Multi-party collaborative enterprise data AI intelligent analysis storage method

Abstract

The present application relates to the technical field of data analysis storage, and more particularly to a multi-party cooperative enterprise data AI intelligent analysis storage method, which fills the "information blind area" of the single factor method through the double-dimensional attribute of data; adopts multi-modal feature combined with graph attention network to depict indirect value, explicitly quantifies the implicit value of "being cited / cocurrent by high-value objects", and realizes "cooperative value amplification"; the calculation of data update influence degree converts "high priority x high update" into a substantial impact index of the system, the calculation of correlation-load coupling degree supports the matching of high-correlation data to low-load nodes, and reduces access delay; the calculation of data value degree adopts a three-layer weight structure, which is more consistent with the real business value distribution; by constructing a storage node adaptation feature matrix and calculating a node adaptation degree score vector, permission-load double factor mapping is realized; the actual utilization efficiency of data is introduced as a feature, and "high value but not utilized" becomes an observable object.

Description

Technical Field

[0001] This invention relates to the field of data analysis and storage technology, and in particular to a multi-party collaborative enterprise data AI intelligent analysis and storage method. Background Technology

[0002] In enterprise-level data management, the value of the same data varies in different time windows and collaboration scenarios. Moreover, the value is affected by multiple factors such as business priority, update frequency, collaborative contribution, and time decay, making value assessment difficult. The storage media (SSD / HDD / tape) and the real-time load (capacity / concurrency) of nodes change dynamically, making static tiering or fixed strategies difficult to be effective in the long term, and resource matching difficult. Furthermore, high privacy data and high-performance access often conflict, requiring a fine-grained trade-off between "accessibility (permissions)" and "load tolerance (load)," resulting in a contradiction between compliance and efficiency.

[0003] Existing approaches include static hot / cold stratification, single-dimensional scoring, role-based coarse-grained access control (RBAC), and simple migration strategies based on capacity thresholds. These approaches share common characteristics: few elements, weak coupling, and insufficient real-time performance, making them unsuitable for supporting complex enterprise-level scenarios involving cross-departmental collaboration, federated compliance, and fine-grained scheduling. Summary of the Invention

[0004] The main objective of this invention is to provide a multi-party collaborative enterprise data AI intelligent analysis and storage method. It supplements the "information blind spots" of single-factor methods by using dual-dimensional attributes of data; it employs multimodal features combined with graph attention networks to characterize indirect value, explicitly quantifying the implicit value of "being referenced / co-occurring with high-value objects," thus achieving "collaborative value amplification"; the calculation of data update impact transforms "high priority × high update" into a substantial impact indicator on the system, and the calculation of correlation-load coupling supports adapting highly correlated data to low-load nodes, reducing access latency; the calculation of data value adopts a three-layer weight structure, aggregating "business priority," "update × coupling," and "collaboration × timeliness" in layers, more closely aligning with the distribution of real business value; by constructing a storage node attribute matrix, user permission vector, permission-storage adaptation vector, and performance-storage adaptation vector, it further constructs a storage node adaptation feature matrix and calculates a node adaptation score vector, realizing permission-load dual-factor mapping; and it introduces the feature of actual data utilization efficiency, making "high-value but unutilized" data observable.

[0005] The technical solution of the present invention is as follows:

[0006] Firstly, a multi-party collaborative enterprise data AI intelligent analysis and storage method is proposed, which includes the following steps:

[0007] S1. Collect heterogeneous data streams in the collaborative process of each collaborative enterprise and construct two-dimensional attribute vectors under different time windows. The heterogeneous data streams include structured data and unstructured data, and the two-dimensional attribute vectors include basic attribute vectors and collaborative attribute vectors.

[0008] S2. For heterogeneous data collected within the same time window, extract the basic feature vectors of different business categories of data. Construct a multi-party data association graph with a single business category of data as a node and the sharing frequency between different business categories of data as the edge weight. Use a graph attention network to calculate the association importance of nodes of different categories of data and output the association importance vector. Then, fuse the basic feature vector and the association importance vector.

[0009] S3. For data of different business categories, calculate the impact of data updates, obtain the comprehensive load rate of storage nodes to be allocated to data of different business categories, calculate the correlation-load coupling degree of data of different business categories, and further calculate the data value of data of different business categories and classify the value attributes.

[0010] S4. Obtain the storage node attribute matrix, user permission vector, permission-storage adaptation vector, and performance-storage adaptation vector. Further construct the storage node adaptation feature matrix, calculate the node adaptation score vector, select the storage node with the highest adaptation score in the node adaptation score vector as the target storage node, and output the target storage node ID, level, and overall load rate.

[0011] S5. Calculate the actual data utilization efficiency of different business categories of data within the same time window, and issue instructions based on threshold evaluation.

[0012] A further improvement of this invention lies in that the specific content of S1 is: collecting heterogeneous data streams from the collaborative process of each collaborative enterprise, specifically including structured and unstructured data, and constructing a two-dimensional attribute vector, wherein the two-dimensional attribute vector is represented as... ;in, For the d-th class of data at time window t; The basic attribute vector of the d-th class of data. , The privacy sensitivity of the d-th data type is represented by a value between 0 and 1. The business priority for data of type d is 0-1. The value represents the freshness of the d-th data type, ranging from 0 to 1. Let d be the collaborative attribute vector of the d-th class of data. , The collaborative contribution of the d-th data type is represented by a value between 0 and 1. This represents the number of other data categories associated with the d-th data category, and its value is a positive integer, where d is the business category label to which the data belongs.

[0013] A further improvement of the present invention is that step S2 includes the following specific steps:

[0014] S21. For heterogeneous data collected at time window t, for structured data, the XGBoost model is used to extract the field importance weight matrix; for unstructured text data, the BERT model is used to extract semantic feature vectors; for unstructured image and video data, the ResNet model is used to extract visual feature vectors. The field importance weight matrix, semantic feature vectors, and visual feature vectors are then normalized to a [0,1] dimension to form the basic feature matrix. ,in, Let be the basic feature vector of the d-th class of data at time window t. The dimension is d×3;

[0015] S22. Using single-category data as nodes and the sharing frequency between different categories of data as edge weights, construct a multi-party data association graph. Calculate the association importance of each data category node using a graph attention network. The formula for calculating the association importance is:

[0016] ;

[0017] in, The association importance of the d-th data at time window t. Let be the edge weight between the d-th data type and the i-th associated data type. Output the historical value of the i-th type of associated data; output the association importance matrix. ,in, Let d be the association importance vector of the data in class d. , The dimension is d×3;

[0018] S23, the basic feature matrix With the importance matrix Perform weighted fusion and output the final feature matrix. The final feature matrix The calculation formula is:

[0019] ;

[0020] in, For fusion weighting coefficients.

[0021] A further improvement of the present invention is that step S3 includes the following specific steps:

[0022] S31. For data of different business categories, calculate the impact of data updates. The formula for calculating the impact of data updates is as follows:

[0023] ;

[0024] in, For the data update impact of the d-th type of data, Let d be the number of times the data is updated per unit time. The business impact coefficient is 1.5 when the d-th data is core business data and 0.8 when the d-th data is not core business data.

[0025] S32. Obtain the overall load rate of the storage nodes to be allocated for different business categories of data. The formula for calculating the overall load rate is as follows:

[0026] ;

[0027] in, The overall load rate of the storage nodes to be allocated for the d-th type of data. This represents the used capacity of the storage node. This represents the maximum capacity of the storage node. This represents the number of real-time access requests to the storage node. This represents the maximum number of access requests that a storage node can support. This is the request-capacity conversion factor, with a value of 0.01.

[0028] S33. Calculate the correlation-load coupling degree of the d-th type of data. The formula for calculating the correlation-load coupling degree is as follows:

[0029] ;

[0030] in, For the correlation-load coupling degree of the d-th type of data, The inherent correlation between the d-th type of data and the core business is represented by a value of 0-1.

[0031] A further improvement of the present invention is that S3 further includes:

[0032] S34. Calculate the data value of data for different business categories. The formula for calculating the data value is as follows:

[0033]

[0034] in, The data value of the d-th type of data at time window t. Let d be the most recent update time of the data in class d. The validity period of the d-th type of data;

[0035] S35. Divide data into value attributes for different business categories, when... When, the data of category d is assigned to the archived value pool, when When, the data of category d is assigned to the regular value pool, when At that time, the data of category d will be assigned to the core value pool.

[0036] A further improvement of the present invention is that step S4 includes the following specific steps:

[0037] S41. Obtain the storage node attribute matrix The dimension is 3×n, where n is the number of storage nodes, and the three columns correspond to the storage node ID, level, and overall load rate of the storage node, respectively.

[0038] S42. Based on the user's permission type, which includes financial permissions, operational permissions, and administrative permissions in sequence, construct a user permission vector. When the user's permission type is financial, ， When the user's permission type is "operational" When the user's permission type is administrative, ;

[0039] S43. The permission-storage adaptation vector is constructed as follows: Where n is the number of storage nodes, when the user permissions of the nth storage node match, When the user permissions of the nth storage node do not match, ;

[0040] S44. Construct the load-storage adaptation vector as follows: When the overall load rate of the nth storage node is not greater than the overall load rate threshold, When the overall load rate of the nth storage node is greater than the overall load rate threshold, .

[0041] S45. The performance-storage adaptation vector is constructed as follows: When the nth storage node is an SSD node, When the nth storage node is an HDD node, ,otherwise, ;

[0042] S46. The output storage node adaptation feature matrix is: .

[0043] A further improvement of the present invention is that S4 further includes:

[0044] S47. Calculate the node fitness score vector, wherein the calculation formula for the node fitness score vector is:

[0045] ;

[0046] in, Let n be the node fit score vector, with dimension n×1. This is the feature weight vector for storing nodes; ; This is the permission weighting factor, with a value of 0.2. This is the load weighting factor, with a value of 0.5. This is the performance weighting factor, with a value of 0.3.

[0047] S48. Select the storage node with the highest fit score in the node fit score vector as the target storage node, and output the target storage node ID, level and overall load rate.

[0048] A further improvement of the present invention is that step S5 includes the following specific steps:

[0049] S51. Using [t-N+1,t] as the time window interval, obtain the data value of different business categories within N consecutive time windows, and calculate the actual data utilization efficiency of different business categories at time window t. The calculation formula is as follows:

[0050] ;

[0051] in, The actual utilization efficiency of the d-th type of data at time window t. It is a binary function, which takes the value 1 when the d-th type of data is accessed at time window k, and takes the value 0 otherwise;

[0052] S52, when the d-th type of data is core value pool data and When the utilization efficiency is less than the actual utilization efficiency threshold, a warning is issued that high-value data is not being utilized; when the overall load rate of the target storage node for the d-th type of data is greater than the overall load rate threshold, a node switching instruction is issued; when the d-th type of data is archived value pool data and the collaborative contribution is less than the collaborative contribution threshold, a data archiving instruction is issued.

[0053] Secondly, a computer-readable storage medium is proposed, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned multi-party collaborative enterprise data AI intelligent analysis and storage method is implemented.

[0054] Thirdly, an electronic device is proposed, including a memory for storing instructions and a processor for executing the instructions, causing the device to perform the aforementioned multi-party collaborative enterprise data AI intelligent analysis and storage method.

[0055] The technical effects of this invention are as follows:

[0056] A multi-party collaborative enterprise data AI intelligent analysis and storage method was constructed, which fills the "information blind spot" of single-factor methods by using the dual-dimensional attributes of data; it adopts multimodal features combined with graph attention networks to characterize indirect value, and explicitly quantifies the implicit value of "being referenced / co-occurring by high-value objects", thus realizing "collaborative value amplification"; the calculation of data update impact transforms "high priority × high update" into a substantial impact indicator on the system, and the calculation of correlation-load coupling degree supports the adaptation of highly correlated data to low-load nodes, reducing access latency; the calculation of data value degree adopts a three-layer weight structure, which will aggregate "business priority", "update × coupling", and "collaboration × timeliness" in layers, which is more in line with the actual business value distribution; by constructing a storage node attribute matrix, user permission vector, permission-storage adaptation vector, and performance-storage adaptation vector, a storage node adaptation feature matrix is ​​further constructed, and a node adaptation degree score vector is calculated, realizing permission-load dual-factor mapping; the feature of actual data utilization efficiency is introduced, making "high value but unutilized" observable objects. Attached Figure Description

[0057] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0058] Figure 1 This is a flowchart illustrating the multi-party collaborative enterprise data AI intelligent analysis and storage method of Embodiment 1 of the present invention. Detailed Implementation

[0059] Example 1: This example proposes a multi-party collaborative enterprise data AI intelligent analysis and storage method. It supplements the "information blind spot" of single-factor methods by using dual-dimensional attributes of data; it employs multimodal features combined with graph attention networks to characterize indirect value, explicitly quantifying the implicit value of "being referenced / co-occurring with high-value objects," thus achieving "collaborative value amplification"; the calculation of data update impact transforms "high priority × high update" into a substantial impact indicator on the system, and the calculation of correlation-load coupling supports adapting highly correlated data to low-load nodes, reducing access latency; the calculation of data value adopts a three-layer weight structure, aggregating "business priority," "update × coupling," and "collaboration × timeliness" in layers, better reflecting the distribution of real business value; by constructing a storage node attribute matrix, user permission vector, permission-storage adaptation vector, and performance-storage adaptation vector, it further constructs a storage node adaptation feature matrix and calculates a node adaptation score vector, realizing permission-load dual-factor mapping; it introduces the feature of actual data utilization efficiency, making "high-value but unutilized" observable objects. Specifically, such as... Figure 1As shown, the multi-party collaborative enterprise data AI intelligent analysis and storage method proposed in this embodiment includes the following specific steps:

[0060] S1. Collect heterogeneous data streams in the collaborative process of each collaborative enterprise and construct two-dimensional attribute vectors under different time windows. The heterogeneous data streams include structured data and unstructured data, and the two-dimensional attribute vectors include basic attribute vectors and collaborative attribute vectors.

[0061] In this embodiment, S1 specifically involves: collecting heterogeneous data streams from each collaborative enterprise's collaborative process, specifically including structured and unstructured data, and constructing a two-dimensional attribute vector, wherein the two-dimensional attribute vector is represented as... ;in, For the d-th class of data at time window t; The basic attribute vector of the d-th class of data. , The privacy sensitivity of the d-th data type is represented by a value between 0 and 1. The business priority for data of type d is 0-1. The value represents the freshness of the d-th data type, ranging from 0 to 1. Let d be the collaborative attribute vector of the d-th class of data. , The collaborative contribution of the d-th data type is represented by a value between 0 and 1. This represents the number of other data categories associated with the d-th data category, and its value is a positive integer, where d is the business category label to which the data belongs.

[0062] S2. For heterogeneous data collected within the same time window, extract the basic feature vectors of different business categories of data. Construct a multi-party data association graph with a single business category of data as a node and the sharing frequency between different business categories of data as the edge weight. Use a graph attention network to calculate the association importance of nodes of different categories of data and output the association importance vector. Then, fuse the basic feature vector and the association importance vector.

[0063] In this embodiment, S2 includes the following specific steps:

[0064] S21. For heterogeneous data collected at time window t, for structured data, the XGBoost model is used to extract the field importance weight matrix; for unstructured text data, the BERT model is used to extract semantic feature vectors; for unstructured image and video data, the ResNet model is used to extract visual feature vectors. The field importance weight matrix, semantic feature vectors, and visual feature vectors are then normalized to a [0,1] dimension to form the basic feature matrix. ,in, Let be the basic feature vector of the d-th class of data at time window t. The dimension is d×3;

[0065] S22. Using single-category data as nodes and the sharing frequency between different categories of data as edge weights, construct a multi-party data association graph. Calculate the association importance of each data category node using a graph attention network. The formula for calculating the association importance is:

[0066] ;

[0067] in, The association importance of the d-th data at time window t. Let be the edge weight between the d-th data type and the i-th associated data type. Output the historical value of the i-th type of associated data; output the association importance matrix. ,in, Let d be the association importance vector of the data in class d. , The dimension is d×3;

[0068] S23, the basic feature matrix With the importance matrix Perform weighted fusion and output the final feature matrix. The final feature matrix The calculation formula is:

[0069] ;

[0070] in, For fusion weighting coefficients.

[0071] S3. For data of different business categories, calculate the impact of data updates, obtain the comprehensive load rate of storage nodes to be allocated to data of different business categories, calculate the correlation-load coupling degree of data of different business categories, and further calculate the data value of data of different business categories and classify the value attributes.

[0072] In this embodiment, S3 includes the following specific steps:

[0073] S31. For data of different business categories, calculate the impact of data updates. The formula for calculating the impact of data updates is as follows:

[0074] ;

[0075] in, For the data update impact of the d-th type of data, Let d be the number of times the data is updated per unit time. The business impact coefficient is 1.5 when the d-th data is core business data and 0.8 when the d-th data is not core business data.

[0076] S32. Obtain the overall load rate of the storage nodes to be allocated for different business categories of data. The formula for calculating the overall load rate is as follows:

[0077] ;

[0078] in, The overall load rate of the storage nodes to be allocated for the d-th type of data. This represents the used capacity of the storage node. This represents the maximum capacity of the storage node. This represents the number of real-time access requests to the storage node. This represents the maximum number of access requests that a storage node can support. This is the request-capacity conversion factor, with a value of 0.01.

[0079] S33. Calculate the correlation-load coupling degree of the d-th type of data. The formula for calculating the correlation-load coupling degree is as follows:

[0080] ;

[0081] in, For the correlation-load coupling degree of the d-th type of data, The inherent correlation between the d-th type of data and the core business is represented by a value of 0-1.

[0082] S34. Calculate the data value of data for different business categories. The formula for calculating the data value is as follows:

[0083]

[0084] in, The data value of the d-th type of data at time window t. Let d be the most recent update time of the data in class d. The validity period of the d-th type of data;

[0085] S35. Divide data into value attributes for different business categories, when... When, the data of category d is assigned to the archived value pool, when When, the data of category d is assigned to the regular value pool, when At that time, the data of category d will be assigned to the core value pool.

[0086] S4. Obtain the storage node attribute matrix, user permission vector, permission-storage adaptation vector, and performance-storage adaptation vector. Further construct the storage node adaptation feature matrix, calculate the node adaptation score vector, select the storage node with the highest adaptation score in the node adaptation score vector as the target storage node, and output the target storage node ID, level, and overall load rate.

[0087] In this embodiment, S4 includes the following specific steps:

[0088] S41. Obtain the storage node attribute matrix The dimension is 3×n, where n is the number of storage nodes, and the three columns correspond to the storage node ID, level, and overall load rate of the storage node, respectively.

[0089] S42. Based on the user's permission type, which includes financial permissions, operational permissions, and administrative permissions in sequence, construct a user permission vector. When the user's permission type is financial, ， When the user's permission type is "operational" When the user's permission type is administrative, ;

[0090] S43. The permission-storage adaptation vector is constructed as follows: Where n is the number of storage nodes, when the user permissions of the nth storage node match, When the user permissions of the nth storage node do not match, ;

[0091] S44. Construct the load-storage adaptation vector as follows: When the overall load rate of the nth storage node is not greater than the overall load rate threshold, When the overall load rate of the nth storage node is greater than the overall load rate threshold, .

[0092] S45. The performance-storage adaptation vector is constructed as follows: When the nth storage node is an SSD node, When the nth storage node is an HDD node, ,otherwise, ;

[0093] S46. The output storage node adaptation feature matrix is: .

[0094] A further improvement of the present invention is that S4 further includes:

[0095] S47. Calculate the node fitness score vector, wherein the calculation formula for the node fitness score vector is:

[0096] ;

[0097] in, Let n be the node fit score vector, with dimension n×1. This is the feature weight vector for storing nodes; ; This is the permission weighting factor, with a value of 0.2. This is the load weighting factor, with a value of 0.5. This is the performance weighting factor, with a value of 0.3.

[0098] S48. Select the storage node with the highest fit score in the node fit score vector as the target storage node, and output the target storage node ID, level and overall load rate.

[0099] S5. Calculate the actual data utilization efficiency of different business categories of data within the same time window, and issue instructions based on threshold evaluation.

[0100] In this embodiment, S5 includes the following specific steps:

[0101] S51. Using [t-N+1,t] as the time window interval, obtain the data value of different business categories within N consecutive time windows, and calculate the actual data utilization efficiency of different business categories at time window t. The calculation formula is as follows:

[0102] ;

[0103] in, The actual utilization efficiency of the d-th type of data at time window t. It is a binary function, which takes the value 1 when the d-th type of data is accessed at time window k, and takes the value 0 otherwise;

[0104] S52, when the d-th type of data is core value pool data and When the utilization efficiency is less than the actual utilization efficiency threshold, a warning is issued that high-value data is not being utilized; when the overall load rate of the target storage node for the d-th type of data is greater than the overall load rate threshold, a node switching instruction is issued; when the d-th type of data is archived value pool data and the collaborative contribution is less than the collaborative contribution threshold, a data archiving instruction is issued.

[0105] The threshold and weight settings can be based on the default settings of this invention, or they can be set by the operator.

[0106] Example 2: This example provides an electronic device, including a processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the above-mentioned multi-party collaborative enterprise data AI intelligent analysis and storage method by calling the computer program stored in the memory.

[0107] The electronic device can vary considerably depending on its configuration or performance. It may include one or more Central Processing Units (CPUs) and one or more memories, wherein the memory stores at least one computer program, which is loaded and executed by the processor to implement the multi-party collaborative enterprise data AI intelligent analysis and storage method provided in the above-described embodiment. The electronic device may also include other components for implementing its functions; for example, it may have wired or wireless network interfaces and input / output interfaces for data input and output. Details will not be elaborated upon in this embodiment.

[0108] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this disclosure can be embodied in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0109] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0110] This invention is described with reference to flowchart illustrations and block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and block diagrams, as well as combinations of blocks in the flowchart illustrations and block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0111] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and boxes Figure 1 The steps of the function specified in one or more boxes.

[0112] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A multi-party collaborative enterprise data AI intelligent analysis and storage method, characterized by: The specific steps include the following: S1. Collect heterogeneous data streams in the collaborative process of each collaborative enterprise and construct two-dimensional attribute vectors under different time windows. The heterogeneous data streams include structured data and unstructured data, and the two-dimensional attribute vectors include basic attribute vectors and collaborative attribute vectors. S2. For heterogeneous data collected within the same time window, extract the basic feature vectors of different business categories of data. Construct a multi-party data association graph with a single business category of data as a node and the sharing frequency between different business categories of data as the edge weight. Use a graph attention network to calculate the association importance of nodes of different categories of data and output the association importance vector. Then, fuse the basic feature vector and the association importance vector. S3. For data of different business categories, calculate the impact of data updates, obtain the comprehensive load rate of storage nodes to be allocated to data of different business categories, calculate the correlation-load coupling degree of data of different business categories, and further calculate the data value of data of different business categories and classify the value attributes. S3 includes the following specific steps: S31. For data of different business categories, calculate the impact of data updates. The formula for calculating the impact of data updates is as follows: ； in, For the data update impact of the d-th type of data, Let d be the number of times the data is updated per unit time. The business impact coefficient is 1.5 when the d-th data is core business data and 0.8 when the d-th data is not core business data. S32. Obtain the overall load rate of the storage nodes to be allocated for different business categories of data. The formula for calculating the overall load rate is as follows: ； in, The overall load rate of the storage nodes to be allocated for the d-th type of data. This represents the used capacity of the storage node. This represents the maximum capacity of the storage node. This represents the number of real-time access requests to the storage node. This represents the maximum number of access requests that a storage node can support. This is the request-capacity conversion factor, with a value of 0.

01. S33. Calculate the correlation-load coupling degree of the d-th type of data. The formula for calculating the correlation-load coupling degree is as follows: ； in, For the correlation-load coupling degree of the d-th type of data, The inherent correlation between the d-th type of data and the core business is represented by a value of 0-1. S34. Calculate the data value of data for different business categories. The formula for calculating the data value is as follows: ； in, The data value of the d-th type of data at time window t. Let d be the most recent update time of the data in class d. The validity period of the d-th type of data; S35. Divide data into value attributes for different business categories, when... When, the data of category d is assigned to the archived value pool, when When, the data of category d is assigned to the regular value pool, when At that time, the data of category d will be assigned to the core value pool; S4. Obtain the storage node attribute matrix, user permission vector, permission-storage adaptation vector, and performance-storage adaptation vector. Further construct the storage node adaptation feature matrix, calculate the node adaptation score vector, select the storage node with the highest adaptation score in the node adaptation score vector as the target storage node, and output the target storage node ID, level, and overall load rate. S4 includes the following specific steps: S41. Obtain the storage node attribute matrix The dimension is 3×n, where n is the number of storage nodes, and the three columns correspond to the storage node ID, level, and overall load rate of the storage node, respectively. S42. Based on the user's permission type, which includes financial permissions, operational permissions, and administrative permissions in sequence, construct a user permission vector. When the user's permission type is financial, ， When the user's permission type is "operational" When the user's permission type is administrative, ; S43. The permission-storage adaptation vector is constructed as follows: Where n is the number of storage nodes, when the user permissions of the nth storage node match, When the user permissions of the nth storage node do not match, ; S44. Construct the load-storage adaptation vector as follows: When the overall load rate of the nth storage node is not greater than the overall load rate threshold, When the overall load rate of the nth storage node is greater than the overall load rate threshold, ; S45. The performance-storage adaptation vector is constructed as follows: When the nth storage node is an SSD node, When the nth storage node is an HDD node, ,otherwise, ; S46. The output storage node adaptation feature matrix is: ; S47. Calculate the node fitness score vector, wherein the calculation formula for the node fitness score vector is: ； in, Let n be the node fit score vector, with dimension n×1. This is the feature weight vector for storing nodes; ; This is the permission weighting factor, with a value of 0.

2. This is the load weighting factor, with a value of 0.

5. This is a performance weighting factor with a value of 0.

3. S48. Select the storage node with the highest adaptation score in the node adaptation score vector as the target storage node, and output the target storage node ID, level and overall load rate. S5. Calculate the actual data utilization efficiency of different business categories of data within the same time window, and issue instructions based on the threshold evaluation.

2. The multi-party collaborative enterprise data AI intelligent analysis and storage method according to claim 1, characterized in that: The specific content of S1 is as follows: Collect heterogeneous data streams from the collaborative process of each collaborating enterprise, specifically including structured and unstructured data, and construct a two-dimensional attribute vector, wherein the two-dimensional attribute vector is represented as... ;in, For the d-th class of data at time window t; For the basic attribute vector of the d-th class of data, , The privacy sensitivity of the d-th data type is represented by a value between 0 and 1. The business priority for data of type d is 0-1. The value represents the freshness of the d-th data type, ranging from 0 to 1. Let d be the collaborative attribute vector of the d-th class of data. , The collaborative contribution of the d-th data type is represented by a value between 0 and 1. This represents the number of other data categories associated with the d-th data category, and its value is a positive integer, where d is the business category label to which the data belongs.

3. The multi-party collaborative enterprise data AI intelligent analysis and storage method according to claim 2, characterized in that: S2 includes the following specific steps: S21. For heterogeneous data collected at time window t, for structured data, the XGBoost model is used to extract the field importance weight matrix; for unstructured text data, the BERT model is used to extract semantic feature vectors; for unstructured image and video data, the ResNet model is used to extract visual feature vectors. The field importance weight matrix, semantic feature vectors, and visual feature vectors are then normalized to a [0,1] dimension to form the basic feature matrix. ,in, Let be the basic feature vector of the d-th class of data at time window t. The dimension is d×3; S22. Using single-category data as nodes and the sharing frequency between different categories of data as edge weights, construct a multi-party data association graph. Calculate the association importance of each data category node using a graph attention network. The formula for calculating the association importance is: ； in, The association importance of the d-th data at time window t. Let be the edge weight between the d-th data type and the i-th associated data type. Output the historical value of the i-th type of associated data; output the association importance matrix. ,in, Let d be the association importance vector of the data in class d. , The dimension is d×3; S23, The basic feature matrix Association importance matrix Perform weighted fusion and output the final feature matrix. The final feature matrix The calculation formula is: ； in, For fusion weighting coefficients.

4. The multi-party collaborative enterprise data AI intelligent analysis and storage method according to claim 3, characterized in that: S5 includes the following specific steps: S51. Using [t-N+1,t] as the time window interval, obtain the data value of different business categories within N consecutive time windows, and calculate the actual data utilization efficiency of different business categories at time window t. The calculation formula is as follows: ； in, The actual utilization efficiency of the d-th type of data at time window t. It is a binary function, which takes the value 1 when the d-th type of data is accessed at time window k, and takes the value 0 otherwise; S52, when the d-th type of data is core value pool data and When the utilization efficiency is less than the actual utilization efficiency threshold, a warning is issued that high-value data is not being utilized; when the overall load rate of the target storage node for the d-th type of data is greater than the overall load rate threshold, a node switching instruction is issued; when the d-th type of data is archived value pool data and the collaborative contribution is less than the collaborative contribution threshold, a data archiving instruction is issued.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multi-party collaborative enterprise data AI intelligent analysis and storage method as described in any one of claims 1-4.

6. An electronic device, characterized in that, It includes a memory for storing instructions; and a processor for executing the instructions, causing the device to perform the multi-party collaborative enterprise data AI intelligent analysis and storage method as described in any one of claims 1 to 4.

Images