Method for constructing online social platform topic topology fingerprint

Abstract

The application provides a kind of online social platform topic propagation topology fingerprint graph construction method, comprising: constructing topic propagation network graph, determining source node, extracting maximum connected subgraph, calculating subgraph each node forwarding volume, node hop count, adjacent edge forwarding intensity;From each subgraph source node, with forwarding intensity as weight, the shortest weighted path from source node as filter function, forward search out connected component sequence, record each connected component start time step and duration, the same connected component is merged;From each subgraph recorded search end node, start reverse search, get ring sequence from reverse search, record each ring structure start time step and duration, the same ring structure is merged;According to connected component sequence set, ring sequence set constructs topic propagation fingerprint graph.The application can effectively extract the significant topology of connected path and loop formed in the process of online social platform topic propagation, and quantify its duration.

Description

Technical Field

[0001] This invention belongs to the technical field of big data analysis of topic dissemination on online social platforms, specifically relating to a method for constructing a topic topology fingerprint map on online social platforms. Background Technology

[0002] With the advent of the Internet 2.0 era, new media such as online social platforms have gradually become the source and primary tool for the dissemination of topics. Currently, the dissemination of topics on online social platforms exhibits a trend and characteristics of weaponization. As a tool for topic dissemination, online social platforms can also serve as powerful psychological intervention weapons in the field of cyber information warfare. Therefore, timely and effective discovery of fractal topological structures that reveal the potential for ordered dissemination within irregular and complex phenomena during the dissemination of topics on online social platforms, and the construction of topological "fingerprint" maps, are crucial for understanding the dissemination of topics on online social platforms. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for constructing a topic topology fingerprint graph for online social platforms. This method can effectively extract the significant topological structure of connected paths and loops formed during the spread of topics on online social platforms, quantify their duration, and represent them as coordinate points in the graph, thereby facilitating further subsequent task processing and assisting in topic analysis and prediction.

[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0005] A method for constructing a topic topology fingerprint graph on an online social platform includes the following steps:

[0006] Step 1: For target topic data on online social platforms, construct a topic propagation network graph based on the information propagation and forwarding relationship between social platform node accounts, determine the topic source node account, extract the maximum connected subgraph starting from the source node, calculate the information forwarding volume of each node in the subgraph and the number of hops from the node to the source node, and on this basis, calculate the forwarding strength of adjacent nodes and use it as the edge weight.

[0007] Step 2: Based on the subgraph extracted from the topic propagation network graph, starting from any source node of the subgraph, perform a forward search with forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function until all nodes in the subgraph have been visited, and record the search end node; filter out the connected component sequence on the subgraph through this search process; with the source node of the subgraph as the 0th time step and the number of hops from each node to the source node as the relative time step, record the start time step and duration of the connected components in each set of sequences;

[0008] Step 3: Starting from other subgraph source nodes that have not been searched in the forward direction, execute Step 2 in sequence; form a sequence obtained from the search of multiple subgraph source nodes, forming a set of connected component sequences containing the start time step and duration; merge the same connected components obtained from the search of each subgraph source node, take the minimum start time, and take the duration as the sum of the durations of each connected component in each sequence;

[0009] Step 4: For each subgraph, start the reverse search from the end node of the record, use the forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function, until return to the source node, filter out the cycle sequence on the topic propagation network graph, and record the relative time steps of the connected components and the start and end of the cycle in each set of sequences.

[0010] Step 5: Starting from the end node of other subgraph records that have not been searched in reverse order, execute Step 4 in sequence; thus forming a sequence of multiple source node searches, forming a set of ring sequences containing the start time step and duration. Merge the same rings obtained by searching each subgraph source node, take the minimum start time, and take the duration as the sum of the durations of each ring structure in each sequence.

[0011] Step 6: Extract the start time step and duration of each sequence in the connected component sequence set and the ring sequence set to construct coordinate pairs, and merge them to form a coordinate set; plot each point in a two-dimensional plane coordinate system to construct a fingerprint map of the key topological characteristics of the target topic propagation.

[0012] Furthermore, step 1 specifically includes:

[0013] Step 1.1: For target topic data on online social platforms, construct a node set V of the topic propagation network graph based on the node account information involved in the propagation process of the target topic, and construct a set of directed edges E of the cascade graph based on the topic propagation and forwarding relationship between nodes. Based on this, construct the topic propagation network graph G(V,E).

[0014] Step 1.2: Extract the user nodes that published original blog posts in the topic propagation network graph, and determine that the user nodes are the message source accounts for the propagation of the target topic; based on each source account, i.e., the source node, use the depth-first algorithm to extract the largest connected subgraph G'(V',E') connected to the source node;

[0015] Step 1.3: Based on the extracted subgraph G'(V',E'), starting from the source account, calculate the forwarding volume of each node according to the node forwarding relationship, and calculate the forwarding strength of adjacent nodes. The calculation method for the forwarding volume of each node is as follows:

[0016] fw(v)=OD(v),v∈V'

[0017] Where node v is any node in the node set V' of subgraph G', fw(v) is the forwarding number of node v, and OD(v) is the out-degree of node;

[0018] Step 1.4: Taking the source node as the 0th hop, calculate the hop count between the node and the source node, and calculate the forwarding strength of adjacent nodes:

[0019]

[0020] Where node v is any node in the node set V' of subgraph G', s(v) is the forwarding strength of node v, OD(v) is the out-degree of node v, and ID(v) is the in-degree of node v.

[0021] Furthermore, the filtering function in step 2, which uses forwarding strength as the edge weight and the shortest weighted path from the source node, is as follows:

[0022] f(v i )=min d(v i ,v s )

[0023] Among them, v i For any node in subgraph G', v s For the current source node, d(v) i v s ) is node v i to source node v s The distance is denoted by , and min is the minimum value function.

[0024] Furthermore, during the forward search process in step 2, the current node v i The search strategy is: select the node with a distance v from the source node. s The neighbor node v with the smallest increase in shortest weighted path distance j To access, that is:

[0025] f(v i ,v j )=min(f(v i ),f(v j ))

[0026] Among them, v i v j For any node in subgraph G', f(v i ) is node v i Distance from source node v s Shortest weighted path length, f(v) j ) is node v j Distance from source node v s The shortest weighted path length, min is the minimum value function.

[0027] Furthermore, step 3 includes the following sub-steps:

[0028] Step 3.1: Starting from other unsearched subgraph source nodes, execute Step 2 sequentially; thereby obtaining multiple connected component sequences searched from subgraph source nodes, forming a set of connected component sequences containing the start time step and duration;

[0029] Step 3.2: In the set of connected component sequences obtained in Step 3.1, merge the identical connected components obtained from each subgraph starting from the source node. The start time is taken as the minimum start time, and the duration is the sum of the durations of each connected component in each sequence. That is:

[0030] t begin =min(t) begin (c mi ),t begin (c nj )),

[0031] t p =t p (c mi )+t p (c nj )

[0032] if c mi ==c nj

[0033] Among them, c mi For v sm c is the i-th connected component obtained by filtering the source node; nj For v sn For the j-th connected component obtained by filtering the source node, t begin (c mi ) is a connected component c mi The start time step, t begin (c nj ) is a connected component c nj The start time step, t p (c mi ) is a connected component c mi The duration, t p (c nj ) is a connected component c nj The duration.

[0034] Furthermore, step 4 is implemented as follows:

[0035] Step 4.1: Based on the search performed in Step 3, perform a reverse search for each subgraph starting from the end node of the record and moving towards the source node, using forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function;

[0036] Step 4.2: Execute step 4.1 until backtracking to the source node v s Based on the above process, for the loop sequences filtered out on subgraph G', with the source node of the subgraph as time step 0 and the hop count of each node from the source node as the relative time step, the time step t at which the loop begins in each set of sequences is recorded. begin End time step t end And based on this, the duration t of the ring structure is calculated. p The ring sequence is represented as follows:

[0037] [(ring 1, t) begin , t p ), (ring 2, t begin , t p ),…(ring n,t) begin , t p )]

[0038] Among them, t p =t begin -t end .

[0039] Furthermore, the reverse search strategy is as follows:

[0040] During the reverse search, starting from the current node v i Starting from the source node, select the neighbor node v with the largest decreasing shortest weighted path distance from the source node. j To access, that is:

[0041] f(v i ,v j )=max(f(v i ),f(v j ))

[0042] Among them, v i v j For any node in subgraph G', f(v i ) is node v i Distance from source node v s Shortest weighted path length, f(v) j ) is node v j Distance from source node v s The shortest weighted path length, where max is the maximum value function.

[0043] Furthermore, step 5 includes the following sub-steps:

[0044] Step 5.1: Starting from the end node of other unsearched subgraph records, execute step 4 in sequence to obtain the ring sequences found by reverse searching of multiple subgraph source nodes, forming a set of ring sequences containing the start time step and duration.

[0045] Step 5.2: In the set of ring sequences obtained in Step 5.1, merge the identical ring structures obtained through the search of each subgraph source node. The start time is taken as the minimum start time, and the duration is taken as the sum of the durations of each connected component in each sequence, that is:

[0046] t begin =min(t) begin (l mi ),t begin (l nj )),

[0047] t p =t p (l mi )+t p (l nj )

[0048] if l mi ==l nj

[0049] Among them, l mi For v sm The i-th ring obtained by filtering the source node; l nj For v sn For the j-th ring obtained by filtering the source node, t begin (l mi ) is a ring l mi The start time step, t begin (l nj ) is a ring l nj The start time step, t p (l mi ) is a ring l mi The duration, t p (l nj ) is a ring l nj The duration.

[0050] Further, in step 6, the union of the set of connected component sequences and the set of ring sequences is extracted. The start time step and duration of each sequence are retained in the union, and coordinate pairs are constructed based on this. These are then merged to form a coordinate set, which takes the form:

[0051] {(t begin1 , t p1 ), (t begin2 , t p2 )…(t bN , tpN )}

[0052] Among them, t begini t is the time step at which the union begins. pi Let i be the duration of the union, where i = 1 - N, and N is the number of sequences in the union set.

[0053] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention proposes a method for constructing a topological fingerprint of topic propagation on online social platforms. Based on target topic propagation data, a topic propagation forwarding network is constructed. Connected components and loop structures in the network are extracted through forward and backward search, and the start time step and duration of each topological structure are recorded. This extracts topological structures with significant characteristics in the topic propagation network and their duration of action, which are then converted into two-dimensional planar coordinates to create a fingerprint. This fingerprint can then be used for downstream task processing using point cloud, graph machine learning, and deep learning techniques. This invention effectively extracts the significant topological structures of connected paths and loops formed during topic propagation on online social platforms and quantifies their duration of action, representing them as coordinate points in the graph, thus facilitating further subsequent task processing. Furthermore, since the constructed topological fingerprint of topic propagation on online social platforms possesses topological saliency and high-order properties, it can provide effective tools and fundamental methods for subsequent topic propagation analysis and prediction. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating a method for constructing a topic topology fingerprint graph for an online social platform, as described in an embodiment of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0056] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0057] The present invention will be further described below with reference to specific embodiments, but these are not intended to limit the scope of the invention.

[0058] like Figure 1 As shown in the figure, this invention discloses a method for constructing a topic topology fingerprint graph on an online social platform, including the following steps:

[0059] Step 1: Based on the information propagation and forwarding relationships between target topic propagation data on online social platforms, construct a topic propagation network graph, determine the topic source node account, extract the maximum connected subgraph originating from the source node, and calculate the information forwarding volume of each node in the subgraph and the number of hops from the source node to the node. Based on this, define and calculate the forwarding strength of adjacent nodes as edge weights. This step includes the following sub-steps:

[0060] Step 1.1: For the target topic propagation data on online social platforms, construct the node set V of the topic propagation network graph based on the node account information involved in the target topic propagation process, and construct the directed edge set E of the cascade graph based on the topic propagation and forwarding relationship between nodes. Based on this, construct the topic propagation network graph G(V,E).

[0061] Step 1.2: Extract the user nodes that published original blog posts in the topic propagation network graph, and determine that the user nodes are the message source accounts for the propagation of the target topic; use a depth-first algorithm based on each source account, i.e., the source node, that is: start from the source node and search for unvisited nodes along the edges of the current node; if there are no unvisited nodes, return to the previous node and continue to try other nodes until all edges of the source node have been searched, and extract the maximum connected subgraph G'(V',E') connected to the source node.

[0062] Step 1.3: Based on the extracted subgraph G'(V',E'), starting from the source account, calculate the forwarding volume of each node according to the node forwarding relationship. The calculation method is as follows:

[0063] fw(v)=OD(v),v∈V'

[0064] Where node v is any node in the node set V' of subgraph G', fw(v) is the forwarding number of node v, and OD(v) is the out-degree of node;

[0065] Taking the source node as the 0th hop, calculate the hop count from the source node to the next node, and calculate the forwarding strength of adjacent nodes. The expression is:

[0066]

[0067] Where node v is any node in the node set V' of subgraph G', s(v) is the forwarding strength of node v, OD(v) is the out-degree of node v, and ID(v) is the in-degree of node v.

[0068] Step 2: Based on the subgraph extracted from the topic propagation network graph, starting from any source node in the subgraph, perform a forward search using forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function. Continue this search process as the edges grow until all nodes in the subgraph have been visited, recording the end node. This search process filters out connected component sequences from the subgraph. Using the source node as time step 0 and the hop count from each node to the source node as the relative time step, record the start time step and duration of the connected components in each sequence. This step specifically includes:

[0069] Step 2.1: From the source node v of subgraph G' s Starting from the source node, a forward search is performed, using forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function. The filtering function is as follows:

[0070] f(v i )=min d(v i ,v s )

[0071] Among them, v i For any node in subgraph G', v s For the current source node, d(v) i v s ) is node v i to source node v s The distance is denoted by , and min is the minimum value function.

[0072] During the forward search, the current node v i The search strategy is: select the node with a distance v from the source node. s The neighbor node v with the smallest increase in shortest weighted path distance j To access, that is:

[0073] f(v i ,v j )=min(f(v i ),f(v j ))

[0074] Among them, v i v j For any node in subgraph G', f(v i ) is node v i Distance from source node v s Shortest weighted path length, f(v) j ) is node v j Distance from source node v s The shortest weighted path length, min is the minimum value function;

[0075] Step 2.2: Execute step 2.1 until all nodes in subgraph G' have been visited, and record the search termination node; based on the above process, filter out a set of connected component sequences from v in subgraph G'. s The connected paths to each node; with the source node of the subgraph as time step 0, and the hop count of each node from the source node as the relative time step, record the start time step t of the connected components in each sequence. begin End time step t end And based on this, calculate the duration t from start to finish. p This forms a sequence of connected components containing a start time step and a duration, such as:

[0076] [(connected component 1, t) begin , t p (connected component 2, t) begin , t p ),…(connected components n, t) begin , t p )]

[0077] Among them, t p =t begin -t end .

[0078] Step 3: Starting from other sub-graph source nodes that have not been searched in the forward direction, proceed with Step 2; this creates a set of connected component sequences obtained from multiple sub-graph source node searches, including the start time step and duration. Merge identical connected components obtained from each sub-graph source node search, taking the minimum start time and the sum of the durations of each connected component in each sequence. This step includes the following sub-steps:

[0079] Step 3.1: Starting from other unsearched subgraph source nodes, proceed with Step 2 sequentially; this yields multiple connected component sequences searched from subgraph source nodes, forming a set of connected component sequences including the start time step and duration. The set of connected component sequences includes:

[0080] {[(v s1 Connected component 1, t begin , t p ),(v s1 Connected component 2, t begin , t p ),…(v s1 Connected components n, t begin , t p )],

[0081] [(v s2 Connected component 1, t begin , t p ),(v s2Connected component 2, t begin , t p ),…(v s2 Connected components n, t begin , t p )],…

[0082] [(v sk Connected component 1, t begin , t p ),(v sk Connected component 2, t begin , t p ),…(v sk Connected components n, t begin , t p )]}

[0083] Among them, t p =t begin -t end .

[0084] Step 3.2: In the set of connected component sequences obtained in Step 3.1, merge the identical connected components obtained from each subgraph starting from the source node. The start time is taken as the minimum start time, and the duration is the sum of the durations of each connected component in each sequence. That is:

[0085] t begin =min(t) begin (c mi ),t begin (c nj )),

[0086] t p =t p (c mi )+t p (c nj )

[0087] if c mi ==c nj

[0088] Among them, c mi For v sm c is the i-th connected component obtained by filtering the source node; nj For v sn The j-th connected component obtained by filtering the source node. begin (c mi ) is a connected component c mi The start time step, t begin (c nj ) is a connected component c nj The start time step, t p (c mi) is a connected component c mi The duration, t p (c nj ) is a connected component c nj The duration.

[0089] Step 4: For each subgraph, start the reverse search from the end node of the record, using forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function, until returning to the source node. Filter out the cycle sequences on the topic propagation network graph and record the connected components and the relative time steps of the start and end of the cycle in each sequence. Step 4 specifically includes:

[0090] Step 4.1: Based on the search performed in Step 3, for each subgraph, perform a reverse search from the end node of the record towards the source node, using forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function. The filtering function is as follows:

[0091] f(v i ,v j )=max(f(v i ),f(v j ))

[0092] Among them, v i v j For any node in subgraph G', f(v i ) is node v i Distance from source node v s Shortest weighted path length, f(v) j ) is node v j Distance from source node v s The shortest weighted path length, where max is the maximum value function;

[0093] The reverse search strategy in this step is as follows: During the reverse search process, starting from the current node v... i Starting from the source node, select the neighbor node v with the largest decreasing shortest weighted path distance from the source node. j Access is permitted.

[0094] Step 4.2: Execute step 4.1 until backtracking to the source node v s Based on the above reverse search process, combined with the connected components obtained from the forward search, a set of cycle sequences is filtered out on the subgraph G'. This cycle sequence contains the topological structure of loops. Taking the source node of the subgraph as the 0th time step and the hop count of each node from the source node as the relative time step, the time step t at which the cycle begins in each set of sequences is recorded. begin End time step t end And based on this, the duration t of the ring structure is calculated. p The details are as follows:

[0095] [(ring 1, t) begin , t p ), (ring 2, t begin , t p ),…(ring n,t) begin , t p )]

[0096] Among them, t p =t begin -t end .

[0097] Step 5: Starting from the end node of each subgraph record that has not been searched in reverse, proceed with Step 4; this creates a sequence of multiple source node searches, forming a set of ring sequences containing the start time step and duration. Merge identical rings obtained from the source node searches of each subgraph, taking the minimum start time and the sum of the durations of each connected component in each sequence; the specific implementation of this step is as follows:

[0098] Step 5.1: Starting from the end node of other unsearched subgraph records, proceed with Step 4; this yields multiple ring sequences obtained by reverse searching from the source nodes of the subgraph, forming a set of ring sequences containing the start time step and duration. The set of ring sequences takes the following form:

[0099] {[(v s1 Ring 1, t begin , t p ),(v s1 Ring 2, t begin , t p ),…(v s1 Ring n, t begin , t p )],

[0100] [(v s2 Ring 1, t begin , t p ),(v s2 Ring 2, t begin , t p ),…(v s2 Ring n, t begin , t p )],…

[0101] [(v sk Ring 1, t begin , t p ),(v sk Ring 2, t begin , t p ),…(v sk Ring n, t begin , t p )]}

[0102] Among them, t p =t begin -t end

[0103] Step 5.2: In the set of ring sequences obtained in Step 5.1, merge the identical ring structures obtained through the search of each subgraph source node. The start time is taken as the minimum start time, and the duration is taken as the sum of the durations of each connected component in each sequence, that is:

[0104] t begin =min(t) begin (l mi ),t begin (l nj )),

[0105] t p =t p (l mi )+t p (l nj )

[0106] if l mi ==l nj

[0107] Among them, l mi For v sm The i-th ring obtained by filtering the source node; l nj For v sn This represents the j-th ring obtained by filtering the source node. begin (l mi ) is a ring l mi The start time step, t begin (l nj ) is a ring l nj The start time step, t p (l mi ) is a ring l mi The duration, t p (l nj ) is a ring l nj The duration.

[0108] Step 6: Extract the start time step and duration of each sequence from the connected component sequence set and the ring sequence set to construct coordinate pairs, and merge them to form a coordinate set; plot each point in a two-dimensional plane coordinate system to construct a "fingerprint" map of the key topological characteristics of topic propagation;

[0109] In this step, the union of the set of connected component sequences and the set of ring sequences is extracted. The start time step and duration of each sequence are retained and concentrated, and coordinate pairs are constructed based on this. These are then merged to form a coordinate set; for example:

[0110] {(t begin1 , t p1 ), (t begin2 , t p2 )…(t bN , t pN )}

[0111] Among them, t begini t is the time step at which the union begins. pi Let i be the duration of the union, where i = 1 - N, and N is the number of sequences in the union set.

[0112] Accordingly, the fractal structure, including connected components and loops, is extracted from the topic propagation network and represented as a series of coordinate points. Based on the start time step and duration information contained in the coordinate points, the start time and duration are used as the x and y coordinates in a two-dimensional coordinate system, respectively, to plot each point in a two-dimensional plane coordinate system, constructing a "fingerprint" map of the key topological characteristics of the topic propagation. This fingerprint map contains significant topological structure information of the topic propagation network, which can be further processed and used with point cloud, graph machine learning, deep learning, and other related technologies to perform subsequent tasks.

[0113] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made based on the content of this specification should be included within the protection scope of the present invention.

Claims

1. A method for constructing a topic topology fingerprint graph on an online social platform, characterized in that, Includes the following steps: Step 1: For target topic data on online social platforms, construct a topic propagation network graph based on the information propagation and forwarding relationship between social platform node accounts, determine the topic source node account, extract the maximum connected subgraph starting from the source node, calculate the information forwarding volume of each node in the subgraph and the number of hops from the node to the source node, and on this basis, calculate the forwarding strength of adjacent nodes and use it as the edge weight. Step 2: Based on the subgraph extracted from the topic propagation network graph, starting from any source node of the subgraph, perform a forward search with forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function until all nodes in the subgraph have been visited, and record the search end node. The search process filters out the connected component sequences on the subgraph; with the source node of the subgraph as the 0th time step and the number of hops from each node to the source node as the relative time step, the start time step and duration of the connected components in each set of sequences are recorded. Step 3: Starting from other subgraph source nodes that have not been searched in the forward direction, execute Step 2 in sequence; form a sequence obtained from the search of multiple subgraph source nodes, forming a set of connected component sequences containing the start time step and duration; merge the same connected components obtained from the search of each subgraph source node, take the minimum start time, and take the duration as the sum of the durations of each connected component in each sequence; Step 4: For each subgraph, start the reverse search from the end node of the record, use the forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function, until return to the source node, filter out the cycle sequence on the topic propagation network graph, and record the relative time steps of the connected components and the start and end of the cycle in each set of sequences. Step 5: Starting from the end node of other subgraph records that have not been searched in reverse order, execute Step 4 in sequence; thus forming a sequence obtained from multiple source node searches, forming a set of ring sequences containing the start time step and duration. Merge the same rings obtained from each subgraph source node search, take the minimum start time, and take the sum of the durations of each ring structure in each sequence. Step 6: Extract the start time step and duration of each sequence in the connected component sequence set and the ring sequence set to construct coordinate pairs, and merge them to form a coordinate set; plot each point in a two-dimensional plane coordinate system to construct a fingerprint map of the key topological characteristics of the target topic propagation.

2. The method for constructing a topic topology fingerprint graph for an online social platform according to claim 1, characterized in that, Step 1 specifically includes: Step 1.1: For target topic data on online social platforms, construct a node set V of the topic propagation network graph based on the node account information involved in the propagation process of the target topic, and construct a set of directed edges E of the cascade graph based on the topic propagation and forwarding relationship between nodes. Based on this, construct the topic propagation network graph G(V, E). Step 1.2: Extract the user nodes that published original blog posts in the topic propagation network graph, and determine that the user nodes are the message source accounts for the propagation of the target topic; based on each source account, i.e., the source node, use the depth-first algorithm to extract the largest connected subgraph G'(V',E') connected to the source node; Step 1.3: Based on the extracted subgraph G'(V',E'), starting from the source account, calculate the forwarding volume of each node according to the node forwarding relationship, and calculate the forwarding strength of adjacent nodes. The calculation method for the forwarding volume of each node is as follows: Where node v is any node in the node set V' of subgraph G', fw(v) is the forwarding number of node v, and OD(v) is the out-degree of node; Step 1.4: Taking the source node as the 0th hop, calculate the hop count from the source node to the node, and calculate the forwarding strength of adjacent nodes: Where node v is any node in the node set V' of subgraph G', s(v) is the forwarding strength of node v, OD(v) is the out-degree of node v, and ID(v) is the in-degree of node v.

3. The method for constructing a topic topology fingerprint graph for an online social platform according to claim 1, characterized in that, The filtering function in step 2, which uses forwarding strength as the edge weight and the shortest weighted path from the source node, is as follows: in, v i For any node in subgraph G', v s For the current source node, d(v) i ，v s ） For nodes v i To the source node v s distance, min It is a minimum value function.

4. The method for constructing a topic topology fingerprint graph for an online social platform according to claim 3, characterized in that, During the forward search in step 2, the current node v i The search strategy is: select the node with a distance v from the source node. s The neighbor node v with the smallest increase in shortest weighted path distance j To access, that is: in, v i 、v j Let be any node in the subgraph G'. f(v i ) For nodes v i Distance from source node v s Shortest weighted path length f (v j ) For nodes v j Distance from source node v s Shortest weighted path length min It is a minimum value function.

5. The method for constructing a topic topology fingerprint graph for an online social platform according to claim 1, characterized in that, Step 3 includes the following sub-steps: Step 3.1: Starting from other unsearched subgraph source nodes in sequence, execute step 2; thereby obtaining multiple connected component sequences searched from subgraph source nodes, forming a set of connected component sequences containing the start time step and duration; Step 3.2: In the set of connected component sequences obtained in Step 3.1, merge the identical connected components obtained from each subgraph starting from the source node. The start time is taken as the minimum start time, and the duration is the sum of the durations of each connected component in each sequence. That is: in, c mi For v sm The first obtained by filtering the source node i One connected component; c nj For v sn The first obtained by filtering the source node j Connected components t begin (c mi ) For connected components c mi The start time step, t begin (c nj ) For connected components c nj The start time step, t p (c mi ) For connected components c mi Duration, t p (c nj ) For connected components c nj The duration.

6. The method for constructing a topic topology fingerprint graph for an online social platform according to claim 1, characterized in that, Step 4 is implemented as follows: Step 4.1: Based on the search performed in Step 3, perform a reverse search for each subgraph starting from the end node of the record and moving towards the source node, using forwarding strength as the edge weight and the shortest weighted path from the source node as the filtering function; Step 4.2: Execute step 4.1 until backtracking to the source node v s Based on the above process, for the loop sequences filtered out on subgraph G', with the source node of the subgraph as the 0th time step and the hop count of each node from the source node as the relative time step, the time step t at which the loop begins in each set of sequences is recorded. begin End time step t end And based on this, the duration t of the ring structure is calculated. p The ring sequence is represented as follows: [(ring 1, t) begin , t p ),( Ring 2, t begin , t p ),…(ring n,t) begin , t p )] Among them, t p = t begin -t end .

7. The method for constructing a topic topology fingerprint graph for an online social platform according to claim 6, characterized in that, The reverse search strategy is: During the reverse search, starting from the current node v i Starting from the source node, select the neighbor node v with the largest decreasing shortest weighted path distance from the source node. j To access, that is: Among them, v i v j For any node in subgraph G', f(v i ) is node v i Distance from source node v s Shortest weighted path length, f(v) j ) is node v j Distance from source node v s The shortest weighted path length, where max is the maximum value function.

8. The method for constructing a topic topology fingerprint graph for an online social platform according to claim 1, characterized in that, Step 5 includes the following sub-steps: Step 5.1: Starting from the end node of other unsearched subgraph records, execute step 4 in sequence to obtain the ring sequences found by reverse searching of multiple subgraph source nodes, forming a set of ring sequences containing the start time step and duration. Step 5.2: In the set of ring sequences obtained in Step 5.1, merge the identical ring structures obtained through the search of each subgraph source node. The start time is taken as the minimum start time, and the duration is taken as the sum of the durations of each connected component in each sequence, that is: in, l mi For v sm The first obtained by filtering the source node i One ring; l nj For v sn The first obtained by filtering the source node j One ring, t begin (l mi ) For ring l mi The start time step, t begin (l nj ) For ring l nj The start time step, t p (l mi ) For ring l mi Duration, t p (l nj ) For ring l nj The duration.

9. The method for constructing a topic topology fingerprint graph for an online social platform according to claim 1, characterized in that, Step 6 extracts the union of the connected component sequence set and the ring sequence set, retains the start time step and duration of each sequence in the union, and constructs coordinate pairs based on this. These coordinate pairs are then merged to form a coordinate set, which takes the form: {(t begin1 ,t p1 ),(t begin2 ,t p2 )…(t beginN ,t pN )} Among them, t begini t is the time step at which the union begins. pi Let be the duration of the union, where i = 1, 2, ..., N, and N is the number of sequences in the union set.

Images