Evolution sequence diagram-oriented stable unit discovery and evaluation method and device

Abstract

The invention discloses a stable unit discovery and evaluation method and device oriented to an evolution sequence diagram, and the method comprises the steps: firstly, after a complex network with time information in the real world is given, a collection time window needs to be set, and all related node information is obtained; secondly, setting the time period length of the time slices, reading the edge connection information in each time slice, and representing the time sequence diagram as a multi-layer time sequence network in a multi-dimensional tensor form; then, calculating a stability index of each node and a connecting edge in the time sequence diagram, which is obtained by combining an on-chip importance measurement and an inter-chip continuity measurement, so as to obtain key components in the time sequence diagram; and finally, designing a performance measurement index, namely time robustness measurement, of the time sequence diagram, and evaluating the importance of a stable unit according to the performance measurement index. According to the method, the in-chip structure importance measurement of the connected edges (or nodes) in a single snapshot and the inter-chip dynamic continuity measurement in the time sequence diagram evolution process can be effectively combined to realize the connected edge importance sorting which is more consistent with the practical significance.

Description

Technical Field

[0001] The present invention relates to the technical field of important edge sorting analysis of complex time series networks, and in particular to a stable unit discovery and evaluation method and device for evolving time series graphs. Background Art

[0002] Temporal networks are networks where nodes or edges appear and disappear intermittently over time. Compared with static networks, temporal networks can more reasonably and effectively characterize the dynamic process of real networks. For example, in a mobile phone communication network, the edge between two people only exists when they are talking, and the edge disappears after the call ends. Currently, we observe that people are increasingly interested in understanding how real-world network systems behave. It is challenging to evaluate the dynamics of these real-world network complex systems (i.e., the changes of nodes or edges over time) because these systems usually present a large number of elements with complex interactions. In addition, these complex systems may evolve rapidly. Temporal graphs, also known as time-varying graphs or time-dependent networks in the literature, make it possible to study and understand the dynamics of many real-world network complex systems. In fact, temporal graphs allow people to represent and better understand the dynamics of complex networks. Centrality is a key concept in the analysis scenario of complex temporal networks.

[0003] Detecting stable units in temporal networks plays an important role in gaining insights into their influence and functionality. The impact of a single node or edge on the network structure depends not only on its structural centrality at the snapshot it is in, but also on its temporal importance over the time dimension. Understanding these interrelationships remains one of the key challenges in network science.

[0004] A Chinese patent with announcement number CN107292751A (a method and device for mining the importance of nodes in a time series network) discloses a method and device for mining the importance of nodes in a time series network. First, in this method, a time series network is first constructed according to the time relationship of the interaction between N nodes, and then the time series network is sliced ​​according to the preset time window size to obtain a slice network sequence, which includes: m slice networks corresponding to m time windows, and then a weight value is calculated for all the edges in each slice network, and the weight values ​​corresponding to all the edges in each slice network are accumulated according to different edges to obtain the weight accumulation value of the edges between the node pairs in the time series network, and the weight accumulation value of the edges between each node pair is updated to the time series network to obtain a weighted network, and finally the importance index of each node in the weighted network is calculated through the edges including the weight accumulation value, and the importance index of each node includes: the sum of the weight accumulation values ​​corresponding to all the edges of the corresponding node. Although this invention utilizes the weight values ​​of the k-core decomposition within the slice and uses it as the intra-slice structural measurement for edge sorting in the temporal network, it lacks relevant considerations in the time dimension and its exploration of the temporal network is relatively rudimentary.

[0005] The Chinese patent with announcement number CN118413452A (a method and device for discovering key nodes of Tor based on a time series network) discloses a method and device for discovering key nodes of a Tor time series network. First, after establishing a stable connection with the Tor network through the deployed Tor honeypot relay node, the Tor network topology is collected, and the time information is recorded when the node plays a forwarding role. Then, the connection relationship data Q between the connection time, entry node, relay node and exit node in the collected Tor network is preprocessed to draw a Tor node connection topology diagram, and the connection timestamp is embedded in the edge of the Tor node connection topology diagram, so that each edge carries the specific time series information of the interaction between nodes, thereby constructing a Tor node time series network model. After that, the topology diagram of any day will be selected, M time periods will be set, and the node time series values ​​in the M time periods will be calculated. Finally, the Tor nodes are sorted from large to small according to the time series values, and the sorting rankings are recorded, and the Tor key nodes are identified according to the rankings. Although this invention also utilizes time information to sort nodes in a time series network, its usage is relatively superficial and is not scalable in some other time series networks. Summary of the invention

[0006] The present invention aims to overcome the above-mentioned shortcomings of the prior art and provide a stable unit discovery and evaluation method and device for evolving time series graphs, which realizes a more realistic edge importance ranking by combining the intra-chip structural importance measurement of edges (or nodes) in a single snapshot and the inter-chip dynamic continuity measurement during the evolution of the time series graph. The technical solution of the present invention is as follows:

[0007] A first aspect of the present invention relates to a stable unit discovery and evaluation method for an evolving time series graph, comprising the following steps:

[0008] S1. Timing diagram data collection:

[0009] Given a complex network with time information in the real world, it is necessary to set the time window for collection and obtain information about all nodes involved;

[0010] S2. Construction of timing graph network:

[0011] Set the time period length of the time slice, read the edge information in each time slice, and represent the time series graph as a multi-layer time series network in the form of a multi-dimensional tensor;

[0012] S3, Timing diagram stable unit found:

[0013] Calculate the stability index of each node and edge in the timing graph - obtained by combining the intra-chip importance metric with the inter-chip persistence metric to obtain the key components in the timing graph;

[0014] S4, Timing diagram stable unit evaluation:

[0015] A performance metric for timing diagrams, the temporal robustness metric, is designed to evaluate the importance of stable units.

[0016] Further preferably, the stated step S1 specifically includes:

[0017] Set the time window for method application - specify the start time t min and end time t max , get the All the node information contained in it constitutes the time series graph node set V = {1, 2, ..., N}. If the node contains attribute information, the attribute is attached to it.

[0018] Further preferably, the stated step S2 specifically includes:

[0019] S2.1: Set the snapshot period length Δt, and discretize the continuous time window of the temporal complex network into m snapshots N1, N2, …, N m , the number of snapshots is

[0020] S2.2: Read snapshot time period [t i ,t i +Δt](i=1,2,…,m) constitutes the edge set of this time slice The snapshot network can be represented using the adjacency matrix

[0021] S2.3: Perform the snapshot network construction operation of S2.2 for each snapshot time period in turn, and finally obtain m adjacency matrices, which are arranged in chronological order. If the edge has attributes of a dimensions, the time series graph can be represented by a tensor of dimension (N,N,m,a), that is,

[0022] Further preferably, the stated step S3 specifically includes:

[0023] S3.1: Calculation of intra-chip structural importance metrics:

[0024] At each time slice t i In the network, the betweenness centrality (Betweenness Centrality) BC(e) is calculated. The calculation formula is:

[0025]

[0026] Among them, σ(s,t) represents the total number of shortest paths from node s to node t, and σ(s,t|e) is the number of shortest paths from node s to node t through edge e. Through the adjacency matrix of each time slice, the betweenness centrality BC(e) of each edge (matrix element) is calculated. The betweenness centrality matrix generated by each time slice will be used as a new dimension to finally form a time series graph tensor containing the importance measure of the intra-slice structure, whose dimension is (n,n,t,2), where n is the number of nodes and t is the number of time slices. ;

[0027] S3.2: Inter-slice dynamic continuity metric calculation, including three indicators, namely "total number of occurrences", "number of consecutive occurrences", and "maximum number of consecutive occurrences", where

[0028] Total number of occurrences Tt otal :

[0029] Each edge e i,j The total number of occurrences of represents the number of times the edge appears in all time slices of the timing graph;

[0030] Continuous occurrence times T consecutive :

[0031] Each edge e i,j The number of consecutive occurrences of refers to the situation where it exists simultaneously in two or more consecutive time slices. For example, given the time sequence If edge e exists at t0, t1, t3, and t5, that is, the edge appears only once continuously, so its number of consecutive appearances is 1. The number of consecutive appearances of each edge is calculated in this way;

[0032] Maximum number of consecutive occurrences T max :

[0033] Edge i,j The maximum number of consecutive occurrences of refers to the longest duration of its occurrence in multiple consecutive time slices. For example, the edge e exists at t0, t1, t2, t4, and t5, its consecutive occurrence number is 2, and the maximum consecutive occurrence number is 3;

[0034] S3.3 Standardization and weighted combination:

[0035] In order to comprehensively consider the intra-slice structural importance measure (betweenness centrality) and the inter-slice dynamic continuity measure (total number of occurrences, consecutive number of occurrences, and maximum consecutive number of occurrences), each measure needs to be standardized to unify its dimensions. The standardization method uses Min-Max standardization to scale the value of each measure to the range of 0 to 1. The formula is as follows:

[0036]

[0037] For the betweenness centrality BC(e), the total number of occurrences T total , consecutive occurrence times T consecutive and the maximum number of consecutive occurrences T max Standardize separately.

[0038] Next, these standardized metrics are combined using a weighted combination method to form a comprehensive stability score S(e). The comprehensive stability score formula is:

[0039] S(e)=α·BC norm (e)+

[0040] β·T total,norm (e)+γ·T consecutive,norm (e)+δ·T max,norm (e) (3.3)

[0041] Among them, α, β, γ, and δ are adjustable weight parameters used to control the influence of each metric in the overall stability score. The specific weights can be adjusted according to the actual application scenario to ensure that the influence of intra-chip structure and inter-chip continuity is balanced;

[0042] S3.4 Stable unit sorting:

[0043] Use the comprehensive stability scoring formula mentioned in S4.3 to calculate the score of each edge in turn, and sort them in descending order according to the score. The higher the score, the more stable and important it is in the timing diagram, and it is called a stable unit;

[0044] Further preferably, the stated step S4 specifically includes:

[0045] S4.1: Define a temporal distance metric to measure the number of time steps d required for information to propagate between two nodes within a specific time window. AB =t d -t1, where t1 is the departure time from node A, t d is the time to reach node B;

[0046] S4.2: Calculate the time efficiency as the average of the inverse of the time distance between all node pairs, and its formula is

[0047]

[0048] S4.3: After the network suffers some damage D, the time efficiency is recalculated, and the time robustness metric R is defined G (D), as the relative loss of time efficiency before and after the damage

[0049]

[0050] in, It is the time efficiency difference before and after the damage. The time robustness value is between 0 and 1, where 1 means the network is not affected and 0 means the network is completely ineffective.

[0051] S4.4: Select the corresponding stable unit to delete as the damage D and calculate the time robustness metric R G (D), Evaluate the value of the selected stable unit.

[0052] The second aspect of the present invention relates to a stable unit discovery and evaluation device for evolutionary timing graphs, comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement a stable unit discovery and evaluation method for evolutionary timing graphs of the present invention.

[0053] A third aspect of the present invention relates to a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements a stable unit discovery and evaluation method for an evolving timing diagram of the present invention.

[0054] Compared with the prior art, the advantages of the present invention are:

[0055] Based on the premise of focusing on the intra-chip structural characteristics in the network, the present invention considers the frequency of occurrence, continuous frequency of occurrence and maximum duration of the edge in the temporal evolution process, and integrates the temporal characteristics in the temporal evolution process into the edge importance evaluation system, so that the stability index of each edge can be considered more in the time dimension, so as to distinguish and obtain stable units. In addition, a performance measurement index based on the timing graph, the temporal robustness measurement, is designed to better measure the importance of stable units in the timing graph. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings are used to provide an intuitive description of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0057] Figure 1 It is an algorithm flow chart of the method of the present invention;

[0058] Figure 2 This is a schematic timing diagram of a file transfer link between 10 servers in a computer room according to Embodiment 1 of the present invention, and includes sub-diagrams at five moments;

[0059] Figure 3 These are three stable edges found in the timing diagram of the file transfer link in the computer room according to the first embodiment of the present invention, and are marked with a dotted line, a dashed line, and a dash-dot line respectively;

[0060] Figure 4 It is the corresponding timing robustness metric value calculated by using the evaluation function after individually deleting the corresponding edge in the computer room file transmission link timing diagram of embodiment 2 of the present invention.

[0061] Figure 5 It is a structural diagram of the device of the present invention. DETAILED DESCRIPTION

[0062] The embodiments are used to explain the purpose, technical methods and advantages of the present invention in more detail. The following embodiments are used to explain the present invention together with the embodiments of the present invention. The specific embodiments described here are only used to explain the present invention and do not constitute a limitation of the present invention.

[0063] Example 1

[0064] The present invention provides a communication network key node identification method based on a stable unit discovery and evaluation technology for an evolutionary time sequence graph, see Figure 1 , Figure 2 and Figure 3 , the steps are:

[0065] S1. Timing graph data collection: Given a complex network with time information in the real world, it is necessary to set the collection time window and obtain all the node information involved;

[0066] This embodiment adopts a real network, taking the file transfer communication in the computer room as an example. The task discovered by the stability unit is to discover and evaluate the communication value of each host in the communication network, so as to identify the key hosts, which belongs to the key node identification task. The specific information of the network is shown in Table 1. Table 1 Real network data set information and experimental results

[0067]

[0068] The time window for collection is set for the complex network in Table 1, and the information of all involved nodes is obtained; the specific steps are: Δt = 2h.

[0069] S1.1. Set the time window for method application - specify the start time t min = 00:00 and end time t max = 22:00, get the time window All node information contained in it constitutes the time series graph node set V = {0, 1, ..., 9}. If a node does not contain attribute information, it does not need to be attached to it;

[0070] S2. Construction of the time series graph network: Set the time period length of the time slice, read the edge information in each time slice, and represent the time series graph as a multi-layer time series network in the form of a multi-dimensional tensor. The specific steps are:

[0071] S2.1. Set the snapshot time period Δt = 2h, and discretize the continuous time window of the temporal complex network into m snapshots N1, N2, ..., N m , the number of snapshots is

[0072] S2.2, read snapshot time period [t,t i +02:00](i=1,2,…,m) constitutes the edge set of this time slice The snapshot network can be represented using the adjacency matrix

[0073] S2.3, perform the snapshot network construction operation of S2.2 for each snapshot time period in turn, and finally obtain m = 5 adjacency matrices, which are arranged in chronological order. Finally, the time series graph can be represented by a tensor with a dimension of (10, 10, 5, 1), that is,

[0074] S3, Timing Graph Stability Unit Discovery: Calculate the stability index of each node and edge in the timing graph - obtained by combining the intra-chip importance metric with the inter-chip persistence metric, so as to obtain the key components in the national AS network. The specific steps are:

[0075] S3.1: Calculation of intra-chip structural importance metrics:

[0076] At each time slice t i In the network, the betweenness centrality (Betweenness Centrality) BC(e) is calculated. The calculation formula is:

[0077]

[0078] Among them, σ(s,t) represents the total number of shortest paths from node s to node t, and σ(s,t|e) is the number of shortest paths from node s to node t through edge e. Through the adjacency matrix of each time slice, the betweenness centrality BC(e) of each edge (matrix element) is calculated. The betweenness centrality matrix generated by each time slice will be used as a new dimension to finally form a time series graph tensor containing the importance measure of the intra-slice structure, whose dimension is (n,n,t,2), where n is the number of nodes and t is the number of time slices. ;

[0079] S3.2: Inter-slice dynamic continuity metric calculation, including three indicators, namely "total number of occurrences", "number of consecutive occurrences", and "maximum number of consecutive occurrences", where

[0080] Total number of occurrences T total :

[0081] Each edge e i,j The total number of occurrences of represents the number of times the edge appears in all time slices of the timing graph;

[0082] Continuous occurrence times T consecutive :

[0083] Each edge e i,j The number of consecutive occurrences of refers to the situation where it exists simultaneously in two or more consecutive time slices. For example, given the time sequence If edge e exists at t0, t1, t3, and t5, that is, the edge appears only once continuously, so its number of consecutive appearances is 1. The number of consecutive appearances of each edge is calculated in this way;

[0084] Maximum number of consecutive occurrences T max :

[0085] Edge i,jThe maximum number of consecutive occurrences of refers to the longest duration of its occurrence in multiple consecutive time slices. For example, the edge e exists at t0, t1, t2, t4, and t5, its consecutive occurrence number is 2, and the maximum consecutive occurrence number is 3;

[0086] S3.3 Standardization and weighted combination:

[0087] In order to comprehensively consider the intra-slice structural importance measure (betweenness centrality) and the inter-slice dynamic continuity measure (total number of occurrences, consecutive number of occurrences, and maximum consecutive number of occurrences), each measure needs to be standardized to unify its dimensions. The standardization method uses Min-Max standardization to scale the value of each measure to the range of 0 to 1. The formula is as follows:

[0088]

[0089] For the betweenness centrality BC(e), the total number of occurrences T total , consecutive occurrence times T consecutive and the maximum number of consecutive occurrences T max Standardize separately.

[0090] Next, these standardized metrics are combined using a weighted combination method to form a comprehensive stability score S(e). The comprehensive stability score formula is:

[0091] S(e)=α·BC norm (e)+β·T total,norm (e)+γ·T consecutive,norm (e)+δ·T max,norm (e) (3.3)

[0093] The corresponding weights are set to α=0.5, β=0.3, γ=0.2, δ=0.1, and finally the top three edges are selected, with corresponding scores of ((1,5), 0.8), ((3,9), 0.716), ((6,7), 0.692). The selected edges are as follows: Figure 3 As indicated by the logo;

[0094] S4. Timing diagram stable unit evaluation: Use the "timing robustness metric" of the timing diagram to evaluate the importance of the stable unit in the timing diagram. The steps are:

[0095] S4.1: Calculate the time efficiency as the average of the inverse of the time distance between all node pairs, and its formula is

[0096]

[0097] S4.2: After the network suffers some damage D, the time efficiency is recalculated, and the time robustness metric R is definedG (D), as the relative loss of time efficiency before and after the damage

[0098]

[0099] in, It is the time efficiency difference before and after the damage. The time robustness value is between 0 and 1, where 1 means the network is not affected and 0 means the network is completely ineffective.

[0100] S4.4: Select the corresponding stable unit to delete as the damage D and calculate the time robustness metric R G (D), evaluate the value of the selected stable unit. The results are as follows Figure 4 shown.

[0101] Step S5: input a new communication network to realize key host identification;

[0102] When applied to different other types of communication networks, the corresponding graph information and the set time period are applied to the method, and hosts with high timing importance in the communication network can be obtained, thereby identifying high-risk and high-importance server hosts.

[0103] This technology dynamically monitors and analyzes the communication links between servers to identify the links that play a key role in the communication network in different time periods. The identification of key links helps operation and maintenance personnel accurately locate potential communication bottlenecks or single points of failure, ensuring efficient and stable communication between servers. Especially in environments with high concurrency and massive data transmission, key links can be used as reference points for operation and maintenance optimization, and the network's fault tolerance can be effectively improved by reinforcing or diverting traffic on these key links. In addition, this method can also reveal the evolution of the communication structure between servers over time, providing data support for the dynamic optimization of the network topology. This not only improves the overall stability of the network, but also effectively reduces the risk of network interruption.

[0104] Example 2

[0105] like Figure 5 This embodiment relates to a stable unit discovery and evaluation device for an evolutionary timing diagram, including a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement a stable unit discovery and evaluation method for an evolutionary timing diagram of embodiment 1.

[0106] Example 3

[0107] This embodiment relates to a computer-readable storage medium on which a program is stored. When the program is executed by a processor, a stable unit discovery and evaluation method for an evolving timing diagram of embodiment 1 is implemented.

[0108] The above description is only an embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent substitution, improvement, etc. made within the concept and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A stable unit discovery and evaluation method for evolving time series graphs, characterized in that: The following steps are involved: S1. Timing graph data collection: Given a complex network with time information in the real world, it is necessary to set the collection time window and obtain all the node information involved; S2. Construction of the time series graph network: set the time period length of the time slice, read the edge information in each time slice, and represent the time series graph as a multi-layer time series network in the form of a multi-dimensional tensor; S3, Timing Graph Stability Unit Discovery: Calculate the stability index of each node and edge in the timing graph, which is obtained by combining the intra-chip importance metric with the inter-chip persistence metric to obtain the key components in the timing graph; S4. Timing diagram stable unit evaluation: Design a performance measurement indicator for timing diagrams - temporal robustness measurement, and use it to evaluate the importance of stable units.

2. The stable unit discovery and evaluation method for evolving time series graph according to claim 1, characterized in that: In step S1, given a complex network with time information in the real world, it is necessary to set the time window for collection and obtain all the node information involved, including: Set the time window for method application - specify the start time t min and end time t max , get the All the node information contained in it constitutes a time-series graph node set V = {1, 2, ..., N}. If the node contains attribute information, the attribute is attached to it.

3. The stable unit discovery and evaluation method for evolving time series graph according to claim 1, characterized in that: In step S2, the time period length of the time slice is set, the edge information in each time slice is read, and the time sequence graph is represented as a multi-layer time sequence network in the form of a multi-dimensional tensor, which specifically includes: S2.1: Set the snapshot period length Δt, and discretize the continuous time window of the temporal complex network into m snapshots N1, N2, ..., N m , the number of snapshots is S2.2: Read snapshot time period [t i , t i +Δt](i=1,2,…,m) constitutes the edge set of this time slice The snapshot network can be represented using the adjacency matrix S2.3: Perform the snapshot network construction operation of S2.2 for each snapshot time period in turn, and finally obtain m adjacency matrices, which are arranged in chronological order. If the edge has attributes of a dimensions, the time series graph can be represented by a tensor of dimension (N, N, m, a), that is, 4. The stable unit discovery and evaluation method for evolving time series graph according to claim 1, characterized in that: In step S3, the stability index of each node and edge in the timing graph is calculated, which is obtained by combining the intra-chip importance metric with the inter-chip continuity metric, so as to obtain the key components in the national AS network, including: S3.1: Calculate the importance metric of the intra-chip structure: At each time slice t i In the network, the betweenness centrality (Betweenness Centrality) BC(e) is calculated. The calculation formula is: Among them, σ(s, t) represents the total number of shortest paths from node s to node t, and σ(s, t|e) is the number of shortest paths from node s to node t through edge e. Through the adjacency matrix of each time slice, the betweenness centrality BC(e) of each edge (matrix element) is calculated. The betweenness centrality matrix generated by each time slice will be used as a new dimension to finally form a time series graph tensor containing the importance measure of the intra-slice structure, whose dimension is (n, n, t, 2), where n is the number of nodes and t is the number of time slices. ; S3.2: Calculate the inter-slice dynamic continuity metric, including three indicators, namely "total number of occurrences", "number of consecutive occurrences", and "maximum number of consecutive occurrences", where Total number of occurrences T total : Each edge e i,j The total number of occurrences of represents the number of times the edge appears in all time slices of the timing graph; Continuous occurrence times T consecutive : Each edge e i,j The number of consecutive occurrences of refers to the situation where it exists simultaneously in two or more consecutive time slices. For example, given the time sequence If edge e exists at t0, t1, t3, and t5, that is, the edge appears only once continuously, so its number of consecutive appearances is 1. The number of consecutive appearances of each edge is calculated in this way; Maximum number of consecutive occurrences T max : Edge i，j The maximum number of consecutive occurrences of refers to the longest duration of its occurrence in multiple consecutive time slices. For example, edge e exists at t0, t1, t2, t4, and t5, its consecutive occurrence number is 2, and its maximum consecutive occurrence number is 3; S3.3 Standardization and weighted combination: In order to comprehensively consider the intra-slice structural importance measure (betweenness centrality) and the inter-slice dynamic continuity measure (total number of occurrences, consecutive number of occurrences, and maximum consecutive number of occurrences), each measure needs to be standardized to unify its dimensions. The standardization method uses Min-Max standardization to scale the value of each measure to the range of 0 to 1. The formula is as follows: For the betweenness centrality BC(e), the total number of occurrences T total , consecutive occurrence times T consecutive and the maximum number of consecutive occurrences T max Standardize separately. Next, these standardized metrics are combined using a weighted combination method to form a comprehensive stability score S(e). The comprehensive stability score formula is: S(e)=α·BC norm (e)+β·T total，norm (e)+γ·T consecutive，norm (e)+δ·T max，norm (e) (3.3) Among them, α, β, γ, and δ are adjustable weight parameters used to control the influence of each metric in the overall stability score. The specific weights can be adjusted according to the actual application scenario to ensure that the influence of intra-chip structure and inter-chip continuity is balanced; S3.4 Stable unit sorting: Use the comprehensive stability scoring formula mentioned in S3.3 to calculate the score of each edge in turn, and sort them in descending order according to the score size. The higher the score, the more stable and important it is in the timing diagram, and it is called a stable unit.

5. The stable unit discovery and evaluation method for evolving time series graph according to claim 1, characterized in that: In step S4, a "timing robustness metric" of the timing diagram in the time window [t1, t2] is customized to evaluate the importance of the stable unit in the timing diagram, specifically including: S4.1: Define a temporal distance metric to measure the number of time steps d required for information to propagate between two nodes within a specific time window. AB =t d -t1, where t1 is the departure time from node A, t d is the time to reach node B; S4.2: Calculate the time efficiency as the average of the inverse of the time distance between all node pairs, and its formula is S4.3: After the network suffers some damage D, the time efficiency is recalculated, and the time robustness metric R is defined G (D), as the relative loss of time efficiency before and after the damage in, It is the time efficiency difference before and after the damage. The time robustness value is between 0 and 1, where 1 means the network is not affected and 0 means the network is completely ineffective. S4.4: Select the corresponding stable unit to delete as the damage D and calculate the time robustness metric R G (D), Evaluate the value of the selected stable unit.

6. A stable unit discovery and evaluation device for evolving time series graphs, characterized in that: It comprises a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, it is used to implement a stable unit discovery and evaluation method for an evolutionary timing diagram according to any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, a stable unit discovery and evaluation method for an evolving timing diagram according to any one of claims 1 to 5 is implemented.

Images