CN112667784A - Rumor restraining method based on weighted reverse sampling - Google Patents
Rumor restraining method based on weighted reverse sampling Download PDFInfo
- Publication number
- CN112667784A CN112667784A CN202110049624.2A CN202110049624A CN112667784A CN 112667784 A CN112667784 A CN 112667784A CN 202110049624 A CN202110049624 A CN 202110049624A CN 112667784 A CN112667784 A CN 112667784A
- Authority
- CN
- China
- Prior art keywords
- node
- rumor
- nodes
- wrr
- graph
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a rumor restraining method based on weighted inverse sampling. The invention provides a weighted inverse sampling frame based on node weight by analyzing and modeling a rumor-true phase competition cascade propagation process, and designs an efficient approximate algorithm solving result set which meets two conditions: is a set of points of scale no less than k; given a set of rumor node sets, the result set is a set of points for initial propagation of true facies, and a greater number of innocent nodes can be protected under the rumor-true facies competition cascade propagation model. The method comprehensively considers the weight of the user and the pertinence of user protection, realizes efficient rumor interdiction, and has great social value in the real world and the comprehensive social network.
Description
Technical Field
The invention belongs to the technical field of multimedia data mining, and particularly relates to a method for searching a result set in a social network based on a weighted reverse sampling frame to achieve a rumor suppression maximization effect.
Background
With the rapid development of online social platforms, opinions and innovative technologies can be rapidly propagated through social networks, but this also provides a medium for the propagation of negative information or rumors. Rumors, a ubiquitous phenomenon in social networks, not only waste information transmission resources, but also easily cause social confusion and panic. Therefore, effective rumor containment strategies have significant social value. In the information dissemination process, users tend to accept the first opinion when they receive two opposite opinions. Based on this, once an error message or rumor is detected, an effective rumor containment method is to introduce true phase nodes that compete with the rumor. Given a budget of k, the rumor containment problem requires finding k node propagation truth phases, ultimately maximizing the number of protected nodes.
Disclosure of Invention
The invention aims to provide a rumor suppression method based on weighted inverse sampling, aiming at the defects of the prior art.
The purpose of the invention is realized by the following technical scheme:
according to a first aspect of the present invention, there is provided a rumor suppression method based on weighted inverse sampling, comprising the steps of:
step 1: giving an original directed graph G (V, E), wherein V represents a node set in the graph, and E represents an edge set in the graph; the figure has n nodes and m edges; for each point V ∈ V, use din(v) Represents the in-degree of point v at G; the following preprocessing operations are performed on the original directed graph G to obtain a graph G':
1.1 obtaining the weight of each point by using the PageRank algorithm, namely, any point v in G has a weight wvE [0, 1) representing the importance degree of the node;
1.2 for any one edge<vi，vj>Setting the probability p on this edgeij＝1/din(vj) And p isijE [0, 1) represents node viInfluencing node vjThe probability of (d);
1.3 randomly select k points from G, set these points as rumor nodes and the other nodes as innocent nodes.
Step 2: designing a weighted backward sampling frame, wherein the frame defines a weighted competition backward reachable set which is marked as a WRR set; let the WRR set for v Point be TvThen T isvConsists of a set of points that can protect the initial node v from rumor nodes; the steps for generating a WRR are as follows:
2.1 on the graph G', selecting a node v as an initial node by taking the weight ratio of the node as the extraction probability;
2.2 reverse operation is carried out on the graph G 'to obtain a graph G', and the probability of the edge is kept unchanged;
2.3 traversing on the graph G' by taking the initial node v as a starting point; in the traversal process, each edge<vi，vj>With probability pjiThe points which pass through in the traversal process are stored in the WRR; once no new innocent nodes are traversed, the traversal process is stopped and the set WRR is returned, or when a rumor node is traversed, the result set WRR stored before the rumor node is traversed is returned.
And step 3: generating a set T1 containing n1 WRRs by using the weighted inverse sampling frame in the step 2; obtaining a temporary result set S' through a greedy algorithm on the set T1; and judging whether the number of WRR sets in the T1 covered by the temporary result set S' is enough, if so, executing the step 4, otherwise, doubling the sampling number of the T1, namely, the next time n1 is twice of the current time n1, and repeating the step 3.
And 4, step 4: generating another set T2 containing at most n2 WRRs by a weighted inverse sampling framework, and verifying the correctness of the temporary result set S' based on the set T2, which specifically comprises:
4.1 initialize i ═ 0 and c ═ 0, each time generating a WRR set TvI is incremented by one and x (S', T) is determinedv) If it is 1, if x (S', T)v) 1, then the value of c is increased by one; wherein, when set S' and TvIn the presence of the same point, x (S', T)v) 1, otherwise, x (S', T)v)＝0；
4.2 judging the WRR set in T2 covered by the temporary result set SWhether the number is enough and the number of i exceeds n2, if the number meets the requirement and the number of i does not exceed n2, calculating the temporary influence estimated value f '(S') -c- (Sigma)u∈ Vwu) I, otherwise, the impact estimate is set to
4.3, judging whether the obtained temporary result set S' has approximate guarantee, if not, doubling the sampling number of T1, namely, the next time n1 is twice of the current time n1, and repeating the step 3-4; if the approximate guarantee is satisfied, the set S' is returned as the final result set.
Further, in the step 2, the node weight ratio w is used on the graph Gv/∑u∈VwuAnd selecting the node v as an initial node for extracting the probability.
Further, in the step 3, the greedy algorithm is divided into k rounds, one point is found in each round and is stored in S', so that when the node is used as a true phase node, under the rumor-true phase competition cascade propagation model, the boundary increment is the largest for the found true phase node, that is, the number of the innocent nodes to be protected is increased the most.
Further, in step 3, in the rumor-true phase competition cascade propagation model, three node types including a rumor node, a true phase node and an innocent node are shared, wherein the rumor node and the true phase node propagate simultaneously, and once the innocent node is influenced by the rumor node or the true phase node, the node is marked as the rumor node or the true phase node and the state is kept unchanged; an innocent node is said to be protected if it is affected by a phase node.
Further, in step 3, it is determined whether the number of WRR sets in T1 covered by the temporary result set S' is enough, that is, the inequality is determined
And whether the result is true or not is determined, wherein e is a natural constant, e represents a set error coefficient, and delta represents a set result probability coefficient.
Further, in the step 4,
where e is a natural constant and e represents a set error coefficient.
Further, in step 4, it is determined whether the number of WRR sets in T2 covered by the temporary result set S' is sufficient, that is, the inequality is determined
And whether the result is true or not is determined, wherein e is a natural constant, e represents a set error coefficient, and delta represents a set result probability coefficient.
Further, in said step 4.3, returning the set S' as the final result set, the number of innocent nodes can be maximally protected in the graph by determining the k users to initiate true phase node propagation.
According to a second aspect of the present invention, there is provided a computer device comprising a memory and a processor, the memory having stored therein computer readable instructions which, when executed by the processor, cause the processor to perform the steps of the weighted inverse sampling based rumor suppression method described above.
According to a third aspect of the present invention, there is provided a storage medium having stored thereon computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the weighted inverse sampling based rumor suppression method described above.
The invention has the beneficial effects that: in order to quickly find a high-quality node set as a true phase node set, the invention firstly provides a weighted inverse sampling framework and designs an efficient approximation algorithm based on the framework. The true phase node set obtained by the approximation algorithm has stronger influence, and innocent nodes can be protected from being influenced by rumor nodes. The algorithm provided by the invention can provide a new thought and method for the rumor-splitting work on the social platform.
Drawings
FIG. 1 is a flow chart of an approximation algorithm based on a weighted inverse sampling framework provided by an embodiment of the present invention;
FIG. 2 is a reverse graph after pre-processing;
FIG. 3 is a schematic diagram of randomly generating a set of WRRs on a reverse graph;
FIG. 4 is a schematic diagram of the generation of a temporary result set based on a given set of WRRs;
fig. 5 is a diagram comparing the number of protection nodes of a plurality of algorithms on the real social network epion.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways than those specifically described and will be readily apparent to those of ordinary skill in the art without departing from the spirit of the present invention, and it is therefore not limited to the specific embodiments disclosed below.
As shown in fig. 1, a rumor suppression method based on weighted inverse sampling according to an embodiment of the present invention includes the steps of:
step 1: given an original directed graph G (V, E), where V represents a set of nodes in the graph and E represents a set of edges in the graph. The figure has n nodes and m edges in total. For each point V ∈ V, use din(v) Represents the in-degree of v at point G. For the original directed graph G, the following preprocessing operation is done to obtain a graph G'.
1.1 obtaining the weight of each point by using the PageRank algorithm, namely, any point v in G has a weight wvE 0, 1) represents the importance of this node.
1.2 for any one edge<vi，vj>Setting the probability p on this edgeij＝1/din(vj) And p isijE [0, 1) represents node viInfluencing node vjThe probability of (c).
1.3 randomly select k points from G, set these points as rumor nodes and the other nodes as innocent nodes, i.e. without any state.
Step 2: in graph G', a new weighted inverse sampling framework is designed based on the conventional inverse sampling framework. The framework defines a weighted contention reverse reachable set, abbreviated WRR set. Let the WRR set for v Point be TvThen T isvConsisting of a set of points that can protect the initial node v from rumor nodes. The specific steps for generating a WRR are as follows:
2.1 on graph G', in terms of node weight ratio (w)v/∑u∈Vwu) And selecting the node v as an initial node for extracting the probability.
2.2 reverse the graph G', i.e. change the direction of each edge, resulting in graph G ", the probability of the edge remaining unchanged.
2.3 go through the graph G' starting from the initial node v. In the traversal process, each edge<vi，vj>With probability pjiIs traversed and the points passed through during the traversal are stored in the WRR. Once no new innocent node is traversed, the traversal process is stopped and the set WRR is returned, or when a rumor node is traversed, the result set WRR stored before the rumor node is traversed is returned.
Based on a weighted inverse sampling framework, the approximation algorithm of the invention provides a lemma:
introduction 1: let f (S)t) Is a set of nodes StRepresenting a set of nodes StAs an initial set of true phases, the number of innocent nodes that would otherwise be affected by the rumor node can be protected. E [ f (S)t)]Denotes StThe desired value of the influence of. Let T bevIs a WRR set with v as an initial node, x (S, T)v) When 1 denotes the sets S and TvThere is the same point in (i.e., S can protect point v). Representing the set of all WRRs extracted by R, the unbiased estimate of the rumor containment problem proposed by the present invention is:
and step 3: using the weighted inverse sampling framework of step 2, a set T1 of n1 WRRs is generated. On the set T1, a temporary result set S' is obtained by a greedy algorithm.
3.1 the greedy algorithm is divided into k rounds, each round needs to find a point, which is stored in S', so that when it is used as a true phase node, under the rumor-true phase competition cascade propagation model, the boundary increment is maximum for the found true phase node, i.e. the number of the added protected innocent nodes is maximum. In the rumor-true phase competition cascade propagation model, three node types of rumor nodes, true phase nodes and innocent nodes are shared. Wherein the rumor node and the true phase node are propagated simultaneously, once an innocent node is influenced by the rumor node or the true phase node, the node is marked as the rumor node or the true phase node, and the state is kept unchanged. An innocent node is said to be protected if it is affected by a phase node.
3.2 judge whether the number of WRR sets in T1 covered by the temporary result set S' is enough, i.e. judge the inequality
Or not, where T1 represents a set of n1 WRRs. When set S' and TvIn the presence of the same point, x (S', T)v) 1, otherwise, x (S', T)v) 0. e is a natural constant, e represents a set error coefficient and is generally set to 0.1, and δ represents a set result probability coefficient and is generally set to 1/n. If the judgment condition is satisfied, continuing the step 4 and the subsequent steps, otherwise, doubling the sampling number of the T1, namely, the next time n1 is twice of the current time n1, and repeating the step 3.
And 4, step 4: generating another set T2 containing at most n2 WRRs by a weighted inverse sampling framework, and verifying the correctness of the temporary result set S' based on the set T2
The specific process is as follows:
4.1 initialize i ═ 0 and c ═ 0, each time generating a WRR set TvI is incremented by one and x (S', T) is determinedv) If it is 1, if x (S', T)v) 1, the value of c is incremented by one.
4.2 judging whether the number of WRR sets in T2 covered by the temporary result set S' is enough and the number of i exceeds n2, namely judging whether the number of WRR sets in T2 is enough or not
And i ≦ n2, where e is a natural constant, e represents the set error coefficient, typically set to 0.1, and δ represents the set resulting probability coefficient, typically set to 1/n. If this condition is satisfied, a temporary influence estimation value f '(S') is calculated as c · (∑ e)u∈Vwu) I, otherwise, the impact estimate is set to
4.3 judging whether the obtained temporary result set S' has approximate guarantee, namely judging the inequality
Whether or not this is true. If not, the number of samples of T1 is doubled, i.e., the next time n1 is twice as large as n1 at this time, and steps 3-4 are repeated. If yes, returning the set S' as a final result set. By determining the k users to initiate the true phase node propagation, the number of innocent nodes can be maximally protected in the graph.
In fig. 2, white circular nodes represent innocent nodes, and white square nodes represent rumor nodes. The node sequence number and the weight thereof are represented in the node, and the probability on the edge represents the propagation probability.
In fig. 3, white circular nodes represent innocent nodes, white square nodes represent rumor nodes, and gray circular nodes represent true phase nodes. By Tv1For example, the weighted inverse sampling framework generates the WRR set as follows:
1. extract v with a probability of 0.11Traversing in a reverse graph for the initial node whenv1Successfully influencing v with probabilities of 0.5 and 0.5, respectively6And v7V is to be6And v7Deposit Tv1。
2. Similarly, v6And v7Successfully influencing v with probabilities of 0.5 and 0.5, respectively9And v8V is to be8And v9Deposit Tv1。
3. When v is9Traverse to rumor node v4Then, the traversal process is stopped and the point before the rumor node is encountered is stored as Tv1。
In fig. 4, white circular nodes represent innocent nodes, white square nodes represent rumor nodes, and gray circular nodes represent true phase nodes. Assuming that k is 1, the process of selecting the temporary result set in step 3.1 is as follows:
1. firstly, respectively extracting initial nodes v with a certain probability1，v6，v7Three random WRR sets are generated and stored in R.
2. According to the observation, node v9Is the most frequently occurring node in the WRR set, then the ability of this node to suppress rumors in the original graph is maximized, { v9I.e. a temporary result set.
FIG. 5 is the experimental effect on the real data set, Epinions, with points 75,879 and edges 508,837. The experiment compares the BD algorithm with the RS algorithm, wherein the BD algorithm randomly selects k nodes from the top 20% of points with the highest degree as a return result, and the RS algorithm randomly selects the k nodes as the return result. The abscissa in fig. 5 represents the number k of selected phase nodes, and the ordinate represents the number of innocent nodes that can be protected by the k phase nodes selected by the respective algorithms, respectively. In the figure, the advantage of the algorithm (marked as WSC) provided by the invention is obvious compared with a simple heuristic algorithm. In the data set, the algorithm of the present invention is approximately 30% faster than the BD algorithm and approximately 75% faster than the RS algorithm.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having stored therein computer readable instructions, which when executed by the processor, cause the processor to perform the steps of the weighted inverse sampling based rumor containment method of the above embodiments.
In one embodiment, a storage medium is provided that stores computer readable instructions, which when executed by one or more processors, cause the one or more processors to perform the steps of the weighted inverse sampling based rumor containment method of the embodiments described above. The storage medium may be a nonvolatile storage medium.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.
The foregoing is only a preferred embodiment of the present invention, and although the present invention has been disclosed in the preferred embodiments, it is not intended to limit the present invention. Those skilled in the art can make numerous possible variations and modifications to the present teachings, or modify equivalent embodiments to equivalent variations, without departing from the scope of the present teachings, using the methods and techniques disclosed above. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention are still within the scope of the protection of the technical solution of the present invention, unless the contents of the technical solution of the present invention are departed.
Claims (10)
1. A rumor suppression method based on weighted inverse sampling, the method comprising the steps of:
step 1: giving an original directed graph G (V, E), wherein V represents a node set in the graph, and E represents an edge set in the graph; the figure has n nodes and m edges; for each point V ∈ V, use din(v) Represents the in-degree of point v at G; the following preprocessing operations are performed on the original directed graph G to obtain a graph G':
1.1 obtaining each by using the PageRank algorithmThe weights of the points, i.e. any point v in G has a weight wvE [0, 1) representing the importance degree of the node;
1.2 for any one edge<vi，vj>Setting the probability p on this edgeij＝1/din(vj) And p isijE [0, 1) represents node viInfluencing node vjThe probability of (d);
1.3 randomly select k points from G, set these points as rumor nodes and the other nodes as innocent nodes.
Step 2: designing a weighted backward sampling frame, wherein the frame defines a weighted competition backward reachable set which is marked as a WRR set; let the WRR set for v Point be TvThen T isvConsists of a set of points that can protect the initial node v from rumor nodes; the steps for generating a WRR are as follows:
2.1 on the graph G', selecting a node v as an initial node by taking the weight ratio of the node as the extraction probability;
2.2 reverse operation is carried out on the graph G 'to obtain a graph G', and the probability of the edge is kept unchanged;
2.3 traversing on the graph G' by taking the initial node v as a starting point; in the traversal process, each edge<vi，vj>With probability pjiThe points which pass through in the traversal process are stored in the WRR; once no new innocent nodes are traversed, the traversal process is stopped and the set WRR is returned, or when a rumor node is traversed, the result set WRR stored before the rumor node is traversed is returned.
And step 3: generating a set T1 containing n1 WRRs by using the weighted inverse sampling frame in the step 2; obtaining a temporary result set S' through a greedy algorithm on the set T1; and judging whether the number of WRR sets in the T1 covered by the temporary result set S' is enough, if so, executing the step 4, otherwise, doubling the sampling number of the T1, namely, the next time n1 is twice of the current time n1, and repeating the step 3.
And 4, step 4: generating another set T2 containing at most n2 WRRs by a weighted inverse sampling framework, and verifying the correctness of the temporary result set S' based on the set T2, which specifically comprises:
4.1 initialize i ═ 0 and c ═ 0, each time generating a WRR set TvI is incremented by one and x (S', T) is determinedv) If it is 1, if x (S', T)v) 1, then the value of c is increased by one; wherein, when set S' and TvIn the presence of the same point, x (S', T)v) 1, otherwise, x (S', T)v)＝0；
4.2, judging whether the number of WRR sets in T2 covered by the temporary result set S ' is enough and whether the number of i exceeds n2, and if the number meets the requirement and the number of i does not exceed n2, calculating a temporary influence estimation value f ' (S ') -c · (∑ Σ)u∈ Vwu) I, otherwise, the impact estimate is set to
4.3, judging whether the obtained temporary result set S' has approximate guarantee, if not, doubling the sampling number of T1, namely, the next time n1 is twice of the current time n1, and repeating the step 3-4; if the approximate guarantee is satisfied, the set S' is returned as the final result set.
2. The method of claim 1, wherein in step 2, w is weighted by node weight on graph Gv/∑u∈VwuAnd selecting the node v as an initial node for extracting the probability.
3. The method for suppression of rumors based on weighted inverse sampling as claimed in claim 1, wherein in said step 3, the greedy algorithm is divided into k rounds, each round needs to find a point, which is stored in S', so that when it is used as a true phase node, the boundary increment is the largest for the found true phase node, i.e. the number of the added protected innocent nodes is the largest under the rumor-true phase competition cascade propagation model.
4. The rumor suppression method based on weighted inverse sampling of claim 1, wherein in the rumor-true phase competition cascade propagation model, in step 3, there are three node types of rumor node, true phase node and innocent node, wherein the rumor node and true phase node propagate simultaneously, and once the innocent node is affected by the rumor node or true phase node, the node is marked as the rumor node or true phase node and the state is kept unchanged; an innocent node is said to be protected if it is affected by a phase node.
5. The method for restraining rumors based on weighted inverse sampling of claim 1, wherein in said step 3, it is determined whether the number of WRR sets in T1 covered by the temporary result set S' is enough, i.e. the inequality is determined
And whether the result is true or not is determined, wherein e is a natural constant, e represents a set error coefficient, and delta represents a set result probability coefficient.
6. The method of claim 1, wherein in step 4,
where e is a natural constant and e represents a set error coefficient.
7. The method of claim 1, wherein in step 4, the inequality is determined by determining whether the number of WRR sets in T2 covered by the temporary result set S' is sufficient
And whether the result is true or not is determined, wherein e is a natural constant, e represents a set error coefficient, and delta represents a set result probability coefficient.
8. The method of claim 1, wherein in step 4.3, the set S' is returned as the final result set, and the number of innocent nodes in the graph can be protected to the maximum extent by determining the k users to initiate true phase node propagation.
9. A computer device comprising a memory and a processor, the memory having stored therein computer readable instructions that, when executed by the processor, cause the processor to perform the steps of the weighted inverse sampling based rumor containment method of any one of claims 1-8.
10. A storage medium having computer-readable instructions stored thereon which, when executed by one or more processors, cause the one or more processors to perform the steps of the weighted inverse sampling based rumor containment method of any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110049624.2A CN112667784B (en) | 2021-01-14 | 2021-01-14 | Rumor restraining method based on weighted reverse sampling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110049624.2A CN112667784B (en) | 2021-01-14 | 2021-01-14 | Rumor restraining method based on weighted reverse sampling |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112667784A true CN112667784A (en) | 2021-04-16 |
| CN112667784B CN112667784B (en) | 2022-04-05 |
Family
ID=75415188
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110049624.2A Active CN112667784B (en) | 2021-01-14 | 2021-01-14 | Rumor restraining method based on weighted reverse sampling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112667784B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108520337A (en) * | 2018-03-22 | 2018-09-11 | 清华大学深圳研究生院 | A kind of rumour methods of risk assessment based on network risks entropy difference |
| CN110069718A (en) * | 2019-04-15 | 2019-07-30 | 哈尔滨工程大学 | A kind of social networks rumour suppressing method based on theme |
| CN112069397A (en) * | 2020-08-21 | 2020-12-11 | 三峡大学 | Rumor detection method combining self-attention mechanism with generation of confrontation network |
-
2021
- 2021-01-14 CN CN202110049624.2A patent/CN112667784B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108520337A (en) * | 2018-03-22 | 2018-09-11 | 清华大学深圳研究生院 | A kind of rumour methods of risk assessment based on network risks entropy difference |
| CN110069718A (en) * | 2019-04-15 | 2019-07-30 | 哈尔滨工程大学 | A kind of social networks rumour suppressing method based on theme |
| CN112069397A (en) * | 2020-08-21 | 2020-12-11 | 三峡大学 | Rumor detection method combining self-attention mechanism with generation of confrontation network |
Non-Patent Citations (4)
| Title |
|---|
| JIANMING ZHU等: "Robust rumor blocking problem with uncertain rumor sources in social networks", 《WORLD WIDE WEB》 * |
| QIHAO SHI等: "Adaptive Influence Blocking: Minimizing the Negative Spread by Observation-Based Policies", 《2019 IEEE 35TH INTERNATIONAL CONFERENCE ON DATA ENGINEERING (ICDE)》 * |
| 李亚琼: "社交网络阻断影响最大化问题的研究", 《中国优秀博硕士学位论文全文数据库(硕士)基础科学辑》 * |
| 郑建国: "在线社交网络中基于结构优化的谣言信息传播抑制", 《中国优秀博硕士学位论文全文数据库(硕士)基础科学辑》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112667784B (en) | 2022-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102061987B1 (en) | Risk Assessment Method and System | |
| Tesfahun et al. | Intrusion detection using random forests classifier with SMOTE and feature reduction | |
| CN112085046A (en) | Intrusion detection method and system based on sampling and feature reduction for unbalanced data set conversion | |
| Zhao et al. | A neural-network based DDoS detection system using hadoop and HBase | |
| CN107231382A (en) | A kind of Cyberthreat method for situation assessment and equipment | |
| CN111680291A (en) | Countermeasure sample generation method and device, electronic equipment and storage medium | |
| CN112446634A (en) | Method and system for detecting influence maximization node in social network | |
| CN113407425B (en) | Internal user behavior detection method based on BiGAN and OTSU | |
| CN112667784B (en) | Rumor restraining method based on weighted reverse sampling | |
| CN115391674B (en) | Method, device, equipment and storage medium for efficiently suppressing false information of network community | |
| CN107622201B (en) | A kind of Android platform clone's application program rapid detection method of anti-reinforcing | |
| JPH08221429A (en) | Automatic document sorter | |
| CN112000948B (en) | Password strength evaluation method based on sequence generation antagonistic neural network | |
| CN114417811A (en) | Similarity calculation method and device based on semantics and storage medium | |
| He et al. | Leveraging weighted automata in compositional reasoning about concurrent probabilistic systems | |
| CN110766338A (en) | DPOS (distributed data processing) bifurcation prediction model method based on artificial intelligence and EOS (Ethernet over Ethernet) and IO (input/output) of block chain technology | |
| CN112115509A (en) | Data generation method and device | |
| CN117457017B (en) | Voice data cleaning method and electronic equipment | |
| CN111614695B (en) | Network intrusion detection method and device of generalized inverse Dirichlet mixed HMM model | |
| Alsinet et al. | Argumentation reasoning with graph neural networks for reddit conversation analysis | |
| CN110674390B (en) | Confidence-based group discovery method and device | |
| CN114650435B (en) | Method and device for searching repeated segments in video and related equipment | |
| Allander et al. | Qualitative Image Selection with Active Learning | |
| KR100926876B1 (en) | Rank Learning Model Creation Method and Rank Learning Model Creation System Using Rank Occurrence Probability | |
| CN108710618B (en) | Propagation seed node selection method based on concurrent process and influence expectation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |