Enterprise-level knowledge graph construction method for heterogeneous data fusion

Abstract

The present application relates to the technical field of data processing, and more particularly to a kind of enterprise-level knowledge graph construction method for heterogeneous data fusion, by constructing two-dimensional state space criterion, data freshness and write lock competition state are included in unified decision system, the joint diagnosis of cross-module running state is realized.Aiming at three different abnormal scenes of state alignment deviation too large, resource competition intensifies, control target conflict, respectively output different dynamic regulation and control strategy, can simultaneously consider the timeliness of graph data and the stability of back-end write.The normalized calculation is used for freshness decay ratio, which can adapt to the delay requirements of different business scenarios, and has strong versatility;Write lock competition index fusion processing rate, topological influence range and database lock characteristics three dimensions can more accurately predict lock conflict risk and trigger regulation and control action in advance.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an enterprise-level knowledge graph construction method for heterogeneous data fusion. Background Technology

[0002] As enterprises deepen their digital transformation, knowledge graphs have become the foundational data carrier for core scenarios such as supply chain management and business risk control. Enterprise-level knowledge graphs utilize diverse heterogeneous data sources, including business systems, log files, and third-party interfaces. The construction process typically involves two core links: front-end data extraction and parsing, and back-end topology fusion; and persistent graph writing. Distributed cluster architectures are commonly used for deployment. Currently, enterprise businesses have increasingly stringent requirements for the timeliness of graph data, while the write performance of graph databases is significantly constrained by locking mechanisms. Existing architectures often employ an independent scheduling model at both ends, relying on queue length thresholds for simple backpressure control. This makes it difficult to dynamically adapt between data injection speed and write processing capacity. Maintaining back-end write stability while ensuring data freshness has become a common technical challenge that enterprise-level knowledge graph construction systems urgently need to address.

[0003] In existing enterprise-level knowledge graph construction systems, the front-end extraction and back-end writing links employ independent scheduling logics, lacking a cross-module joint awareness mechanism for operational status. The system triggers a single anti-pressure strategy based solely on queue backlog length, failing to distinguish between two different types of operational anomalies: data freshness exceeding limits and write lock resource contention. This results in a lack of targeted control measures. When business data volume fluctuates, either back-end processing delays lead to excessively long data dwell times, causing graph data freshness to decline and hindering real-time business decisions; or a sudden surge in write pressure triggers a surge in graph database lock conflicts, causing a non-linear decrease in write throughput and exacerbating data backlog. In extreme scenarios where both problems occur simultaneously, a single control method becomes ineffective, failing to simultaneously alleviate the conflict between freshness and performance. This results in insufficient overall system robustness and hinders optimal hardware resource utilization. Summary of the Invention

[0004] The main objective of this invention is to provide an enterprise-level knowledge graph construction method for heterogeneous data fusion. This solution constructs a two-dimensional state space criterion, incorporating data freshness and write lock contention status into a unified judgment system to achieve joint diagnosis of cross-module operating status. For three different abnormal scenarios—excessive state alignment deviation, intensified resource contention, and complete conflict of control objectives—differentiated dynamic control strategies are output, simultaneously considering the timeliness of graph data and the stability of backend writes. The freshness decay ratio is calculated using normalization, adapting to the latency requirements of different business scenarios and possessing strong versatility. The write lock contention index integrates three dimensions: processing rate, topology influence range, and database lock characteristics, enabling more accurate prediction of lock conflict risks and early triggering of control actions. The system can automatically adjust frontend and backend operating parameters based on real-time operating status, dynamically adapting to data traffic fluctuations, effectively improving system robustness and resource utilization efficiency, and ensuring the quality of knowledge graph construction and service continuity.

[0005] The technical solution of the present invention is as follows: Firstly, a method for constructing an enterprise-level knowledge graph for heterogeneous data fusion is proposed, which includes the following steps: S1. Obtain the underlying parameter set for the current period to provide basic data input; the underlying parameter set includes the time consumption of single batch heterogeneous data extraction, absolute timestamp of candidate subgraph generation, current absolute timestamp, time consumption of single global topology fusion and writing, average degree of the graph nodes to be updated, maximum data delay time, and graph database write lock sensitivity coefficient. S2. Perform a difference operation based on the current absolute timestamp and the absolute timestamp generated by the candidate subgraph to output the data residence time; S3. Based on the data residence time and the maximum data delay time, determine and obtain the freshness decay ratio; at the same time, based on the single global topology fusion and write time, the single batch heterogeneous data extraction time, the average degree of the graph nodes to be updated, and the graph database write lock sensitivity coefficient, calculate and obtain the write lock contention index. S4. Combine the freshness decay ratio with the write lock contention index into two-dimensional coordinate points to generate a cross-module freshness and write lock contention state space criterion. S5. Perform a logical comparison of the cross-module freshness and write lock contention state space criterion, determine its relationship with the preset state space boundary, and output the connection conflict diagnosis result. S6. When the connection conflict diagnosis result indicates that the state alignment deviation is too large, the resource competition is intensified, or the control target is completely conflicted, the control command correction amount is calculated based on the connection conflict diagnosis result, and a set of control correction commands is generated and issued to dynamically adjust the operating parameters of the front-end extraction computing power allocation scheduler and the back-end graph update executor.

[0006] A further improvement of the present invention is that the specific content of S1 is as follows: by collecting operation logs through probes deployed on the front-end extraction server cluster, and by combining the Coordinated Universal Time synchronized by the Network Time Protocol, the absolute timestamp of the candidate subgraph generation and the current absolute timestamp are obtained; the transaction execution logs are obtained through the built-in transaction log interface of the back-end graph database, and the running status of the back-end graph database is obtained through the query interface; the global configuration file of the supply chain business is obtained through the configuration center; and by combining the above information, the underlying parameter set of the current period is output.

[0007] A further improvement of the present invention is that the specific content of S2 is: based on the current absolute timestamp Generate absolute timestamps with the candidate subgraphs Perform a difference operation and output the data residence time after the time base alignment is completed. .

[0008] A further improvement of the present invention is that step S3 includes the following specific steps: S31. Divide the data dwell time by the maximum data delay time. The freshness decay ratio was obtained. The calculation formula is: ; S32, The time consumed by the single global topology fusion and writing is calculated. Divide by the time consumed in extracting heterogeneous data in a single batch and the average degree of the nodes in the graph to be updated. and the graph database write lock sensitivity coefficient Multiply by each other to obtain the write lock contention index. The calculation formula is: .

[0009] A further improvement of the present invention is that S4 includes: based on the freshness decay ratio Contention index with the write lock Combined into a two-dimensional coordinate point; generate and output the cross-module freshness and write lock contention state space criterion. .

[0010] A further improvement of the present invention is that S5 includes: a state space criterion based on the cross-module freshness and write lock contention. It is logically compared with the preset state space boundary; whereby the freshness warning boundary is set as... Set the write lock security threshold to ;like and The output connection conflict diagnosis result indicates that the state alignment deviation is too large; if and The output connection conflict diagnosis result indicates that resource contention has intensified; if and The output connection conflict diagnosis result indicates that the control target is in complete conflict.

[0011] A further improvement of this invention is that the operating parameters include the batch size of the backend graph database fetch, the batch size of the write operation, the transaction lock granularity, the topology fusion depth, and the number of concurrent threads in the frontend extraction server cluster; step S6 includes: generating an instruction to increase the batch size of the backend graph database fetch in response to the excessive state alignment deviation, wherein the adjustment range of the batch size is related to... The difference is proportional; in response to the increased resource competition, instructions are generated to reduce the write batch size of the backend graph database and split the transaction lock granularity, and frontend backpressure rate limiting instructions are generated. The reduction in write batch size is proportional to... The difference is proportional; in response to a complete conflict with the control objective, an instruction is generated to reduce the number of concurrent threads in the front-end extraction server cluster, and the reduction in the number of threads is proportional to... and The product is proportional to the value of the product, and instructions are generated to reduce the topology fusion depth of the backend graph database. The reduction in fusion depth is proportional to the time consumed by a single global topology fusion and write operation. The current value is proportional to the output; the output is a set of control correction instructions for the front-end extraction computing power allocation scheduler and the back-end graph update executor.

[0012] Secondly, a computer-readable storage medium is proposed, on which a computer program is stored. When the computer program is executed by a processor, it implements the above-mentioned enterprise-level knowledge graph construction method for heterogeneous data fusion.

[0013] 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 above-described enterprise-level knowledge graph construction method for heterogeneous data fusion.

[0014] The technical effects of this invention are as follows: This paper presents an enterprise-level knowledge graph construction method for heterogeneous data fusion. The solution incorporates data freshness and write lock contention status into a unified judgment system by constructing a two-dimensional state space criterion, enabling joint diagnosis of cross-module operational status. Differentiated dynamic control strategies are output for three different abnormal scenarios: excessive state alignment deviation, intensified resource contention, and complete conflict of control objectives, simultaneously considering the timeliness of graph data and the stability of backend writes. The freshness decay ratio is calculated using normalization, adapting to the latency requirements of different business scenarios and possessing strong versatility. The write lock contention index integrates three dimensions: processing rate, topology influence range, and database lock characteristics, enabling more accurate prediction of lock conflict risks and early triggering of control actions. The system can automatically adjust frontend and backend operating parameters based on real-time operational status, dynamically adapting to data traffic fluctuations, effectively improving system robustness and resource utilization efficiency, and ensuring the quality of knowledge graph construction and service continuity. Attached Figure Description

[0015] 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: Figure 1 This is a flowchart illustrating an enterprise-level knowledge graph construction method for heterogeneous data fusion according to Embodiment 1 of the present invention. Detailed Implementation

[0016] Example 1: This example proposes an enterprise-level knowledge graph construction method for heterogeneous data fusion. This solution constructs a two-dimensional state space criterion, incorporating data freshness and write lock contention status into a unified judgment system to achieve joint diagnosis of cross-module operating status. For three different abnormal scenarios—excessive state alignment deviation, intensified resource contention, and complete conflict of control objectives—differentiated dynamic adjustment strategies are output, simultaneously considering the timeliness of graph data and the stability of backend writes. The freshness decay ratio is calculated using normalization, adapting to the latency requirements of different business scenarios and possessing strong versatility. The write lock contention index integrates three dimensions: processing rate, topology influence range, and database lock characteristics, enabling more accurate prediction of lock conflict risks and early triggering of adjustment actions. The system can automatically adjust frontend and backend operating parameters based on real-time operating status, dynamically adapting to data traffic fluctuations, effectively improving system robustness and resource utilization efficiency, and ensuring the quality of knowledge graph construction and service continuity. Specifically, such as... Figure 1 As shown in the figure, the enterprise-level knowledge graph construction method for heterogeneous data fusion proposed in this embodiment includes the following specific steps: S1. Obtain the underlying parameter set for the current period to provide basic data input; the underlying parameter set includes the time consumption of single batch heterogeneous data extraction, absolute timestamp of candidate subgraph generation, current absolute timestamp, time consumption of single global topology fusion and writing, average degree of the graph nodes to be updated, maximum data delay time, and graph database write lock sensitivity coefficient.

[0017] In this embodiment, the specific content of S1 is as follows: Run logs are collected by probes deployed on the front-end extraction server cluster, and the absolute timestamp of the candidate subgraph and the current absolute timestamp are obtained by combining the Coordinated Universal Time (UTC) synchronized by the Network Time Protocol; transaction execution logs are obtained through the built-in transaction log interface of the back-end graph database, and the running status of the back-end graph database is obtained through the query interface; the global configuration file of the supply chain business is obtained through the configuration center; and the underlying parameter set of the current period is output by combining the above information.

[0018] In this embodiment, at the beginning of each scheduling cycle, the system first obtains the underlying parameter set of the current cycle as the basic data input for subsequent state determination and control adjustment. The underlying parameter set includes the time consumption of a single batch of heterogeneous data extraction, the absolute timestamp of candidate subgraph generation, the current absolute timestamp, the time consumption of a single global topology fusion and write, the average degree of the graph nodes to be updated, the maximum data latency, and the graph database write lock sensitivity coefficient. The absolute timestamp of candidate subgraph generation and the current absolute timestamp are obtained by collecting operation logs from probes deployed on the front-end extraction server cluster and combining them with Coordinated Universal Time (UTC) synchronized by the network time protocol. The absolute timestamp of candidate subgraph generation corresponds to the moment when a single batch of heterogeneous data is extracted, parsed, and candidate graph subgraphs are generated. The current absolute timestamp corresponds to the moment when the state determination operation is performed in the current scheduling cycle. The time consumption of a single global topology fusion and write, and the average degree of the graph nodes to be updated are obtained by obtaining transaction execution logs through the built-in transaction log interface of the back-end graph database and statistically analyzing the running status obtained through the graph database query interface. The write time corresponds to the total time taken for a single batch of candidate subgraphs to complete topological relationship fusion and persistently write to the graph database. The average degree of the graph nodes to be updated corresponds to the average number of connected edges of all existing graph nodes involved in this write operation. The maximum data latency and graph database write lock sensitivity coefficient are obtained by reading the global configuration file of the supply chain business from the configuration center. The maximum data latency corresponds to the longest allowed lag time for graph data under the current business scenario. Graph data exceeding this time cannot meet the timeliness requirements of business decisions. The graph database write lock sensitivity coefficient corresponds to the multiplier characteristic of the probability of lock conflicts in the currently deployed graph database under the target load as the write pressure increases. This step integrates three types of information: front-end running logs, back-end transaction and status data, and global configuration files, to output a complete set of underlying parameters for the current period, providing all inputs for subsequent calculations.

[0019] S2. Perform a difference operation based on the current absolute timestamp and the absolute timestamp generated by the candidate subgraph to output the data residence time.

[0020] In this embodiment, the specific content of S2 is: based on the current absolute timestamp Generate absolute timestamps with the candidate subgraphs Perform a difference operation and output the data residence time after the time base alignment is completed. .

[0021] In this embodiment, a difference operation is performed based on the current absolute timestamp and the absolute timestamp generated by the candidate subgraph, and the data residence time after the time base alignment is completed is output. This is the current absolute timestamp. Generate absolute timestamps for candidate subgraphs. This is the data dwell time, which reflects the length of time a candidate subgraph remains from the time it is generated until the current time before it has been merged and written. The longer the dwell time, the lower the data freshness.

[0022] S3. Based on the data residence time and the maximum data delay time, determine and obtain the freshness decay ratio; at the same time, based on the single global topology fusion and write time, the single batch heterogeneous data extraction time, the average degree of the graph nodes to be updated, and the graph database write lock sensitivity coefficient, calculate and obtain the write lock contention index.

[0023] In this embodiment, step S3 includes the following specific steps: S31. Divide the data dwell time by the maximum data delay time. The freshness decay ratio was obtained. The calculation formula is: ; S32, The time consumed by the single global topology fusion and writing is calculated. Divide by the time consumed in extracting heterogeneous data in a single batch and the average degree of the nodes in the graph to be updated. and the graph database write lock sensitivity coefficient Multiply by each other to obtain the write lock contention index. The calculation formula is: .

[0024] In this embodiment, after obtaining the data residency time, two calculations are performed simultaneously, outputting the freshness decay ratio and the write lock contention index, respectively. The freshness decay ratio is calculated by dividing the data residency time by the maximum data latency time, as shown in the formula. The freshness decay ratio, The maximum data latency is a normalized relative value ranging from 0 to positive infinity. A larger value indicates a more severe decline in data freshness. When the value is greater than or equal to 1, the data has exceeded the maximum allowable latency. In the supply chain scenario of this embodiment, the preferred maximum data latency is 300 seconds. This value stems from the general requirements of enterprise supply chain operations for data timeliness, namely, the lag time for inventory and order data should not exceed 5 minutes; otherwise, it will affect the effectiveness of inventory scheduling and order fulfillment decisions. The write lock contention index is calculated by dividing the time spent on a single global topology fusion and write operation by the time spent extracting heterogeneous data in a single batch, and then multiplying this by the average degree of the graph nodes to be updated and the graph database write lock sensitivity coefficient. The calculation formula is as follows: To write the lock contention index, For the time consumed by a single global topology fusion and write operation, The time required for extracting heterogeneous data in a single batch. The average degree of the nodes in the graph to be updated. This refers to the write lock sensitivity coefficient for the graph database. The write lock contention index comprehensively reflects the relative speed of backend write processing, the topological impact range of write operations, and the lock conflict characteristics of the database itself. The higher the value, the higher the risk of write operations causing lock resource contention. In this embodiment, for mainstream distributed graph databases using optimistic locking mechanisms, the preferred value for the graph database write lock sensitivity coefficient is 0.12. This value is obtained through stress testing of the corresponding version of the graph database under the target hardware configuration, and can accurately reflect the increase in the probability of lock contention when the write load increases.

[0025] S4. Combine the freshness decay ratio and the write lock contention index into two-dimensional coordinate points to generate a cross-module freshness and write lock contention state space criterion.

[0026] In this embodiment, step S4 includes: based on the freshness decay ratio Contention index with the write lock Combined into a two-dimensional coordinate point; generate and output the cross-module freshness and write lock contention state space criterion. .

[0027] In this embodiment, after calculating the two indices, the system combines the freshness decay ratio and the write lock contention index into two-dimensional coordinate points to generate a cross-module freshness and write lock contention state space criterion, which is expressed as follows: This criterion maps two independent operational metrics, originally belonging to the front-end data extraction link and the back-end graph writing link, to the same two-dimensional state plane, realizing a unified representation of cross-module operational status and providing a carrier for subsequent joint state determination.

[0028] S5. Perform a logical comparison of the cross-module freshness and write lock contention state space criteria, determine its relationship with the preset state space boundary, and output the connection conflict diagnosis result.

[0029] In this embodiment, S5 includes: based on the cross-module freshness and write lock contention state space criterion. It is logically compared with the preset state space boundary; whereby the freshness warning boundary is set as... Set the write lock security threshold to ;like and The output connection conflict diagnosis result indicates that the state alignment deviation is too large; if and The output connection conflict diagnosis result indicates that resource contention has intensified; if and The output connection conflict diagnosis result indicates that the control target is in complete conflict.

[0030] In this embodiment, after obtaining the state space criterion, the system logically compares it with the preset state space boundary to determine the interval where the current state point is located and outputs the corresponding connection conflict diagnosis result. The preset state space boundary in this embodiment includes two items, one of which is the freshness warning boundary, with a value of... This value originates from the early warning settings conventions of enterprise-level business systems. When data reaches 80% of the maximum latency, it enters the early warning interval, reserving one-fifth of the buffer space for scheduling adjustments to prevent data from directly exceeding timeliness requirements. The second is the write lock security threshold, denoted as... In this embodiment, the preferred value is 2.5. This value was obtained through stress testing of the target graph database under production-grade hardware configuration. When the write lock contention index exceeds this threshold, the database write throughput will decrease non-linearly, the lock wait time will account for more than 30% of the total write time, and the system write stability cannot be guaranteed. During the comparison process, if the current state satisfies... and The output connection conflict diagnosis result is that the state alignment deviation is too large. This result indicates that the data freshness has entered the warning range, but there is no obvious competition for backend write resources. The problem stems from the mismatch between the frontend data generation speed and the backend fusion write speed, with the backend processing rhythm lagging behind the data injection rhythm. If the current state satisfies and The output connection conflict diagnosis result indicates intensified resource contention. This result suggests that data freshness is within a safe range, but the write lock conflict in the backend graph database has reached a warning level. Continued operation will lead to a continuous deterioration in write performance, which in turn will cause a subsequent decrease in data freshness. If the current state satisfies... and The output connection conflict diagnosis result is a complete conflict of control targets. This result indicates that both freshness and write performance are out of the safe range at the same time. Simply adjusting the operating parameters of one end cannot alleviate both types of problems at the same time. Both ends need to work together to reduce load.

[0031] S6. When the connection conflict diagnosis result indicates that the state alignment deviation is too large, the resource competition is intensified, or the control target is completely conflicted, the control command correction amount is calculated based on the connection conflict diagnosis result, and a set of control correction commands is generated and issued to dynamically adjust the operating parameters of the front-end extraction computing power allocation scheduler and the back-end graph update executor.

[0032] In this embodiment, the operating parameters include the batch size of the backend graph database fetch, the write batch size, the transaction lock granularity, the topology fusion depth, and the number of concurrent threads in the frontend extraction server cluster; step S6 includes: generating an instruction to increase the single fetch batch size of the backend graph database for cases where the state alignment deviation is too large, wherein the adjustment range of the batch size is related to... The difference is proportional; in response to the increased resource competition, instructions are generated to reduce the write batch size of the backend graph database and split the transaction lock granularity, and frontend backpressure rate limiting instructions are generated. The reduction in write batch size is proportional to... The difference is proportional; in response to a complete conflict with the control objective, an instruction is generated to reduce the number of concurrent threads in the front-end extraction server cluster, and the reduction in the number of threads is proportional to... and The product is proportional to the value of the product, and instructions are generated to reduce the topology fusion depth of the backend graph database. The reduction in fusion depth is proportional to the time consumed by a single global topology fusion and write operation. The current value is proportional to the output; the output is a set of control correction instructions for the front-end extraction computing power allocation scheduler and the back-end graph update executor.

[0033] In this embodiment, when the connection conflict diagnosis result indicates excessive state alignment deviation, intensified resource contention, or complete conflict of control objectives, the system calculates the control command correction amount based on the corresponding diagnosis result, generates and issues a set of control correction commands, and dynamically adjusts the operating parameters of the front-end extraction computing power allocation scheduler and the back-end graph update executor. Adjustable operating parameters include the back-end graph database pull batch size, write batch size, transaction lock granularity, topology fusion depth, and the number of concurrent threads in the front-end extraction server cluster. For the diagnosis result of excessive state alignment deviation, the system generates a command to increase the single pull batch size of the back-end graph database. The adjustment range of the batch size is related to... The difference is proportional to the value of the data. In this embodiment, the preset adjustment ratio is that for every 0.1 difference, the batch size of a single fetch increases by 15%. By increasing the amount of data processed by the backend in a single transaction, the fusion and writing rhythm is accelerated, the data residence time is shortened, and the freshness is brought back to a safe range. In response to the diagnosis of intensified resource competition, the system generates instructions to reduce the backend graph database write batch and split the transaction lock granularity. At the same time, it generates frontend backpressure rate limiting instructions. The reduction in write batch size is proportional to the value of the data. The difference is proportional to the value of the data. In this embodiment, the preset adjustment ratio is that for every 0.5 difference, the write batch size is reduced by 20%. By reducing the write range of a single transaction, the scope and duration of lock resource occupation are reduced, alleviating lock contention pressure. At the same time, front-end backpressure rate limiting is used to moderately reduce the data injection speed, preventing the back-end pressure from continuously increasing. For the diagnostic result of a complete conflict with the control objective, the system generates an instruction to reduce the number of concurrent threads in the front-end extraction server cluster. The reduction in the number of threads is proportional to the value of the data. and The product is proportional, reducing the number of candidate subgraphs generated per unit time from the source, while generating instructions to reduce the topology fusion depth of the backend graph database. The reduction in fusion depth is proportional to the time consumed by a single global topology fusion and write. The current value is proportional to the value of the algorithm. By reducing the amount of topological correlation calculations in a single fusion, the write time and lock occupation range are reduced. The coordinated adjustment at both ends allows the system state to gradually return to the safe range. After all adjustment instructions are generated, they are uniformly encapsulated into a set of control correction instructions, which are sent to the front-end extraction computing power allocation scheduler and the back-end graph update executor, respectively, so that the corresponding running parameters can be dynamically applied.

[0034] 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-described enterprise-level knowledge graph construction method for heterogeneous data fusion by calling the computer program stored in the memory.

[0035] 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 enterprise-level knowledge graph construction method for heterogeneous data fusion 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.

[0036] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be implemented 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, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0037] 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 conjunction with an instruction execution system, apparatus, or device.

[0038] 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.

[0039] 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.

[0040] 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 method for constructing an enterprise-level knowledge graph for heterogeneous data fusion, characterized in that: The specific steps include the following: S1. Obtain the underlying parameter set for the current period to provide basic data input; the underlying parameter set includes the time consumption of single batch heterogeneous data extraction, absolute timestamp of candidate subgraph generation, current absolute timestamp, time consumption of single global topology fusion and writing, average degree of the graph nodes to be updated, maximum data delay time, and graph database write lock sensitivity coefficient. S2. Perform a difference operation based on the current absolute timestamp and the absolute timestamp generated by the candidate subgraph to output the data residence time; S3. Based on the data residence time and the maximum data delay time, determine and obtain the freshness decay ratio; at the same time, based on the single global topology fusion and write time, the single batch heterogeneous data extraction time, the average degree of the graph nodes to be updated, and the graph database write lock sensitivity coefficient, calculate and obtain the write lock contention index. S4. Combine the freshness decay ratio with the write lock contention index into two-dimensional coordinate points to generate a cross-module freshness and write lock contention state space criterion. S5. Perform a logical comparison of the cross-module freshness and write lock contention state space criterion, determine its relationship with the preset state space boundary, and output the connection conflict diagnosis result. S6. When the connection conflict diagnosis result indicates that the state alignment deviation is too large, the resource competition is intensified, or the control target is completely conflicted, the control command correction amount is calculated based on the connection conflict diagnosis result, and a set of control correction commands is generated and issued to dynamically adjust the operating parameters of the front-end extraction computing power allocation scheduler and the back-end graph update executor.

2. The enterprise-level knowledge graph construction method for heterogeneous data fusion according to claim 1, characterized in that: The specific content of S1 is as follows: collecting operation logs by probes deployed on the front-end extraction server cluster, and obtaining the absolute timestamp of the candidate subgraph generation and the current absolute timestamp by combining the Coordinated Universal Time synchronized by the Network Time Protocol; obtaining transaction execution logs through the built-in transaction log interface of the back-end graph database, and obtaining the running status of the back-end graph database through the query interface; obtaining the global configuration file of the supply chain business through the configuration center; and outputting the underlying parameter set of the current period by combining the above information.

3. The enterprise-level knowledge graph construction method for heterogeneous data fusion according to claim 2, characterized in that: The specific content of S2 is: based on the current absolute timestamp Generate absolute timestamps with the candidate subgraphs Perform a difference operation and output the data residence time after the time base alignment is completed. .

4. The enterprise-level knowledge graph construction method for heterogeneous data fusion according to claim 3, characterized in that: S3 includes the following specific steps: S31. Divide the data dwell time by the maximum data delay time. The freshness decay ratio was obtained. The calculation formula is: ; S32, The time consumed by the single global topology fusion and writing is calculated. Divide by the time consumed in extracting heterogeneous data in a single batch and the average degree of the nodes in the graph to be updated. and the graph database write lock sensitivity coefficient Multiply by each other to obtain the write lock contention index. The calculation formula is: .

5. The enterprise-level knowledge graph construction method for heterogeneous data fusion according to claim 4, characterized in that: S4 includes: based on the freshness decay ratio Contention index with the write lock Combined into a two-dimensional coordinate point; generate and output the cross-module freshness and write lock contention state space criterion. .

6. The enterprise-level knowledge graph construction method for heterogeneous data fusion according to claim 5, characterized in that: S5 includes: based on the cross-module freshness and write lock contention state space criterion. It is logically compared with the preset state space boundary; whereby the freshness warning boundary is set as... Set the write lock security threshold to ;like and The output connection conflict diagnosis result indicates that the state alignment deviation is too large; if and The output connection conflict diagnosis result indicates that resource contention has intensified; if and The output connection conflict diagnosis result indicates that the control target is in complete conflict.

7. The enterprise-level knowledge graph construction method for heterogeneous data fusion according to claim 5, characterized in that, The operating parameters include the batch size of the backend graph database fetch, the write batch size, the transaction lock granularity, the topology fusion depth, and the number of concurrent threads in the frontend extraction server cluster; step S6 includes: generating an instruction to increase the single fetch batch size of the backend graph database for cases where the state alignment deviation is too large, the adjustment range of the batch size being... The difference is proportional; in response to the increased resource competition, instructions are generated to reduce the write batch size of the backend graph database and split the transaction lock granularity, and frontend backpressure rate limiting instructions are generated. The reduction in write batch size is proportional to... The difference is proportional; in response to a complete conflict with the control objective, an instruction is generated to reduce the number of concurrent threads in the front-end extraction server cluster, and the reduction in the number of threads is proportional to... and The product is proportional to the value of the product, and instructions are generated to reduce the topology fusion depth of the backend graph database. The reduction in fusion depth is proportional to the time consumed by a single global topology fusion and write operation. The current value is proportional to the output; the output is a set of control correction instructions for the front-end extraction computing power allocation scheduler and the back-end graph update executor.

8. 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 an enterprise-level knowledge graph construction method for heterogeneous data fusion as described in any one of claims 1-7.

9. 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 an enterprise-level knowledge graph construction method for heterogeneous data fusion as described in any one of claims 1 to 7.

Images