, and in particular to a server-based distributed block digital data diversion method. Background Art

[0002] With the rapid development of Internet technology and the innovation of multifunctional server applications, its application in finance, supply chain management, smart contracts and other fields has been continuously expanded, attracting more and more attention and investment. The digital, diversified and highly secure characteristics of distributed blocks make it an ideal choice for many industries to solve trust and data security issues. However, with the increase in the scale of distributed block digital networks and the amount of data transactions, the traditional distributed block digital storage method faces some challenges. In the traditional central server system, data is usually stored in each node of the distributed network, and the nodes use consensus algorithms to maintain data consistency. Although this distributed block digital storage method has certain advantages, such as the traditional central server unified centralization and anti-attack, it also has some problems. First, the synchronization and transmission of data between nodes may be limited by network bandwidth and latency, resulting in slow data transactions. Secondly, the storage capacity and performance may be uneven, which may become a bottleneck of the system and affect the performance of the overall system. In addition, the communication and consensus algorithms between nodes also consume a lot of computing resources and energy; therefore, it is particularly important to invent a method for innovating the functions of a server-based distributed block digital diversion method.

[0003] The existing block number diversion method has high complexity in data management and maintenance, which reduces the overall manageability and maintainability of the system, reduces the efficiency and speed of data diversion, and increases the delay in user access to data. To this end, we propose a server-based distributed block number diversion method. Summary of the invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose a server-based distributed block number diversion method.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A distributed block digital diversion method based on a server, the specific steps of the digital diversion method are as follows:

[0007] S1: Collect digital node data and extract data features;

[0008] S2: Count data and identify and classify new source data;

[0009] S3: Optimize the layout and precise calculation of digital data storage;

[0010] S4: Regularly update the data of each block and calculate the total number of outflow values.

[0011] As a further solution of the present invention, the specific steps of data feature extraction in S1 are as follows:

[0012] Step 1: Based on the central function operation information of the server, the server storage area maintains the accumulated data of the node at the previous time, and records the accumulated data as Group A 1, collects the new data between the current time point and the previous time point of the node, and records the new data as Group A 2;

[0013] Step 2: Collect the data from Group A 1 and Group A 2, preprocess the collected data, divide the preprocessed data into training set, validation set and test set, and then build a CNN network model based on the TensorFlow deep learning framework;

[0014] Step 3: The CNN network model receives the training set data, and performs convolution operations with each group of data through the convolution layer to generate multiple groups of feature maps, and then processes each group of feature maps through the ReLU activation function, downsampling the activated feature maps to reduce the dimension of the feature maps;

[0015] Step 4: Flatten the feature map into a one-dimensional vector, multiply the feature vector by the weight matrix to obtain the final output vector, convert the original output of the network into a probability distribution as the model output through the softmax activation function, use the cross entropy loss function to calculate the loss value between the model output and the true label, and then start from the output layer and calculate the gradient of each parameter in the network according to the loss value;

[0016] Step 5: Update the parameters in the network according to the gradient through stochastic gradient descent. After each training cycle, verify the model performance through the validation set data. When the model performance no longer improves, interrupt the training process in advance. Otherwise, repeat the training of the CNN network model until the preset number of training times is reached, and then evaluate the final performance and generalization ability of the model through the test set.

[0017] Step 6: The trained CNN network model uses the learned convolutional layer and pooling layer to extract the data features in Group A1 and Group A2, processes the extracted feature information through feature fusion or feature selection, and represents the extracted features in vector or matrix form.

[0018] As a further solution of the present invention, the specific steps of identification and classification in S2 are as follows:

[0019] Step 1: Count the numbers and total data of Group A 1 and Group A 2, and store the calculated total data in the server area B, and record the data collection method of area B, which will be used as the recording digital mechanism of the block module of the data entering the server;

[0020] Step 2: The server initializes a group of immune cells based on the number of features extracted from each group of data, and represents each immune cell as a potential category or pattern, and calculates the affinity between the data from the new source and each immune cell through cosine similarity;

[0021] Step 3: Based on the calculated affinity, the immune cells whose affinity exceeds the preset threshold are cloned to generate multiple sets of copies, and the immune cells whose affinity is lower than the preset threshold are mutated. After cloning and mutation, multiple sets of representative cells are selected from the immune cell population through a competitive selection strategy as the final classifier;

[0022] Step 4: Calculate the affinity between the new data and each group of final classifiers, classify the data from the new source, output the corresponding category label or probability distribution based on the classification results of the classifier on the new data, and filter out different data based on the output results. Then, based on the identified identical data, update Group A1 and Group A2 data, and then input them into the server's set B area for storage.

[0023] As a further solution of the present invention, the specific steps of digital data storage layout and precise calculation optimization in S3 are as follows:

[0024] Step ①: Initialize the storage layout of the block digital data by randomly generating it and use it as the initial solution. Then, according to the delay of different block data diversions, define an energy function to measure the quality of the current layout, and then initialize a set of temperature parameters and annealing rate parameters;

[0025] Step ②: In each iteration, a set of neighborhood solutions is randomly generated by exchange, insertion and deletion, and the energy difference between the new solution and the current solution is calculated through the energy function. If the energy of the new solution is lower, the new solution is accepted and the current solution information is updated. Otherwise, the acceptance probability P of the worse solution is calculated by the Metropolis criterion, and a random number r is generated. If r≤P, the worse solution is accepted, otherwise the worse solution is rejected.

[0026] Step ③: After each iteration, the temperature is gradually reduced according to the annealing rate parameter. When the current temperature is reduced to the preset termination temperature, the iteration is stopped, and the storage layout corresponding to the optimal solution is output as the optimized block digital data storage layout.

[0027] As a further solution of the present invention, the specific steps for calculating the total number of outflow values ​​in S4 are as follows:

[0028] Step I: The server center performs effective automatic statistics and calculations at the preset operation cycle and time end point, summing up the total number of Group A 1 data in the accumulated blocks and the total number of Group A 2 data added to the server's Area B database;

[0029] Step II: According to the sum of the numbers in the block B area of ​​the same number identified by the server center and the needs of the block numbers, set the golden point ratio of the database in the storage area in the server. After periodic statistics, the remainder of the numbers that receive valid command signals in the server center for effective calculations will be used as the total number of outflow values ​​in the mechanism, and summarized as the sum of the K area data in the numerical area flowing out of the server. The remaining digital data in the storage area will be used as the digital value of the mechanism for the next server center cloud computing;

[0030] Step III: When the server center identifies the data of group B blocks, the data of group B is summarized into the total of the H zone database in the server. At the same time, the sum of the data in zone H is proportional to the amount of data in zone B, and the sum of the numbers in zone H must be greater than the sum of the numbers in zone B. Then, the block number entry and exit mechanism is executed in the server, and the data in zones B and H are updated synchronously according to the data identified by the server center function.

[0031] As a further solution of the present invention, the specific calculation formula of the outflow value in step II is as follows:

[0032]

[0033] In the formula, K n Represents the outflow value of K zone; G 甲1 Represents the cumulative data of Group A 1 in Zone B; N 甲2 Group A 2 represents the newly added data of Zone B; K o represents the outflow value of the previous K zone; R% represents the golden point ratio; G H Represents the total data of Group B in Zone H.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The server-based distributed block number diversion method is based on the central function operation information of the server. The server storage area maintains the cumulative data of the node at the previous time, and records the cumulative data as Group A 1. The new data between the current time point and the previous time point of the node is collected, and the new data is recorded as Group A 2. The numbers and total data of Group A 1 and Group A 2 are counted, and the calculated total data is stored in the server area B. The server center performs effective automatic statistics and calculations at the preset operation cycle and time end point. After periodic statistics, the remainder numbers that receive effective command signals in the server center for effective calculations are used as the total number of outflow values ​​in the mechanism, and are summarized as the sum of the K-area data of the numerical area flowing out of the server. At the same time, As the mechanism of the next server hub cloud computing, it outputs digital values. When the server hub identifies the data of group B blocks, it summarizes the data of group B into the total of the H zone database in the server, and then executes the block digital entry and exit mechanism in the server. According to the data identified by the server hub function, the data of zones B and H are updated synchronously, which can accurately divert data and reduce the complexity of maintenance, improve the overall digital management and sustainability of the system, effectively improve the efficiency and speed of digital data diversion, enhance digital value in a targeted manner, reduce the digital value cost of users, eliminate duplicate data and invalid data, optimize the digital active utilization rate of data, save the cost of wasted data resources, and better exert the flexibility, scalability and value of digitalization. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0037] Figure 1 The present invention is a flowchart of a server-based distributed block number diversion method. DETAILED DESCRIPTION

[0038] Example 1

[0039] Reference Figure 1 , a distributed block digital diversion method based on a server, the specific steps of the digital diversion method are as follows:

[0040] Collect data from each node of the blockchain and extract data features.

[0041] Specifically, according to the central function operation information of the server, the server storage area maintains the cumulative data of the node at the previous time, and records the cumulative data volume as Group A 1, collects the new data between the current time point and the previous time point of the node, and records the new data as Group A 2, collects each group of data in Group A 1 and Group A 2, and pre-processes the collected data. Thereafter, the pre-processed data is divided into a training set, a validation set, and a test set. A CNN network model is then constructed based on the TensorFlow deep learning framework. The CNN network model receives the training set data, and performs convolution operations with each group of data through a convolution layer to generate multiple groups of feature maps. Each group of feature maps is then processed through a ReLU activation function, and the activated feature map is downsampled to reduce the dimension of the feature map. The feature map is flattened into a one-dimensional vector, and the feature vector is multiplied by a weight matrix to obtain the final input. Output a vector, convert the original output of the network into a probability distribution as the model output through the softmax activation function, use the cross entropy loss function to calculate the loss value between the model output and the true label, and then start from the output layer, calculate the gradient of each parameter in the network according to the loss value, and update the parameters in the network according to the gradient through stochastic gradient descent. After each training cycle, verify the model performance through the validation set data. When the model performance no longer improves, interrupt the training process in advance. Otherwise, repeat the training of the CNN network model until the preset number of training times is reached, and then evaluate the final performance and generalization ability of the model through the test set. The trained CNN network model uses the learned convolutional layer and pooling layer to extract the data features in Group A 1 and Group A 2, and processes the extracted feature information through feature fusion or feature selection, and represents the extracted features in vector or matrix form.

[0042] Statistical data and identify and categorize new source data.

[0043] Specifically, the number and total data of Group A1 and Group A2 are counted, and the calculated total data is stored in Area B of the server. At the same time, the data collection method of Area B is recorded and used as a recording digital mechanism for data entering the block module of the server. The server initializes a group of immune cells based on the number of features of each group of data extracted, and each immune cell represents a potential category or pattern. The affinity between the data from the new source and each immune cell is calculated by cosine similarity. According to the calculated affinity, the immune cells whose affinity exceeds the preset threshold are cloned to generate multiple groups of copies, and the immune cells whose affinity is lower than the preset threshold are mutated. After cloning and mutation, multiple groups of representative cells are selected from the immune cell group through a competitive selection strategy as the final classifier, and the affinity between the new data and each group of final classifiers is calculated. The data from the new source is classified, and the corresponding category label or probability distribution is output according to the classification result of the classifier on the new data, and the different data are screened out according to the output result. Then, according to the identified identical data, the data of Group A1 and Group A2 are updated, and then input into the storage of Area B set by the server.

[0044] Example 2

[0045] Reference Figure 1 , a distributed block digital diversion method based on a server, the specific steps of the digital diversion method are as follows:

[0046] Optimize the layout and arrangement of data storage.

[0047] Specifically, the storage layout of the block digital data is initialized by random generation and used as the initial solution. Then, according to the delay of different block data diversions, an energy function that measures the quality of the current layout is defined, and then a set of temperature parameters and annealing rate parameters are initialized. In each iteration, a set of neighborhood solutions is randomly generated by exchange, insertion and deletion, and the energy difference between the new solution and the current solution is calculated by the energy function. If the energy of the new solution is lower, the new solution is accepted and the current solution information is updated. Otherwise, the acceptance probability P of the worse solution is calculated by the Metropolis criterion, and a random number r is generated. If r≤P, the worse solution is accepted, otherwise the worse solution is rejected. After each iteration, the temperature is gradually reduced according to the annealing rate parameter. When the current temperature is reduced to the preset termination temperature, the iteration is stopped, and the storage layout corresponding to the optimal solution is output as the optimized block digital data storage layout.

[0048] Regularly update block data and calculate the total number of outflow values.

[0049] Specifically, the server center performs effective automatic statistics and calculations at the preset operating cycle and time end point, summarizes the total of the Group A 1 data in the accumulated blocks and the countless newly added Group A 2 data into the total of the B area database in the server, and sets the golden point ratio of the database in the storage area in the server according to the total of the numbers in the B area of ​​the block with the same number recognized by the server center and the needs of the block numbers. After periodic statistics, the remainder numbers that are effectively calculated after receiving the effective command signal in the server center are used as the total number of outflow values ​​in the mechanism, and summarized as the sum of the K area data in the numerical area flowing out of the server. The remaining digital data in the storage area are used as the digital values ​​of the mechanism for the next server center cloud computing. In the data recognition of the Group B blocks by the server center, the Group B data is summarized into the total of the H area database in the server, and the sum of the data in the H area is proportional to the corresponding data volume in the B area, and the digital sum of the H area must be greater than the digital sum of the B area. The block number entry and exit mechanism is then executed in the server, and the data in the B and H areas are updated synchronously according to the data recognized by the server center function.

[0050] It should be further explained that the specific calculation formula for the outflow value is as follows:

[0051]

[0052] In the formula, K n Represents the outflow value of K zone; G 甲1 Represents the cumulative data of Group A 1 in Zone B; N 甲2 Group A 2 represents the newly added data of Zone B; K o represents the outflow value of the previous K zone; R% represents the golden point ratio; G H Represents the total data of Group B in Zone H.

Claims

1. A server-based distributed block number diversion method, characterized in that: The specific steps of the digital splitting method are as follows: S1: Collect digital node data and extract data features; S2: Count data and identify and classify new source data; S3: Optimize the layout and precise calculation of digital data storage; S4: Regularly update the data of each block and calculate the total number of outflow values; The specific steps of data feature extraction described in S1 are as follows: Step 1: Based on the central function operation information of the server, the server storage area maintains the accumulated data of the node at the previous time, and records the accumulated data as Group A 1, collects the new data between the current time point and the previous time point of the node, and records the new data as Group A 2; Step 2: Collect the data from Group A 1 and Group A 2, preprocess the collected data, divide the preprocessed data into training set, validation set and test set, and then build a CNN network model based on the TensorFlow deep learning framework; Step 3: The CNN network model receives the training set data, and performs convolution operations with each group of data through the convolution layer to generate multiple groups of feature maps, and then processes each group of feature maps through the ReLU activation function, downsampling the activated feature maps to reduce the dimension of the feature maps; Step 4: Flatten the feature map into a one-dimensional vector, multiply the feature vector by the weight matrix to obtain the final output vector, convert the original output of the network into a probability distribution as the model output through the softmax activation function, use the cross entropy loss function to calculate the loss value between the model output and the true label, and then start from the output layer and calculate the gradient of each parameter in the network according to the loss value; Step 5: Update the parameters in the network according to the gradient through stochastic gradient descent. After each training cycle, verify the model performance through the validation set data. When the model performance no longer improves, interrupt the training process in advance. Otherwise, repeat the training of the CNN network model until the preset number of training times is reached, and then evaluate the final performance and generalization ability of the model through the test set. Step 6: The trained CNN network model uses the learned convolutional layer and pooling layer to extract the data features in Group A 1 and Group A 2, processes the extracted feature information through feature fusion or feature selection, and represents the extracted features in vector or matrix form; The specific steps of identification and classification described in S2 are as follows: Step 1: Count the numbers and total data of Group A 1 and Group A 2, and store the calculated total data in the server area B, and record the data collection method of area B, which will be used as the recording digital mechanism of the block module of the data entering the server; Step 2: The server initializes a group of immune cells based on the number of features extracted from each group of data, and represents each immune cell as a potential category or pattern, and calculates the affinity between the data from the new source and each immune cell through cosine similarity; Step 3: Based on the calculated affinity, the immune cells whose affinity exceeds the preset threshold are cloned to generate multiple sets of copies, and the immune cells whose affinity is lower than the preset threshold are mutated. After cloning and mutation, multiple sets of representative cells are selected from the immune cell population through a competitive selection strategy as the final classifier; Step 4: Calculate the affinity between the new data and each group of final classifiers, classify the new source data, output the corresponding category label or probability distribution according to the classification result of the classifier on the new data, and filter out the different data according to the output result. Then, based on the identified identical data, update the data of Group A1 and Group A2, and then input them into the storage of Area B set up by the server; The specific steps of digital data storage layout and precise calculation optimization described in S3 are as follows: Step ①: Initialize the storage layout of the block digital data by randomly generating it and use it as the initial solution. Then, according to the delay of different block data diversions, define an energy function to measure the quality of the current layout, and then initialize a set of temperature parameters and annealing rate parameters; Step ②: In each iteration, a set of neighborhood solutions is randomly generated by exchange, insertion and deletion, and the energy difference between the new solution and the current solution is calculated through the energy function. If the energy of the new solution is lower, the new solution is accepted and the current solution information is updated. Otherwise, the acceptance probability of the worse solution is calculated by the Metropolis criterion, and a random number r is generated. If r≤P, the worse solution is accepted, otherwise the worse solution is rejected. Step ③: After each iteration, the temperature is gradually reduced according to the annealing rate parameter. When the current temperature is reduced to the preset termination temperature, the iteration is stopped, and the storage layout corresponding to the optimal solution is output as the optimized block digital data storage layout; The specific steps for calculating the total number of outflow values ​​described in S4 are as follows: Step I: The server center performs effective automatic statistics and calculations at the preset operation cycle and time end point, summing up the total number of Group A 1 data in the accumulated blocks and the total number of Group A 2 data added to the server's Area B database; Step II: According to the sum of the numbers in the block B area of ​​the same number identified by the server center and the needs of the block numbers, set the golden point ratio of the database in the storage area in the server. After periodic statistics, the remainder of the numbers that are effectively calculated after receiving the valid command signal in the server center is used as the total number of outflow values ​​in the in-out mechanism, and it is summarized as the sum of the K area data in the numerical area flowing out of the server. The remaining digital data in the storage area will be used as the digital value of the in-out mechanism of the next server center cloud computing; Step III: When the server center identifies the data of group B blocks, the data of group B is summarized into the total of the H zone database in the server. At the same time, the sum of the data in zone H is proportional to the amount of data in zone B, and the sum of the numbers in zone H must be greater than the sum of the numbers in zone B. Then, the block number entry and exit mechanism is executed in the server, and the data in zones B and H are updated synchronously according to the data identified by the server center function.

2. A server-based distributed block digital diversion method according to claim 1, characterized in that: The specific calculation formula of the outflow value in step II is as follows: In the formula, K n Represents the outflow value of K zone; G 甲1 Represents the cumulative data of Group A 1 in Zone B; N 甲2 Group A 2 represents the newly added data of Zone B; K o represents the outflow value of the previous K zone; R% represents the golden point ratio; G H Represents the total data of Group B in Zone H.

Images