An iterative graph structure data alignment method

Abstract

The present invention discloses an iterative graph structure data alignment method, which relates to the technical field related to automatic alignment of graph structure data. The specific steps are: obtaining a topological map, obtaining an initial alignment relationship between any two topological maps to be aligned; calculating the context of any node in any two topological maps to be aligned with respect to the initial alignment relationship; obtaining the similarity between any pair of nodes of the same type in any two topological maps to be aligned; calculating the average similarity of nodes of any type in any two topological maps to be aligned; obtaining the alignment relationship between any two nodes of any type in the topological maps to be aligned; forming the alignment relationship between any two topological maps to be aligned; judging whether to continue iterative calculation; the application explicitly considers the type information of the nodes in the topological map and the relationship between the nodes and other nodes in the topological map, thereby improving the accuracy of topological map alignment.

Description

Technical Field

[0001] The present invention relates to the technical field related to automatic alignment of graph structured data, and more particularly to an iterative graph structured data alignment method. Background Art

[0002] With the continuous advancement of science and technology, some technologies have also begun to be widely distributed in many fields, and graph structured data is one of them. Graph structured data is typically used in key point topological relationship graphs in face images, social network graphs, biological protein interaction graphs, etc., to determine the connection between the two. The key point topological relationship graph in face images is closest to people's lives. Therefore, taking the key point topological relationship graph in face images as an example: given two key point topological relationship graphs of face images, the distance between the two face images can be quantitatively calculated by graph alignment, and then it can be determined whether the two face images are from the same person. The specific steps are: node pair similarity calculation: using specific technical means, the similarity between all node pairs from the left or right image is calculated; node alignment: using the similarity information between node pairs, a greedy algorithm or a global optimal algorithm is used to form an alignment relationship between the nodes in the left and right images; however, the existing technical solutions do not explicitly consider the type information of the nodes in the graph and the relationship between the nodes and other nodes in the graph, which may lead to incorrect alignment relationships; therefore, a method for aligning graph structured data with higher accuracy is needed, which is a problem that technicians in this field urgently need to solve. Summary of the Invention

[0003] In view of this, the present invention provides an iterative graph structure data alignment method, which explicitly considers the type information of nodes in the topology graph and the relationship between the nodes and other nodes in the topology graph, thereby improving the accuracy of topology graph alignment.

[0004] In order to achieve the above object, the present invention provides the following technical solutions:

[0005] An iterative graph structure data alignment method, the specific steps are as follows:

[0006] Get the topology map;

[0007] Obtaining an alignment relationship between any two topological maps to be aligned, which is defined as an initial alignment relationship; the any two topological maps to be aligned include a first topological map to be aligned and a second topological map to be aligned;

[0008] Calculate the context of any node in any two topological graphs to be aligned with respect to the initial alignment relationship based on the two topological graphs to be aligned and the initial alignment relationship;

[0009] Based on the context of any node with respect to the initial alignment relationship, the similarity between any pair of nodes of the same type in any two topological graphs to be aligned is calculated;

[0010] Based on the similarity between any pair of nodes of the same type in any two topological graphs to be aligned, the average similarity of nodes of any type in any two topological graphs to be aligned is calculated;

[0011] According to the relationship between the average similarity of nodes of any type and the similarity between any pair of nodes of the same type in any type, the alignment relationship between nodes of any type in any two topological graphs to be aligned is obtained;

[0012] Put the alignment relationships between nodes of various types in any two topological graphs to be aligned into a set to form the alignment relationship between any two topological graphs to be aligned;

[0013] Determine whether to continue iterative calculation based on the set iteration conditions.

[0014] Preferably, the topology diagram is represented as:

[0015] (1);

[0016] Where, TV is a node type collection; TE is the edge type set; V is a set of nodes; E is the edge set; type It is a mapping function that maps each node to a node type and each edge to an edge type.

[0017] Its advantage is that it combines the node type and the edge type to establish a directed graph, which provides the necessary elements for image alignment and improves the success rate of image alignment.

[0018] Preferably, the calculation method of the initial alignment relationship is:

[0019] S21. Calculate the type context of any node in any two topological graphs to be aligned. The type context is a vector, specifically:

[0020] The edge type set of any two topological graphs to be aligned TE A collection of all elements and node types TV Combine all the elements in , sort all the elements formed by the combination, and set a number for each element combination;

[0021] node u Any node on any topological graph has its context recorded as ctx _ type ( u ), ctx _ type (u )The length of the vector is , ctx _ type ( u ) vector includes nodes u The type context of the outgoing edge and the type context of the incoming edge; number i The element pairs corresponding to the components of , te is a set of edge types TE The elements in tv A collection of node types TV The elements in represent nodes u Among all the outgoing / incoming edges, those that satisfy the "edge type is te , the other end node type of the edge is tv The total number of outgoing / incoming edges of ”;

[0022] S22. Taking the distance between the type contexts of any pair of nodes of the same type as the distance between any pair of nodes of the same type;

[0023] S23. Using a bipartite graph maximum weight allocation algorithm, obtain an alignment relationship between nodes of each type in any two topological graphs to be aligned; merge the alignment relationships of nodes of each type to form an initial alignment relationship between any two topological graphs to be aligned, and record it as an initial alignment relationship set A.

[0024] Preferably, the type context of any node includes two parts. The first part is the type context of the outgoing edge of any node, and the corresponding numbered components are the total number of edges in the outgoing edge of any node whose other endpoint is the node type of the corresponding element combination and whose edge type is the edge type in the corresponding element combination; the second part is the type context of the incoming edge of any node, and the corresponding numbered components are the total number of edges in the incoming edge of any node whose other endpoint is the node type of the corresponding element combination and whose edge is the edge type in the corresponding element combination.

[0025] Preferably, the specific steps of calculating the context of any node in any two topological graphs to be aligned with respect to the initial alignment relationship are:

[0026] For any node in any two topological graphs to be aligned, the context of the initial alignment relationship set A is recorded as ctx ( u , Gy , A ), whose context is a collection defined as follows:

[0027] ;

[0028] in, Gy is any topology graph to be aligned;u For any topology to be aligned Gy Nodes in .

[0029] Preferably, the similarity between any pair of nodes of the same type in any two topological graphs to be aligned is calculated as follows:

[0030] S51: Calculate the intersection of the context of the initial alignment relationship set A of any node in the first topology graph to be aligned and the context of the initial alignment relationship set A of any corresponding node of the same type in the second topology graph to be aligned;

[0031] S52: Calculate the union of the context of the initial alignment relationship set A of any node in the first topology graph to be aligned and the context of the initial alignment relationship set A of any corresponding node of the same type in the second topology graph to be aligned;

[0032] S53. The ratio of the intersection to the union is the similarity between any pair of nodes of the same type in any two topological graphs to be aligned.

[0033] Preferably, the average similarity of any type of nodes in any two topological graphs to be aligned is calculated as follows:

[0034] S61. Obtain a similarity sum: a sum of similarities of any type of node pairs in the first topology graph to be aligned and the second topology graph to be aligned;

[0035] S62, obtaining a node product: the product of the number of nodes of any type in the first topological graph to be aligned and the number of nodes of any type in the second topological graph to be aligned;

[0036] S63. The ratio of the sum of similarities to the product of the nodes is the average similarity of any type of nodes in any two topological graphs to be aligned.

[0037] Preferably, the alignment relationship between nodes of various types in any two topological graphs to be aligned is calculated, and the calculation process is as follows:

[0038] S71, respectively obtaining the number of nodes of any type in the first topology graph to be aligned and the second topology graph to be aligned;

[0039] S72: Determine whether the number of nodes of any type in the first topology to be aligned is greater than the number of nodes of any type in the second topology to be aligned. If so, add a corresponding number of virtual points of the same type to the second topology to be aligned; if not, add a corresponding number of virtual points of the same type to the first topology to be aligned; ensure that the number of nodes of any type in the first topology to be aligned is the same as the number of nodes of any type in the second topology to be aligned.

[0040] S73, setting the similarity between the virtual point supplemented in any topological graph to be aligned and any type of node in the second topological graph to be aligned as the average similarity of any type of node;

[0041] S74. Obtain an initial alignment relationship using a bipartite graph maximum weight allocation algorithm;

[0042] S75. In the initial alignment relationship, replace the virtual node appearing in each element with the zero element 0;

[0043] S76. Determine whether the similarity of any type of node pair corresponding to each element not containing 0 is less than the average similarity of any type of node. If so, replace the element not containing 0 with two elements each containing one zero element. After the replacement is completed, the alignment relationship of any type of node is obtained.

[0044] Preferably, the iteration condition is a set number of iterations or a difference value of the alignment relationship, and the difference value of the alignment relationship is obtained by the alignment relationship of any type of node and the alignment relationship of any two topological graphs to be aligned.

[0045] Preferably, whether iterative calculation is required is determined based on the set iteration conditions. If iterative calculation is required, the alignment relationship between any two topological graphs to be aligned is defined as the initial alignment relationship, and the aligned node pairs whose similarity is less than the average similarity are split, and the calculation is repeated until the iteration conditions are met.

[0046] It can be seen from the above technical solution that compared with the prior art, the present invention discloses an iterative graph structure data alignment method, which explicitly considers the type information of the nodes in the topology graph and the relationship between the nodes and other nodes in the topology graph, thereby improving the accuracy of topology graph alignment. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0048] Figure 1 Schematic diagram of the method flow of the present invention;

[0049] Figure 2 (a) shows the image numbered 1 in the CMU House Image Sequence dataset and the 30 key points it contains; Figure 2 (b) shows the image numbered 51 in the CMU House Image Sequence dataset and the 30 key points it contains;

[0050] FIG3( a ) is a schematic diagram showing the changes in accuracy of the FRGM-E method and the method of the present invention; FIG3( b ) is a schematic diagram showing the running time of the FRGM-E method and the method of the present invention. DETAILED DESCRIPTION

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. Example 1

[0052] The embodiment of the present invention discloses an iterative graph structure data alignment method, which uses a method of pairing two graphs to complete the alignment among a plurality of graphs to be aligned, wherein the graph is a non-topological graph, and the specific steps are as follows: Figure 1 As shown:

[0053] Step 1: Represent the pair of graphs to be aligned as a directed graph with nodes and edges having types. That is, define a node type set TV and a set of edge types TE ; V Set the type of each node in the graph to a collection TV Elements in E Set the type of each edge in the graph to a set TE For each edge in the graph, determine its source node and target node (i.e., this is an edge starting from the source node and pointing to the target node).

[0054] Any directed graph of node and edge types can be formally represented as ,in, TV and TE Represent the node type set and edge type set respectively; V Represents a collection of nodes; E Represents a set of edges, each edge is a triple Represents a slave node a Pointing to a node b , type is type edge; type It is a function that maps any node or edge to the corresponding type.

[0055] This graph representation is compatible with any graph where nodes / edges are untyped and edges are undirected: if a node is untyped, the types of all nodes are set to the same value; if an edge is untyped, the types of all edges are set to the same value; if an edge is undirected, it is replaced by two edges with opposite directions.

[0056] Step 2: Use any other graph alignment algorithm to obtain a pair of graphs to be aligned. G 1 and Figure G 2 is the topological diagram; G 1=( TV , TE , V 1, E , type 1) G 2=( TV , TE , V 2, E , type 2) An alignment relationship. This alignment relationship can be represented as a set A, in which each element is a tuple, represented by < u , x > indicates that. Among them, u or V An element of 1, or a zero element 0; x or V An element in 2, or a zero element 0; and u and x Cannot be 0 at the same time; < u , x > means: G Node 1 u Align to image G Node in 2 x On; if x If it is 0, it means that the graph G Node in 1 u Not aligned to the image G On any node in 2; if u If it is 0, it means that the graph G Node in 2 x Not aligned to the image G 1 on any node.

[0057] Step 3: Computational Graph G 1 and G The context of each node in 2 on the alignment A ctx For Figure G One node in 1 u , its alignment context on A is a set, denoted as ctx( u , G 1 , A ), which is defined as follows:

[0058] ;

[0059] For the graph G One node in 2 x , its alignment context on A ctx ( u , G 2 , A ), can also be defined in the same way.

[0060] Step 4: Calculate the values ​​from the graph G 1 and Figure G 2. For a pair of nodes of the same type < u , x >, and its similarity is recorded as sim ( u , x ), calculated as follows:

[0061] .

[0062] In another embodiment, the same method as above is used, according to FIG. G 1 and Figure G 2. The similarity is calculated based on the type context of a pair of nodes of the same type.

[0063] Step 5: Computational Graph G 1 and Figure G The average similarity of each type of node in 2. For node type tv , and the average similarity of its nodes is recorded as mean _ sim ( tv ), calculated as follows:

[0064] mean _ sim ( tv )= G 1 and G All 2 tv The sum of similarities of node pairs of type ÷ ( G 1 in tv The number of nodes of type × G 2 in tv The number of nodes of the type

[0065] Step 6: Computational Graph G 1 and G 2. New alignment relationship for each type of node.tv , and its new alignment relationship is recorded as A '( tv ), the calculation process is as follows:

[0066] Assumptions G 1 in tv The number of nodes of type is less than G 2 in tv The number of nodes of type G 1 Add some types tv virtual nodes, so that the two graphs tv The number of nodes of the same type is equal. G 1 each tv Type virtual node and G Each of the 2 tv The similarity of nodes of type mean _ sim ( tv );

[0067] Use the bipartite graph maximum weight allocation algorithm to calculate an initial alignment relationship A '_ tmp ( tv ), where the two node sets of the bipartite graph are respectively G 1 in all tv A collection of nodes of the same type, consisting of G All 2 tv A set of nodes of type . In the edge set of a bipartite graph, each pair of nodes in the two node sets u , x There will be an edge between them, and the weight on the edge is sim ( u , x );

[0068] exist A '_ tmp ( tv ), replace the virtual nodes in each element with zero elements 0; then, for each element that does not contain 0 ( u , x ),if( u , x ) is less than mean _ sim ( tv ), then the element ( u , x ) is replaced by two elements ( u ,0) and (0, x ); Finally, the alignment relationship is obtained A '( tv ).

[0069] Step 7: Forming the Graph G 1 and G Alignment relationship between 2 nodes A '. Each node type tv Alignment A '( tv ) into a new set, forming an alignment relationship A ''.

[0070] Step 8: Determine whether to continue iterative calculation. You can set a suitable iteration termination condition according to the current situation (for example: when A 'and A '' is less than a certain value, or the number of iterations is greater than a certain value, the iteration termination condition is true). If the iteration termination condition is false, A The value of '' is updated to A ', then jump to step 3; otherwise, the algorithm terminates, A ' is the output of the algorithm.

[0071] If no suitable graph alignment method is found to obtain an initial graph alignment result, the following process can be used to form an initial graph alignment result.

[0072] Step 1: Computational Graph G 1 and G The type context of each node in 2 is a vector. TE and TV Combine all the elements in and sort all the elements formed by the combination; in this way, each element pair is assigned a number. G A node in 1 u , whose type context is recorded as ctx _ type ( u ), ctx _ type ( u )The length of the vector is , the first half of the vector records u The type context of the outgoing edge. For the component numbered i, assume that its corresponding TE 、 TV The element pair is < te , tv >, then the number is i The component of stores a non-negative positive value, indicating that the node u Among all outgoing edges, those that satisfy the “edge type is te , the other end node type of the edge is tvThe total number of edges. ctx _ type ( u )The second half of the vector records u The type context of the incoming edge, where the value of each component has a similar meaning to the outgoing edge type context component.

[0073] Step 2: Calculation G 1 and G 2. For pairs of nodes of the same type ( u , x ), whose type contexts are ctx _ type ( u )and ctx _ type ( x ), then the distance between these two vectors can be used as the node pair < u , x > distance.

[0074] Step 3: Use the bipartite graph maximum weight allocation algorithm to obtain the graph G 1 and G 2. Then, the alignment relationships of all types of nodes are merged to form G 1 and G An initial alignment relationship between 2.

[0075] To quantitatively evaluate the effectiveness of the present invention compared to existing techniques, we selected the CMU House Image Sequence dataset, a publicly available test dataset in the field of image alignment, and compared the present invention with FRGM-E, the currently most effective method in this field. The CMU House Image Sequence dataset contains 111 house frame images. These 111 images are generated by rotating the first image sequentially by a fixed angle. Each image contains 30 manually identified key points. These 30 points are triangulated to form a topological map corresponding to each image. The 111 images are numbered starting from 1. The greater the difference between the image numbers, the greater the difference in the topological maps corresponding to the two images. Figures 2(a) and 2(b) show two images numbered 1 and 51 in the CMU House Image Sequence dataset and the 30 key points they contain.

[0076] The experimental results are shown in Figures 3(a) and 3(b). The yellow curve in each figure illustrates the performance of the proposed method, while the gray curve illustrates the performance of the FEGM-E method. Figure 3(a) shows how the accuracy of the two methods changes with increasing topology dissimilarity (the difference between the two topology map numbers). Figure 3(b) shows how the runtime of the two methods changes with increasing topology dissimilarity (the difference between the two topology map numbers). As can be seen, as the topology dissimilarity increases, the accuracy of the proposed method remains slightly higher than that of the FEGM-E method, while significantly reducing runtime. For example, when the difference between the two topology map numbers is 100, the FEGM-E method takes an average of 0.5 seconds to output an alignment result, while the proposed method only takes an average of 0.05 seconds. Example 2

[0077] In order to quantitatively evaluate the effectiveness of the present invention and the existing technology, four public test datasets in the field of social network alignment, Douban, Flickr, Lastfm, and Myspace, were selected, and the present invention was compared with the currently best method in this field, FINAL.

[0078] Using these four datasets, we conducted three social network alignment experiments:

[0079] 1. Aligning Douban's online and offline social networks. We constructed an offline network based on users' co-occurrence in social gatherings. If two users attended the same offline event more than ten times but less than twenty times, a "contact" edge was established between them. If two users attended more than twenty social gatherings at the same time, a "friend" edge was established between them. Using this approach, we constructed an offline social network consisting of 1,118 users. We also randomly extracted an online social subnetwork consisting of 3,906 users from the online social network containing these users. In the alignment experiment between the two networks, we used node degree similarity to establish an initial alignment, which achieved an accuracy of 7.07%.

[0080] 2. Alignment of the Flickr and Lastfm social networks. The first network contains 12,974 users / nodes, and the second network contains 15,436 users / nodes. We use the user's gender as the node type. Then, we sort the nodes according to their PageRank scores, and label the top 1% of nodes as leaders, the next 10% as middle, and the remaining nodes as followers. The edge type between two nodes is formed by the combination of the two node labels: there are six edge types: leader-leader, leader-middle, leader-follower, middle-middle, middle-follower, and follower-follower. Using username similarity to form an initial alignment relationship between the two social networks, it has an accuracy of 61.50%.

[0081] 3. Alignment of the Flickr and Myspace social networks. The node and edge types in the Myspace social network are identical to those in the Flickr and Lastfm social networks. Similarly, username similarity was used to form the initial alignment between the two social networks, achieving an accuracy of 61.8%.

[0082] The experimental results are shown in Table 1:

[0083] Table 1 Comparison of the accuracy of this application and FINAL

[0084]

[0085] In these three alignment experiments, the method of the present invention achieved higher accuracy than the FINAL method: in the first experiment, the accuracy of the method of the present invention was 20.2% higher than the FINAL method (the running time was 3.3 seconds longer); in the second experiment, the accuracy of the method of the present invention was 1.8% higher than the FINAL method (the running time was reduced by nearly 50%); in the third experiment, the accuracy of the method of the present invention was 2.3% higher than the FINAL method (the running time was increased by 16.8%). It can be seen that the method proposed in this invention has a higher accuracy.

[0086] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other. The above description of the disclosed embodiments enables professionals and technicians in this field to implement or use the present invention. Various modifications to these embodiments will be apparent to professionals and technicians in this field, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An iterative graph structure data alignment method, characterized in that: The specific steps are: Get the topology map; Calculate the alignment relationship between any two topological graphs to be aligned, which is defined as the initial alignment relationship; Based on the initial alignment relationship, the context of any node in any two topological graphs to be aligned with respect to the initial alignment relationship is obtained; Based on the context of any node with respect to the initial alignment relationship, the similarity between any pair of nodes of the same type in any two topological graphs to be aligned is calculated; Based on the similarity between any pair of nodes of the same type in any two topological graphs to be aligned, the average similarity of nodes of any type in any two topological graphs to be aligned is calculated; Obtain the alignment relationship between any two nodes of any type in the topological graphs to be aligned based on the relationship between the average similarity of any type of nodes and the similarity between any pair of nodes of the same type in any type; Put the alignment relationships between nodes of various types in any two topological graphs to be aligned into a set to form the alignment relationship between any two topological graphs to be aligned; Determine whether to continue iterative calculation based on the set iteration conditions.

2. The iterative graph structure data alignment method according to claim 1, characterized in that: The topology diagram is represented as: （1）； Where, TV is a node type collection; TE is the edge type set; V is a set of nodes; E is the edge set; type It is a mapping function that maps each node to a node type and each edge to an edge type.

3. The iterative graph structure data alignment method according to claim 1, characterized in that: The calculation method of the initial alignment relationship is: S21. Calculate the type context of any node in any two topological graphs to be aligned, where the type context is a vector. S22. Taking the distance between the type contexts of any pair of nodes of the same type as the distance between any pair of nodes of the same type; S23. Using a bipartite graph maximum weight allocation algorithm, obtain an alignment relationship between nodes of each type in any two topological graphs to be aligned; merge the alignment relationships of nodes of each type to form an initial alignment relationship between any two topological graphs to be aligned, and record it as an initial alignment relationship set A.

4. The iterative graph structure data alignment method according to claim 3, characterized in that: The type context of any node includes two parts. The first part is the type context of the outgoing edge of any node, and the corresponding numbered component is the total number of edges in the outgoing edge of any node whose other endpoint is the node type of the corresponding element combination and the edge type is the edge type in the corresponding element combination; the second part is the type context of the incoming edge of any node, and the corresponding numbered component is the total number of edges in the incoming edge of any node whose other endpoint is the node type of the corresponding element combination and the edge type is the edge type in the corresponding element combination.

5. The iterative graph structure data alignment method according to claim 3, characterized in that: The specific steps for calculating the context of any node in any two topological graphs to be aligned with respect to the initial alignment relationship are: For any node in any two topological graphs to be aligned, align the initial alignment relationship set A The context is recorded as ctx ( u , Gy , A ), whose context is a collection defined as follows: in, Gy is any topological graph to be aligned; u For any topology to be aligned Gy Nodes in .

6. The iterative graph structure data alignment method according to claim 5, characterized in that: Calculate the similarity between any pair of nodes of the same type in any two topological graphs to be aligned, specifically: S51: Calculate the intersection of the context of the initial alignment relationship set A of any node in the first topology graph to be aligned and the context of the initial alignment relationship set A of any corresponding node of the same type in the second topology graph to be aligned; S52: Calculate the union of the context of the initial alignment relationship set A of any node in the first topology graph to be aligned and the context of the initial alignment relationship set A of any corresponding node of the same type in the second topology graph to be aligned; S53. The ratio of the intersection to the union is the similarity between any pair of nodes of the same type in any two topological graphs to be aligned.

7. The iterative graph structure data alignment method according to claim 6, characterized in that: Calculate the average similarity of any type of nodes in any two topological graphs to be aligned. The calculation method is: S61. Obtain a similarity sum: a sum of similarities of any type of node pairs in the first topology graph to be aligned and the second topology graph to be aligned; S62, obtaining a node product: the product of the number of nodes of any type in the first topological graph to be aligned and the number of nodes of any type in the second topological graph to be aligned; S63. The ratio of the sum of similarities to the product of the nodes is the average similarity of any type of nodes in any two topological graphs to be aligned.

8. The iterative graph structure data alignment method according to claim 1, characterized in that: Calculate the alignment relationship between nodes of various types in any two topological graphs to be aligned. The calculation process is as follows: S71, respectively obtaining the number of nodes of any type in the first topology graph to be aligned and the second topology graph to be aligned; S72: Determine whether the number of nodes of any type in the first topology to be aligned is greater than the number of nodes of any type in the second topology to be aligned. If so, add a corresponding number of virtual points of the same type to the second topology to be aligned; if not, add a corresponding number of virtual points of the same type to the first topology to be aligned. Ensure that the number of nodes of any type in the first topology graph to be aligned is the same as the number of nodes of any type in the second topology graph to be aligned; S73, setting the similarity between the virtual point supplemented in any topological graph to be aligned and any type of node in the second topological graph to be aligned as the average similarity of any type of node; S74. Obtain an initial alignment relationship using a bipartite graph maximum weight allocation algorithm; S75. In the initial alignment relationship, replace the virtual node appearing in each element with the zero element 0; S76. Determine whether the similarity of any type of node pair corresponding to each element not containing 0 is less than the average similarity of any type of node. If so, replace the element not containing 0 with two elements each containing one zero element. After the replacement is completed, the alignment relationship of nodes of either type is obtained.

9. The iterative graph structure data alignment method according to claim 1, characterized in that: The iteration condition is the set number of iterations or the difference value of the alignment relationship. The difference value of the alignment relationship is obtained by the alignment relationship of any type of node and the alignment relationship of any two topological graphs to be aligned.

10. The iterative graph structure data alignment method according to claim 1, characterized in that: Based on the set iteration conditions, determine whether iterative calculation is required. If iterative calculation is required, define the alignment relationship of any two topological graphs to be aligned as the initial alignment relationship, and split the aligned node pairs with similarity less than the average similarity, and repeat the calculation until the iteration conditions are met.

Images