Financial time sequence graph processing method and device, computer equipment and storage medium

The graph partitioner selection system predicts the comprehensive time overhead of each graph partitioner and selects the most suitable target graph partitioner, which solves the inefficiency problem caused by traditional random selection and improves the efficiency of distributed graph processing.

CN119026679APending Publication Date: 2024-11-26ZHEJIANG BANGSUN TECH CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411123901.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

In traditional distributed graph processing, randomly selecting target graph partitioners may be inappropriate, resulting in inefficient processing.

Method used

The graph partitioner selection system is adopted to predict the comprehensive time overhead of each graph partitioner through the partition quality predictor, execution time predictor and partition time predictor, so as to select the target graph partitioner with the least comprehensive time overhead for graph partitioning.

Benefits of technology

It improves the efficiency of distributed graph processing, reduces time overhead, and ensures the appropriateness of the selection of graph partitioners.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026679A_ABST
    Figure CN119026679A_ABST
Patent Text Reader

Abstract

The invention relates to a financial time sequence graph processing method and device, computer equipment and a storage medium. The method comprises the steps of obtaining a first time sequence knowledge graph, a first instruction corresponding to a to-be-executed first distributed graph processing algorithm and at least two second instructions corresponding to at least two first graph partitions; inputting the first time sequence knowledge graph, the first instruction and the at least two second instructions to a graph partitioning device selection system to obtain a target graph partitioning device selected by the graph partitioning device selection system; the graph partitioning device selection system is used for selecting a target graph partitioning device corresponding to the predicted minimum comprehensive time overhead from the at least two first graph partitioning devices; performing graph partitioning on the first time sequence knowledge graph based on a target graph partitioning device to obtain a plurality of first time sequence sub-graphs; and performing distributed graph processing on the plurality of first time sequence sub-graphs based on a first distributed graph processing algorithm to obtain a graph processing result. By adopting the method, the distributed graph processing efficiency can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of big data technology, and in particular to a method, device, computer equipment and storage medium for processing financial time series graphs. Background Art

[0002] With the development of big data technology, distributed graph processing technology has emerged. When performing distributed graph processing, it is necessary to first determine the target graph partitioner from at least two graph partitioners, and then use the target graph partitioner to partition the temporal knowledge graph into multiple temporal subgraphs. Each temporal subgraph can be processed one by one by each node in the distributed system. Figure 1 As shown, a time series knowledge graph is provided. The time series knowledge graph includes nodes (such as "Person" and "Room") and edges (such as the connection between "Person" and "Room"). It can be seen that both nodes and edges include time series information, namely "time".

[0003] In conventional techniques, when determining a target graph partitioner from at least two graph partitioners, it is generally selected randomly.

[0004] However, the efficiency of distributed graph processing is affected by graph partitioning. Different graph partitioners may be suitable for different temporal knowledge graphs. The randomly selected target graph partitioner may be inappropriate, resulting in low efficiency of distributed graph processing. Summary of the invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment and storage medium for processing financial time series graphs that can improve the efficiency of distributed graph processing in response to the above technical problems.

[0006] In a first aspect, the present application provides a method for processing a financial time series graph, comprising:

[0007] Obtain a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners;

[0008] Input the first temporal knowledge graph, the first instruction and the at least two second instructions to a graph partitioner selection system to obtain a target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used to select a target graph partitioner corresponding to the predicted minimum comprehensive time overhead from the at least two first graph partitioners; the minimum comprehensive time overhead is predicted based on the first temporal knowledge graph, the first instruction and the at least two second instructions;

[0009] Performing graph partitioning on the first temporal knowledge graph based on the target graph partitioner to obtain a plurality of first temporal subgraphs;

[0010] Based on the first distributed graph processing algorithm, distributed graph processing is performed on the multiple first time-series sub-graphs to obtain a graph processing result.

[0011] In one of the embodiments, the graph partitioner selection system includes a partition quality predictor, an execution duration predictor, a partition duration predictor, and a graph partitioner selection module;

[0012] The execution time predictor is used to predict the graph processing time corresponding to each of the first graph partitioners based on the partition quality predictor;

[0013] The partition duration predictor is used to predict the partition duration corresponding to each of the first graph partitioners;

[0014] The graph partitioner selection module is used to calculate the comprehensive time overhead corresponding to each of the first graph partitioners based on the partition duration and the graph processing duration, and select the target graph partitioner corresponding to the minimum comprehensive time overhead from each of the first graph partitioners.

[0015] In one of the embodiments, the partition quality predictor is specifically used to respectively predict the partition quality of the first graph partitioning performed by each of the first graph partitioners on the first temporal knowledge graph;

[0016] The execution time predictor is specifically used to predict the graph processing time of the first distributed graph processing algorithm to perform distributed graph processing based on the quality of each partition;

[0017] The partition duration predictor is specifically used to predict the partition duration of each of the first graph partitioners for partitioning the first temporal knowledge graph.

[0018] In one embodiment, the step of inputting the first temporal knowledge graph, the first instruction, and the at least two second instructions to the graph partitioner selection system includes:

[0019] Obtain a partition quality predictor to be trained, a partition execution time predictor to be trained, a partition duration predictor to be trained, and a training sample data set corresponding to each predictor;

[0020] Based on the corresponding training sample data sets, the partition quality predictor to be trained, the execution duration predictor to be trained and the partition duration predictor to be trained are trained respectively to obtain the partition quality predictor, the execution duration predictor and the partition duration predictor respectively.

[0021] In one embodiment, obtaining the training sample data sets corresponding to each predictor includes:

[0022] Acquire multiple second temporal knowledge graphs, multiple second graph partitioners, and multiple second distributed graph processing algorithms;

[0023] Based on each of the second temporal knowledge graphs and each of the second graph partitioners, a partition quality label and a partition duration label are determined respectively; each of the second graph partitioners is paired with each of the second temporal knowledge graphs, and each of the pairs corresponds to one of the partition quality labels and one of the partition duration labels;

[0024] Determine a graph processing duration label based on each of the second temporal knowledge graphs, each of the second graph partitioners, and each of the second distributed graph processing algorithms;

[0025] Each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition quality labels are used as the training sample data set corresponding to the partition quality predictor to be trained; each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition duration labels are used as the training sample data set corresponding to the partition duration predictor to be trained; each of the second time series knowledge graphs, the partition quality labels, the second identifier corresponding to each of the second distributed graph processing algorithms, and the graph processing duration labels are used as the training sample data set corresponding to the execution duration predictor to be trained.

[0026] In one embodiment, the determining of the partition quality label and the partition duration label based on each of the second temporal knowledge graphs and each of the second graph partitioners comprises:

[0027] Use each of the second graph partitioners to perform graph partitioning on each of the second temporal knowledge graphs, and measure the partition quality and operation duration corresponding to the graph partition to obtain a corresponding partition quality label and a corresponding partition duration label.

[0028] In one embodiment, determining the graph processing duration label based on each of the second temporal knowledge graphs, each of the second graph partitioners, and each of the second distributed graph processing algorithms includes:

[0029] Use each of the second distributed graph processing algorithms to perform distributed graph processing on multiple second time-series sub-graphs obtained by graph partitioning to obtain corresponding graph processing duration labels.

[0030] In a second aspect, the present application further provides a device for processing a financial time series graph, including:

[0031] A first acquisition module is used to acquire a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners;

[0032] An input module, used for inputting the first temporal knowledge graph, the first instruction and the at least two second instructions into a graph partitioner selection system, and obtaining a target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used for selecting a target graph partitioner corresponding to the predicted minimum comprehensive time cost from the at least two first graph partitioners; the minimum comprehensive time cost is predicted based on the first temporal knowledge graph, the first instruction and the at least two second instructions;

[0033] A graph partitioning module, configured to perform graph partitioning on the first temporal knowledge graph based on the target graph partitioner to obtain a plurality of first temporal subgraphs;

[0034] A graph processing module is used to perform distributed graph processing on the multiple first time-series sub-graphs based on the first distributed graph processing algorithm to obtain a graph processing result.

[0035] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0036] Obtain a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners;

[0037] Input the first temporal knowledge graph, the first instruction and the at least two second instructions to a graph partitioner selection system to obtain a target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used to select a target graph partitioner corresponding to the predicted minimum comprehensive time overhead from the at least two first graph partitioners; the minimum comprehensive time overhead is predicted based on the first temporal knowledge graph, the first instruction and the at least two second instructions;

[0038] Performing graph partitioning on the first temporal knowledge graph based on the target graph partitioner to obtain a plurality of first temporal subgraphs;

[0039] Based on the first distributed graph processing algorithm, distributed graph processing is performed on the multiple first time-series sub-graphs to obtain a graph processing result.

[0040] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0041] Obtain a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners;

[0042] Input the first temporal knowledge graph, the first instruction and the at least two second instructions to a graph partitioner selection system to obtain a target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used to select a target graph partitioner corresponding to the predicted minimum comprehensive time overhead from the at least two first graph partitioners; the minimum comprehensive time overhead is predicted based on the first temporal knowledge graph, the first instruction and the at least two second instructions;

[0043] Performing graph partitioning on the first temporal knowledge graph based on the target graph partitioner to obtain a plurality of first temporal subgraphs;

[0044] Based on the first distributed graph processing algorithm, distributed graph processing is performed on the multiple first time-series sub-graphs to obtain a graph processing result.

[0045] The above-mentioned financial time series graph processing method, device, computer equipment and storage medium do not randomly select the target graph partitioner when selecting the target graph partitioner, but use the graph partitioner selection system to determine the minimum comprehensive time overhead and select according to the minimum comprehensive time overhead to obtain the target graph partitioner. The minimum comprehensive time overhead is predicted based on the first time series knowledge graph, the first instruction and the at least two second instructions. It can be understood that when the first time series knowledge graph is partitioned by the target graph partitioner corresponding to the minimum comprehensive time overhead and the graph processing result is obtained, the time overhead of distributed graph processing can be reduced to a minimum, thereby improving the efficiency of distributed graph processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 A temporal knowledge graph involved in this application;

[0048] Figure 2 It is a flowchart of a method for processing a financial time series graph in one embodiment;

[0049] Figure 3 A structural block diagram of a partitioner selection system in one embodiment;

[0050] Figure 4 A schematic diagram of a process of training each predictor in another embodiment;

[0051] Figure 5 is a structural block diagram of a device for processing a financial time series graph in one embodiment;

[0052] Figure 6 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] In an exemplary embodiment, Figure 2 As shown, a method for processing a financial time series graph is provided, which can be applied to an entry node of a distributed system or other servers connected to the distributed system, including the following steps 202 to 208. The distributed system includes but is not limited to Pregel, Giraph, GraphX ​​and Neo4j. Among them:

[0055] Step 202, obtain a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners.

[0056] Among them, the time series knowledge graph includes but is not limited to the capital chain graph and the transfer relationship graph.

[0057] Among them, the first time-series knowledge graph, the first instruction and at least two second instructions are provided by the terminal.

[0058] The first instruction is for the graph partitioner selection system to identify the first distributed graph processing algorithm to be executed, and the second instruction is for the graph partition selection system to identify each first graph partitioner. The instruction includes at least an identifier, which may be a number, a logo, etc. The distributed graph processing algorithm may be a PageRank algorithm (webpage ranking), a shortest path algorithm, or a community detection algorithm, etc.

[0059] The graph partitioner is implemented by a graph partitioning algorithm, which includes but is not limited to a segmentation algorithm based on a sliding window and a segmentation algorithm based on change point detection.

[0060] Exemplarily, obtaining the first temporal knowledge graph, the first instruction, and at least two second instructions may be a graph processing request sent by a receiving terminal, and then parsed from the graph processing request.

[0061] Step 204, input the first sequential knowledge graph, the first instruction and at least two second instructions to the graph partitioner selection system to obtain the target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used to select the target graph partitioner corresponding to the predicted minimum comprehensive time overhead from at least two first graph partitioners; the minimum comprehensive time overhead is predicted based on the first sequential knowledge graph, the first instruction and at least two second instructions.

[0062] Among them, the graph partitioner selection system can select a target graph partitioner suitable for the first distributed graph processing algorithm to be executed from at least two graph partitioners based on the first temporal knowledge graph, the first instruction and the at least two second instructions. Suitable means that the target graph partitioner is relatively efficient in graph processing by the distributed graph processing algorithm after the first temporal knowledge graph is partitioned, compared with other graph partitioners in the at least two graph partitioners.

[0063] The selection rule is to determine the minimum integrated time cost from at least two integrated time costs corresponding to at least two first graph partitioners, and use the first graph partitioner corresponding to the minimum integrated time cost as the target graph partitioner.

[0064] It can be understood that the efficiency is determined by the amount of comprehensive time overhead, and specifically, the efficiency is negatively correlated with the comprehensive time overhead.

[0065] It can be understood that the graph partitioner selection system can make quantitative predictions of the comprehensive time overhead for different first distributed graph processing algorithms to be executed, that is, it can provide suitable target graph partitioners for various first distributed graph processing algorithms.

[0066] Exemplarily, when inputting the first temporal knowledge graph into the graph partitioner selection system, it can be directly input into the graph partitioner selection system, or the first temporal knowledge graph can be preprocessed first, and the edge information, node sparsity, timing information, etc. obtained by the preprocessing can be input into the graph partitioner selection system.

[0067] Step 206: partition the first temporal knowledge graph based on the target graph partitioner to obtain multiple first temporal subgraphs.

[0068] Exemplarily, when a time-series knowledge graph is processed through a distributed system, since it can be processed simultaneously by each node in the distributed system, it is necessary to partition the first time-series knowledge graph into multiple first time-series sub-graphs, and each first time-series sub-graph can be processed by a node respectively.

[0069] Step 208: Perform distributed graph processing on the plurality of first sequential sub-graphs based on the first distributed graph processing algorithm to obtain a graph processing result.

[0070] The graph processing result varies with different distributed graph processing algorithms. It should be noted that different distributed graph processing algorithms can meet different graph processing requirements. If the distributed graph processing algorithm is the shortest path, the graph processing result is the shortest path from one node to another node in the first temporal knowledge graph.

[0071] Exemplarily, each node performs distributed graph processing on the first time-series subgraphs distributed thereto. In this process, since the first time-series knowledge graph is divided into multiple first time-series subgraphs by the target graph partitioner, there may be a corresponding relationship between the first time-series subgraphs. That is, there is an edge between a node in a first time-series subgraph and a node in another first time-series subgraph in the first time-series knowledge graph, and the edge is cut off by the target graph partitioner, and there may be multiple cut edges, that is, the corresponding relationship is multiple edges.

[0072] This correspondence needs to be used in distributed graph processing, so communication between nodes is required based on this correspondence.

[0073] It should be noted that in practical applications, in addition to randomly selecting a graph partitioner, there is also a technique for selecting a target graph partitioner by cutting size. It should be noted that the graph partitioner partitions the temporal knowledge graph by removing some edges in the temporal knowledge graph. Among them, the cutting size refers to the sum of the weights of all the edges removed. The weight is a numerical attribute of the edge, which represents the importance of the edge. The target graph partitioner corresponds to the smallest cutting size.

[0074] It can be understood that this selection method is different from the concept of the present application. When selecting a graph partitioner, the present application not only considers the factors of the graph partitioner itself, but also considers the impact of the graph partitioner on the graph processing process, specifically the overall execution time, thereby minimizing the duration of distributed graph processing and thereby improving the efficiency of distributed graph processing.

[0075] In the above-mentioned processing method of the financial timing graph, when selecting the target graph partitioner, it is not randomly selected, but the minimum comprehensive time overhead is determined by the graph partitioner selection system and the selection is made according to the minimum comprehensive time overhead, thereby obtaining the target graph partitioner. The minimum comprehensive time overhead is predicted based on the first timing knowledge graph, the first instruction, and at least two second instructions.

[0076] It can be understood that when the first temporal knowledge graph is partitioned by the target graph partitioner corresponding to the minimum comprehensive time overhead and the graph processing result is obtained thereby, the time overhead of distributed graph processing can be reduced to a minimum, thereby improving the efficiency of distributed graph processing.

[0077] In one embodiment, if Figure 3 As shown, a graph partitioner selection system is provided. The graph partitioner selection system 302 includes a partition quality predictor, an execution duration predictor, a partition duration predictor, and a graph partitioner selection module.

[0078] Among them, the execution time predictor is used to predict the graph processing time corresponding to each first graph partitioner based on the partition quality predictor; the partition time predictor is used to predict the partition time corresponding to each first graph partitioner; the graph partitioner selection module is used to calculate the comprehensive time overhead corresponding to each first graph partitioner based on each partition time and each graph processing time, and select the target graph partitioner corresponding to the least comprehensive time overhead from each first graph partitioner.

[0079] Among them, the partition quality predictor, execution time predictor and partition duration predictor are all implemented through a machine learning model, and the machine learning model can be implemented through algorithms such as regression tree and nearest neighbor.

[0080] It should be noted that, for the same first temporal knowledge graph, assuming that it is partitioned by each first graph partitioner and processed by the first distributed graph processing algorithm, the graph processing time corresponding to different first graph partitioners is different.

[0081] Correspondingly, the partitioning duration is similar to the graph processing duration, and different first graph partitioners correspond to different partitioning durations.

[0082] Among them, the graph partitioner selection module selects the target graph partitioner based on the comprehensive time cost. For the process of calculating based on each partition duration and each graph processing duration, it is also for each first graph partitioner, that is, for each first graph partitioner, the sum of its corresponding partition duration and graph processing duration is calculated to obtain the comprehensive time cost corresponding to the first graph partitioner.

[0083] It should be noted that for the graph processing time, the graph processing is performed by a distributed system. When each node in the distributed system processes its own first sequential subgraph in parallel, the processing time of each node may be different due to the different information of each first sequential subgraph. Based on this, the graph processing time can be the maximum value of the processing time of each node, or the average value of the processing time of each node.

[0084] In another embodiment, the graph partitioner selection system may include an execution duration predictor, a partition duration predictor, and a graph partitioner selection module, wherein part of the input of the execution duration predictor may be obtained through measurement, and the part of the input is the partition quality.

[0085] Furthermore, a partition quality predictor is specifically used to predict the partition quality of each first graph partitioner for performing graph partitioning on the first temporal knowledge graph; an execution time predictor is specifically used to predict the graph processing time of the first distributed graph processing algorithm for distributed graph processing based on the quality of each partition; a partition time predictor is specifically used to predict the partition time of each first graph partitioner for performing graph partitioning on the first temporal knowledge graph.

[0086] Prediction means that the output can be obtained through input without performing corresponding operations on the predicted object. For example, the partitioning quality of each first graph partitioner for the first temporal knowledge graph is predicted respectively, where the predicted object is the first temporal knowledge graph and the second instruction corresponding to the first graph partitioner, the corresponding operation is graph partitioning, and the input is the predicted object. The corresponding partitioning quality can be obtained through prediction without the first graph partitioner having to perform graph partitioning on the first temporal knowledge graph and then measuring the partitioning quality.

[0087] Partition quality includes but is not limited to replication factor, edge balance, point balance, source point balance and sink point balance. The replication factor is the sum of the number of vertices in each partition or the number of points in the entire graph, the edge balance is the number of edges in the partition with the largest number of edges or the average number of edges in the partition, the point balance is the number of edges in the partition with the largest number of points or the average number of points in the partition, the source point balance is the number of edges in the partition with the largest number of source points or the average number of source points in the partition, and the sink point balance is the number of edges in the partition with the largest number of sink points or the average number of sink points in the partition.

[0088] It can be understood that the partition quality predictor can provide quantitative predictions of partition quality for different first graph partitioners, and the partition duration predictor can provide quantitative predictions of partition duration for different first graph partitioners. On this basis, the execution duration predictor can provide quantitative predictions of execution duration for different first distributed graph processing algorithms. Based on this, the purpose of providing suitable target graph partitioners for various first distributed graph processing algorithms is achieved.

[0089] In an exemplary embodiment, the partition quality predictor, execution time predictor and partition time predictor as machine learning models all need to be trained on the initial model before they can be used. Figure 4 As shown, before inputting the first temporal knowledge graph, the first instruction and at least two second instructions to the graph partitioner selection system, the method further includes steps 402 to 404. Among them:

[0090] Step 402: Obtain a partition quality predictor to be trained, a partition execution duration predictor to be trained, a partition duration predictor to be trained, and a training sample data set corresponding to each predictor.

[0091] Among them, the partition quality predictor to be trained, the execution time predictor to be trained and the partition duration predictor to be trained can be pre-trained models, or models constructed by algorithms such as regression tree and nearest neighbor.

[0092] Step 404 , based on the corresponding training sample data set, the partition quality predictor to be trained, the execution duration predictor to be trained and the partition duration predictor to be trained are trained respectively to obtain the partition quality predictor, the execution duration predictor and the partition duration predictor respectively.

[0093] Among them, each predictor has its own corresponding training sample data set, that is, each predictor can be trained independently.

[0094] Exemplarily, the partition quality predictor to be trained and the execution time predictor to be trained may also be trained jointly.

[0095] It should be noted that since the output of the partition quality predictor to be trained can be used as the input of the execution time predictor to be trained, in the case of joint training, it is necessary to wait for the single iteration training of the partition quality predictor to be trained to end and output the result before training the execution time predictor to be trained, resulting in a longer training time. However, if the three predictors are trained independently, they can be executed in parallel, and each predictor does not need to wait for the execution results of other predictors, thus improving the training efficiency.

[0096] Exemplarily, in the process of training the partition quality predictor to be trained, the execution time predictor to be trained and the partition duration predictor to be trained respectively based on the corresponding training sample data sets to obtain the partition quality predictor, the execution time predictor and the partition duration predictor respectively, the training process of each predictor is iterative training, and after each round of iteration, it is necessary to determine whether the iteration end condition is met. If it is met, the training ends, if not, it enters the next round of iteration.

[0097] The iteration end condition may be that the number of iterations reaches a preset number, or that the loss value is less than or equal to a loss threshold. The loss value is calculated by the predicted value and the true value.

[0098] The preset number of times and the loss threshold are not limited in this embodiment.

[0099] In one embodiment, obtaining the training sample data sets corresponding to the predictors includes:

[0100] Acquire multiple second temporal knowledge graphs, multiple second graph partitioners, and multiple second distributed graph processing algorithms.

[0101] Among them, each second time series knowledge graph can be a synthetic graph or a real graph in a real business scenario, that is, various different time series knowledge graphs are generated or collected. The use of real graphs can improve the prediction accuracy of each predictor, thereby improving the accuracy of the target graph partitioner selected by the graph partitioner selection system, and further improving the efficiency of distributed graph processing.

[0102] Based on each second temporal knowledge graph and each second graph partitioner, a partition quality label and a partition duration label are determined respectively; each second graph partitioner is paired with each second temporal knowledge graph in pairs, and each pair corresponds to a partition quality label and a partition duration label.

[0103] Among them, each label is the true value mentioned above, that is, each label can be used to determine whether the iteration is completed.

[0104] A graph processing duration label is determined based on each second temporal knowledge graph, each second graph partitioner, and each second distributed graph processing algorithm.

[0105] Among them, for the second distributed graph processing algorithms based on the same second temporal knowledge graph and the same second graph partitioner but different second distributed graph processing algorithms, the determined graph processing time labels are different; similarly, based on the same second temporal knowledge graph and the same second distributed graph processing algorithm but different second graph partitioners, the determined graph processing time labels are different; similarly, based on the same second graph partitioner and the same second distributed graph processing algorithm but different second temporal knowledge graphs, the determined graph processing time labels are different.

[0106] Each second time series knowledge graph, each second graph partitioner corresponding to the first identifier and each partition quality label is used as the training sample data set corresponding to the partition quality predictor to be trained; each second time series knowledge graph, each second graph partitioner corresponding to the first identifier and each partition duration label is used as the training sample data set corresponding to the partition duration predictor to be trained; each second time series knowledge graph, each partition quality label, each second distributed graph processing algorithm corresponding to the second identifier and each graph processing duration label is used as the training sample data set corresponding to the execution duration predictor to be trained.

[0107] Exemplarily, for the above-mentioned joint training case, when training the partition quality predictor to be trained and the execution duration predictor to be trained, no partition quality label is required, that is, the iteration end condition of the two predictors can be determined by the graph processing duration label without being determined by the partition quality label.

[0108] In one embodiment, determining the partition quality label and the partition duration label based on each second temporal knowledge graph and each second graph partitioner can be implemented in the following manner:

[0109] Use each second graph partitioner to perform graph partition on each second temporal knowledge graph, measure the partition quality and operation duration corresponding to the graph partition, and obtain the corresponding partition quality label and the corresponding partition duration label.

[0110] Exemplarily, each second graph partitioner is paired with each second temporal knowledge graph in pairs to obtain a number of graph partition pairs, and then the second temporal knowledge graph is partitioned by the second graph partitioner in each graph partition pair, so as to obtain a partition quality label and a corresponding partition duration label for each graph partition pair.

[0111] In one embodiment, determining the graph processing duration label based on each second temporal knowledge graph, each second graph partitioner, and each second distributed graph processing algorithm may be implemented in the following manner:

[0112] The second distributed graph processing algorithms are used to perform distributed graph processing on the multiple second time-series subgraphs obtained by graph partitioning to obtain corresponding graph processing duration labels.

[0113] Exemplarily, each of the above graph partition pairs corresponds to multiple second temporal subgraphs, that is, different graph partition pairs correspond to different second temporal subgraphs. For example, there are graph partition pairs c1 and graph partition pairs c2, wherein graph partition pair c1 includes graph partitioner b3 and temporal knowledge graph a3, and graph partition pair c1 includes graph partitioner b4 and temporal knowledge graph a4. After graph partitioning the temporal knowledge graph a3 by graph partitioner b3, temporal subgraph d1, temporal subgraph d2 and temporal subgraph d3 are obtained; after graph partitioning the temporal knowledge graph a4 by graph partitioner b4, temporal subgraph d4, temporal subgraph d5 and temporal subgraph d6 are obtained.

[0114] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0115] Based on the same inventive concept, the embodiment of the present application also provides a device for processing a financial time series graph for implementing the method for processing a financial time series graph involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of one or more financial time series graph processing devices provided below can refer to the limitations of the method for processing a financial time series graph above, and will not be repeated here.

[0116] In an exemplary embodiment, Figure 5 As shown, a processing device for a financial time series graph is provided, comprising: a first acquisition module, an input module, a graph partitioning module and a graph processing module, wherein:

[0117] A first acquisition module is used to acquire a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners;

[0118] An input module, used for inputting the first temporal knowledge graph, the first instruction and the at least two second instructions into a graph partitioner selection system, and obtaining a target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used for selecting a target graph partitioner corresponding to the predicted minimum comprehensive time cost from the at least two first graph partitioners; the minimum comprehensive time cost is predicted based on the first temporal knowledge graph, the first instruction and the at least two second instructions;

[0119] A graph partitioning module, configured to perform graph partitioning on the first temporal knowledge graph based on the target graph partitioner to obtain a plurality of first temporal subgraphs;

[0120] A graph processing module is used to perform distributed graph processing on the multiple first time-series sub-graphs based on the first distributed graph processing algorithm to obtain a graph processing result.

[0121] In one embodiment, the graph partitioner selection system includes a partition quality predictor, an execution duration predictor, a partition duration predictor, and a graph partitioner selection module;

[0122] The execution time predictor is used to predict the graph processing time corresponding to each of the first graph partitioners based on the partition quality predictor;

[0123] The partition duration predictor is used to predict the partition duration corresponding to each of the first graph partitioners;

[0124] The graph partitioner selection module is used to calculate the comprehensive time overhead corresponding to each of the first graph partitioners based on the partition duration and the graph processing duration, and select the target graph partitioner corresponding to the minimum comprehensive time overhead from each of the first graph partitioners.

[0125] In one embodiment, the partition quality predictor is specifically used to respectively predict the partition quality of each of the first graph partitioners performing graph partitioning on the first temporal knowledge graph;

[0126] The execution time predictor is specifically used to predict the graph processing time of the first distributed graph processing algorithm to perform distributed graph processing based on the quality of each partition;

[0127] The partition duration predictor is specifically used to predict the partition duration of each of the first graph partitioners for partitioning the first temporal knowledge graph.

[0128] In one embodiment, the financial time series graph processing device further includes:

[0129] A second acquisition module is used to acquire a partition quality predictor to be trained, an execution duration predictor to be trained, a partition duration predictor to be trained, and a training sample data set corresponding to each predictor;

[0130] A training module is used to train the partition quality predictor to be trained, the execution duration predictor to be trained and the partition duration predictor to be trained based on the corresponding training sample data set, and obtain the partition quality predictor, the execution duration predictor and the partition duration predictor respectively.

[0131] In one embodiment, the second acquisition module includes:

[0132] An acquisition unit, used to acquire a plurality of second temporal knowledge graphs, a plurality of second graph partitioners, and a plurality of second distributed graph processing algorithms;

[0133] A first determination unit is used to determine a partition quality label and a partition duration label based on each of the second temporal knowledge graphs and each of the second graph partitioners; each of the second graph partitioners is paired with each of the second temporal knowledge graphs, and each of the pairs corresponds to one of the partition quality labels and one of the partition duration labels;

[0134] A second determining unit, configured to determine a graph processing duration label based on each of the second temporal knowledge graphs, each of the second graph partitioners, and each of the second distributed graph processing algorithms;

[0135] An execution unit is used to use each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition quality labels as the training sample data set corresponding to the partition quality predictor to be trained; use each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition duration labels as the training sample data set corresponding to the partition duration predictor to be trained; use each of the second time series knowledge graphs, the partition quality labels, the second identifier corresponding to each of the second distributed graph processing algorithms, and the graph processing duration labels as the training sample data set corresponding to the execution duration predictor to be trained.

[0136] In one embodiment, the first determining unit is specifically configured to:

[0137] Use each of the second graph partitioners to perform graph partitioning on each of the second temporal knowledge graphs, and measure the partition quality and operation duration corresponding to the graph partition to obtain a corresponding partition quality label and a corresponding partition duration label.

[0138] In one embodiment, the second determining unit is specifically configured to:

[0139] Use each of the second distributed graph processing algorithms to perform distributed graph processing on multiple second time-series sub-graphs obtained by graph partitioning to obtain corresponding graph processing duration labels.

[0140] Each module in the processing device of the above-mentioned financial time series graph can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0141] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 6As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store a first time series knowledge graph, a first instruction and a second instruction. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for processing a financial time series graph is implemented.

[0142] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0143] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0144] Obtain a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners;

[0145] Input the first temporal knowledge graph, the first instruction and the at least two second instructions to a graph partitioner selection system to obtain a target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used to select a target graph partitioner corresponding to the predicted minimum comprehensive time overhead from the at least two first graph partitioners; the minimum comprehensive time overhead is predicted based on the first temporal knowledge graph, the first instruction and the at least two second instructions;

[0146] Performing graph partitioning on the first temporal knowledge graph based on the target graph partitioner to obtain a plurality of first temporal subgraphs;

[0147] Based on the first distributed graph processing algorithm, distributed graph processing is performed on the multiple first time-series sub-graphs to obtain a graph processing result.

[0148] In one embodiment, the graph partitioner selection system includes a partition quality predictor, an execution duration predictor, a partition duration predictor, and a graph partitioner selection module;

[0149] The execution time predictor is used to predict the graph processing time corresponding to each of the first graph partitioners based on the partition quality predictor;

[0150] The partition duration predictor is used to predict the partition duration corresponding to each of the first graph partitioners;

[0151] The graph partitioner selection module is used to calculate the comprehensive time overhead corresponding to each of the first graph partitioners based on the partition duration and the graph processing duration, and select the target graph partitioner corresponding to the minimum comprehensive time overhead from each of the first graph partitioners.

[0152] In one embodiment, the partition quality predictor is specifically used to respectively predict the partition quality of each of the first graph partitioners performing graph partitioning on the first temporal knowledge graph;

[0153] The execution time predictor is specifically used to predict the graph processing time of the first distributed graph processing algorithm to perform distributed graph processing based on the quality of each partition;

[0154] The partition duration predictor is specifically used to predict the partition duration of each of the first graph partitioners for partitioning the first temporal knowledge graph.

[0155] In one embodiment, when the processor executes the computer program, the following steps are also implemented:

[0156] Obtain a partition quality predictor to be trained, a partition execution time predictor to be trained, a partition duration predictor to be trained, and a training sample data set corresponding to each predictor;

[0157] Based on the corresponding training sample data sets, the partition quality predictor to be trained, the execution duration predictor to be trained and the partition duration predictor to be trained are trained respectively to obtain the partition quality predictor, the execution duration predictor and the partition duration predictor respectively.

[0158] In one embodiment, when the processor executes the computer program, the following steps are also implemented:

[0159] Acquire multiple second temporal knowledge graphs, multiple second graph partitioners, and multiple second distributed graph processing algorithms;

[0160] Based on each of the second temporal knowledge graphs and each of the second graph partitioners, a partition quality label and a partition duration label are determined respectively; each of the second graph partitioners is paired with each of the second temporal knowledge graphs, and each of the pairs corresponds to one of the partition quality labels and one of the partition duration labels;

[0161] Determine a graph processing duration label based on each of the second temporal knowledge graphs, each of the second graph partitioners, and each of the second distributed graph processing algorithms;

[0162] Each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition quality labels are used as the training sample data set corresponding to the partition quality predictor to be trained; each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition duration labels are used as the training sample data set corresponding to the partition duration predictor to be trained; each of the second time series knowledge graphs, the partition quality labels, the second identifier corresponding to each of the second distributed graph processing algorithms, and the graph processing duration labels are used as the training sample data set corresponding to the execution duration predictor to be trained.

[0163] In one embodiment, when the processor executes the computer program, the following steps are also implemented:

[0164] Use each of the second graph partitioners to perform graph partitioning on each of the second temporal knowledge graphs, and measure the partition quality and operation duration corresponding to the graph partition to obtain a corresponding partition quality label and a corresponding partition duration label.

[0165] In one embodiment, when the processor executes the computer program, the following steps are also implemented:

[0166] Use each of the second distributed graph processing algorithms to perform distributed graph processing on multiple second time-series sub-graphs obtained by graph partitioning to obtain corresponding graph processing duration labels.

[0167] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0168] Obtain a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners;

[0169] Input the first temporal knowledge graph, the first instruction and the at least two second instructions to a graph partitioner selection system to obtain a target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used to select a target graph partitioner corresponding to the predicted minimum comprehensive time overhead from the at least two first graph partitioners; the minimum comprehensive time overhead is predicted based on the first temporal knowledge graph, the first instruction and the at least two second instructions;

[0170] Performing graph partitioning on the first temporal knowledge graph based on the target graph partitioner to obtain a plurality of first temporal subgraphs;

[0171] Based on the first distributed graph processing algorithm, distributed graph processing is performed on the multiple first time-series sub-graphs to obtain a graph processing result.

[0172] In one embodiment, the graph partitioner selection system includes a partition quality predictor, an execution duration predictor, a partition duration predictor, and a graph partitioner selection module;

[0173] The execution time predictor is used to predict the graph processing time corresponding to each of the first graph partitioners based on the partition quality predictor;

[0174] The partition duration predictor is used to predict the partition duration corresponding to each of the first graph partitioners;

[0175] The graph partitioner selection module is used to calculate the comprehensive time overhead corresponding to each of the first graph partitioners based on the partition duration and the graph processing duration, and select the target graph partitioner corresponding to the minimum comprehensive time overhead from each of the first graph partitioners.

[0176] In one embodiment, the partition quality predictor is specifically used to respectively predict the partition quality of each of the first graph partitioners performing graph partitioning on the first temporal knowledge graph;

[0177] The execution time predictor is specifically used to predict the graph processing time of the first distributed graph processing algorithm to perform distributed graph processing based on the quality of each partition;

[0178] The partition duration predictor is specifically used to predict the partition duration of each of the first graph partitioners for partitioning the first temporal knowledge graph.

[0179] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0180] Obtain a partition quality predictor to be trained, a partition execution time predictor to be trained, a partition duration predictor to be trained, and a training sample data set corresponding to each predictor;

[0181] Based on the corresponding training sample data sets, the partition quality predictor to be trained, the execution duration predictor to be trained and the partition duration predictor to be trained are trained respectively to obtain the partition quality predictor, the execution duration predictor and the partition duration predictor respectively.

[0182] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0183] Acquire multiple second temporal knowledge graphs, multiple second graph partitioners, and multiple second distributed graph processing algorithms;

[0184] Based on each of the second temporal knowledge graphs and each of the second graph partitioners, a partition quality label and a partition duration label are determined respectively; each of the second graph partitioners is paired with each of the second temporal knowledge graphs, and each of the pairs corresponds to one of the partition quality labels and one of the partition duration labels;

[0185] Determine a graph processing duration label based on each of the second temporal knowledge graphs, each of the second graph partitioners, and each of the second distributed graph processing algorithms;

[0186] Each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition quality labels are used as the training sample data set corresponding to the partition quality predictor to be trained; each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition duration labels are used as the training sample data set corresponding to the partition duration predictor to be trained; each of the second time series knowledge graphs, the partition quality labels, the second identifier corresponding to each of the second distributed graph processing algorithms, and the graph processing duration labels are used as the training sample data set corresponding to the execution duration predictor to be trained.

[0187] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0188] Use each of the second graph partitioners to perform graph partitioning on each of the second temporal knowledge graphs, and measure the partition quality and operation duration corresponding to the graph partition to obtain a corresponding partition quality label and a corresponding partition duration label.

[0189] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0190] Use each of the second distributed graph processing algorithms to perform distributed graph processing on multiple second time-series sub-graphs obtained by graph partitioning to obtain corresponding graph processing duration labels.

[0191] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0192] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0193] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0194] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for processing a financial time series graph, characterized in that: The method comprises: Obtain a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners; Input the first temporal knowledge graph, the first instruction and the at least two second instructions to a graph partitioner selection system to obtain a target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used to select a target graph partitioner corresponding to the predicted minimum comprehensive time overhead from the at least two first graph partitioners; the minimum comprehensive time overhead is predicted based on the first temporal knowledge graph, the first instruction and the at least two second instructions; Performing graph partitioning on the first temporal knowledge graph based on the target graph partitioner to obtain a plurality of first temporal subgraphs; Based on the first distributed graph processing algorithm, distributed graph processing is performed on the multiple first time-series sub-graphs to obtain a graph processing result.

2. The method according to claim 1, characterized in that The graph partitioner selection system includes a partition quality predictor, an execution duration predictor, a partition duration predictor, and a graph partitioner selection module; The execution time predictor is used to predict the graph processing time corresponding to each of the first graph partitioners based on the partition quality predictor; The partition duration predictor is used to predict the partition duration corresponding to each of the first graph partitioners; The graph partitioner selection module is used to calculate the comprehensive time overhead corresponding to each of the first graph partitioners based on the partition duration and the graph processing duration, and select the target graph partitioner corresponding to the minimum comprehensive time overhead from each of the first graph partitioners.

3. The method according to claim 2, characterized in that The partition quality predictor is specifically used to respectively predict the partition quality of the first graph partitioning performed by each of the first graph partitioners on the first temporal knowledge graph; The execution time predictor is specifically used to predict the graph processing time of the first distributed graph processing algorithm to perform distributed graph processing based on the quality of each partition; The partition duration predictor is specifically used to predict the partition duration of each of the first graph partitioners for partitioning the first temporal knowledge graph.

4. The method according to claim 2 or 3, characterized in that: The step of inputting the first temporal knowledge graph, the first instruction, and the at least two second instructions to the graph partitioner selection system includes: Obtain a partition quality predictor to be trained, a partition execution time predictor to be trained, a partition duration predictor to be trained, and a training sample data set corresponding to each predictor; Based on the corresponding training sample data sets, the partition quality predictor to be trained, the execution duration predictor to be trained and the partition duration predictor to be trained are trained respectively to obtain the partition quality predictor, the execution duration predictor and the partition duration predictor respectively.

5. The method according to claim 4, characterized in that The step of obtaining the training sample data sets corresponding to the predictors comprises: Obtaining multiple second temporal knowledge graphs, multiple second graph partitioners, and multiple second distributed graph processing algorithms; Based on each of the second temporal knowledge graphs and each of the second graph partitioners, a partition quality label and a partition duration label are determined respectively; each of the second graph partitioners is paired with each of the second temporal knowledge graphs, and each of the pairs corresponds to one of the partition quality labels and one of the partition duration labels; Determine a graph processing duration label based on each of the second temporal knowledge graphs, each of the second graph partitioners, and each of the second distributed graph processing algorithms; Each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition quality labels are used as the training sample data set corresponding to the partition quality predictor to be trained; each of the second time series knowledge graphs, the first identifier corresponding to each of the second graph partitioners, and the partition duration labels are used as the training sample data set corresponding to the partition duration predictor to be trained; each of the second time series knowledge graphs, the partition quality labels, the second identifier corresponding to each of the second distributed graph processing algorithms, and the graph processing duration labels are used as the training sample data set corresponding to the execution duration predictor to be trained.

6. The method according to claim 5, characterized in that The determining of the partition quality label and the partition duration label based on each of the second time series knowledge graphs and each of the second graph partitioners comprises: Use each of the second graph partitioners to perform graph partitioning on each of the second temporal knowledge graphs, and measure the partition quality and operation duration corresponding to the graph partition to obtain a corresponding partition quality label and a corresponding partition duration label.

7. The method according to claim 6, characterized in that The determining of the graph processing duration label based on each of the second time series knowledge graphs, each of the second graph partitioners, and each of the second distributed graph processing algorithms comprises: Use each of the second distributed graph processing algorithms to perform distributed graph processing on multiple second time-series sub-graphs obtained by graph partitioning to obtain corresponding graph processing duration labels.

8. A device for processing a financial time series graph, characterized in that: The device comprises: A first acquisition module is used to acquire a first temporal knowledge graph, a first instruction corresponding to a first distributed graph processing algorithm to be executed, and at least two second instructions corresponding to at least two first graph partitioners; An input module, used for inputting the first temporal knowledge graph, the first instruction and the at least two second instructions into a graph partitioner selection system, and obtaining a target graph partitioner selected by the graph partitioner selection system; the graph partitioner selection system is used for selecting a target graph partitioner corresponding to the predicted minimum comprehensive time cost from the at least two first graph partitioners; the minimum comprehensive time cost is predicted based on the first temporal knowledge graph, the first instruction and the at least two second instructions; A graph partitioning module, configured to perform graph partitioning on the first temporal knowledge graph based on the target graph partitioner to obtain a plurality of first temporal subgraphs; A graph processing module is used to perform distributed graph processing on the multiple first time-series sub-graphs based on the first distributed graph processing algorithm to obtain a graph processing result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Citation Information

Patent Citations

  • Distributed storage and visual query processing method oriented to large-scale financial knowledge graph

    CN113157943A

  • Large-scale distributed graph calculation end-to-end acceleration method and device

    CN113326125A

  • Graph data segmentation method and device, computer equipment and storage medium

    CN117556095A