CN112818125B - Network topic structure evolution discovery method - Google Patents
Network topic structure evolution discovery method Download PDFInfo
- Publication number
- CN112818125B CN112818125B CN202110211155.XA CN202110211155A CN112818125B CN 112818125 B CN112818125 B CN 112818125B CN 202110211155 A CN202110211155 A CN 202110211155A CN 112818125 B CN112818125 B CN 112818125B
- Authority
- CN
- China
- Prior art keywords
- post
- topics
- time period
- posts
- period
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 239000011159 matrix material Substances 0.000 claims abstract description 14
- 238000011156 evaluation Methods 0.000 claims abstract description 6
- 239000013598 vector Substances 0.000 claims description 19
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 13
- 230000009466 transformation Effects 0.000 abstract description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Primary Health Care (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a network topicThe structure evolution discovery method comprises the steps of establishing evaluation functions of posts appearing in a time period t, emerging topics and historical topics before the time period t-delta t, namely a interest function B before the time period t ≤t Further capturing historical topics before t-delta t time period, newly added posts of the historical topics before t time period and newly added topics during t time period, and realizing dynamic structure evolution detection of the topic network during t time period; obtaining new posts of the historical topics before the t-delta t period in the t period and emerging topics before the t period through matrix transformation, classifying the historical topics before the t-delta t period, and obtaining a interest function B according to a classification result t + The topics are divided into two communities according to whether the topics are related to historical topics or not, and then historical categories to which posts related to the historical topics belong and emerging topic categories of the posts independent of the historical topics are obtained. Meanwhile, social attribute inheritance and semantic similarity of the topic network are considered, and the accuracy of network topic evolution detection is improved.
Description
Technical Field
The invention belongs to the technical field of network topic detection methods, and relates to a network topic structure evolution discovery method.
Background
With the rapid development of online social networks, various social platforms, for example, various OSN platforms such as Twitter and microblog, are also becoming increasingly popular. On-line social networks, a lot of important information (on-line topics) is embedded, which is helpful for discovering some important topics or high-value applications in certain aspects or the evolution patterns or laws of hot social phenomena and activities. The OSN platforms have a specific dynamic pattern of information penetration on topics (political and financial affairs, natural disasters, breaking news, etc.), and they have an immediate impact on the daily lives of netizens.
But studies on the evolution of dynamic network structures usually detect topics only by semantic similarity, which ignores the inheritance of social attributes. Current research is not sufficient to truly detect topics because they are one-sided views of research topic detection, resulting in low detection accuracy, they just consider semantic similarity, rather than inheriting social attributes between posts on OSNs.
Disclosure of Invention
The invention aims to provide a network topic structure evolution discovery method, which solves the problems that in the prior art, the topic detection only through semantic similarity leads to low detection accuracy and the problem that the network topic structure evolution process cannot be effectively discovered.
The technical scheme adopted by the invention is that a network topic structure evolution discovery method comprises the following steps:
step 1, establishing a interest function B before t time period ≤t Interest function B ≤t The evaluation function is used for representing that posts appearing in the time period t belong to emerging topics and historical topics before the time period t-delta t, and the time period t refers to the time period from the time period t-delta t to the time period t;
in the above formula, B ≤t-Δt As a function of interest before the t- Δ t period, B t As a function of interest before the t period, I i(≤t)j(≤t) An attribute representing text overlap between posts,key word vectors representing posts i post and j post respectively, F i(t)j(t) Representing a forward comment relationship between the i post and the j post,an indicative function indicating whether online post i and online post j are posts before the t-deltat period,a post that indicates that online post i belongs to the occurrence of the t period and online post j is an indicative function of the occurrence of the post prior to the t-deltat period,an indicative function representing whether the online post i and the online post j are posts occurring during the t period, b i(≤t) Indicates the number of posts, φ, that belong to the same topic as the online post i Ci,Cj An indicative function representing that post i and post j belong to the same topic;
step 2, obtaining a interest function B before the t-delta t time period according to the formula (1) ≤t-Δt :
Obtaining historical topics before a t-delta t time period according to a formula (2);
step 3, simultaneously obtaining a interest function B of the t time period t Comprises the following steps:
acquiring newly added posts of the historical topics in the t period and newly added topics in the t period according to a formula (3);
step 4, according to the step 2, the step 3 and the quadratic theory, the interest function B is processed t Conversion to a form comprising a positive definite matrix Z:
equation (4) is converted to the following form according to equation (6):
in the above formula, L 1 =Z -1 L;
By means of a symmetrical matrixCarrying out calculation and analysis on the characteristic value and the characteristic vector of the topic to obtain a new post of the historical topic in the t time period and a new topic in the t time period before the t-delta t time period;
and 5, converting the formula (2) into:
to formula (11)Analyzing the characteristic value and the characteristic vector to further realize the classification of the historical topics before the t-delta t time period;
step 6, establishing a interest function B of the historical topics before the t-delta t time period and the topics at the t time period t + :
In the above formula, when both the online post i and the online post j belong to posts appearing in the period t, θ i,j 1, otherwise zero;
utilizing the topic classification result before the t-delta t time interval obtained in the step 5 and a profit function B t + Set omega of online posts that appear during t period t Is divided into setsAnd setCollection ofIncluding posts, collections, relating to historical topicsIncluding posts that are not related to historical topics.
Step 7, utilizing the interest function B 1 t To the collectionIn each post, clustering the belonged historical topic categories, and utilizing a interest function B 2 t Pair setClustering emerging topics in each post, and performing a interest function B 1 t Interest function B 2 t Is represented as follows:
in the above formula, L 2 Is composed ofThe vector of the nodes in (1) is,is L 2 Of each post, L 3 Is composed ofThe vector of the nodes in (1) is,is L 3 The vector entries of each post.
The invention is also characterized in that:
in the step 1: if the online post i and the online post j are posts before the t-delta t time period, thenOtherwise, the value is zero; if the online post i belongs to the post appearing in the time interval t and the online post j is the post appearing before the time interval t-delta t, thenOtherwise, the value is zero; if it isIf the online post i and the online post j are posts appearing in the time period t, thenOtherwise it is zero.
In the step 1: if the i post and the j post have a forwarding comment relationship, F i(t)j(t) 1 is ═ 1; otherwise F i(t)j(t) Is zero.
In the step 1: if the overlap ratio of the keywords of the i post and the j post is more than 40 percent of the threshold valueOtherwise
The invention has the beneficial effects that:
the invention relates to a network topic structure evolution discovery method, which establishes evaluation functions of posts appearing in a time period t and emerging topics and historical topics before the time period t-delta t, namely a interest function B before the time period t ≤t Further capturing historical topics before t-delta t time period, newly added posts of the historical topics before t time period and newly added topics during t time period, and realizing dynamic structure evolution detection of the topic network during t time period; obtaining new posts of the historical topics before the t-delta t period in the t period and emerging topics before the t period through matrix transformation, classifying the historical topics before the t-delta t period, and obtaining a interest function B according to a classification result t + Dividing topics into two communities according to whether related to historical topics or not, and then obtaining eachHistory categories to which posts related to history topics belong, and posts which are not related to history topics and the emerging topic categories; meanwhile, social attribute inheritance (historical topics and emerging topics) and semantic similarity (the relationship between emerging topics and the historical topics and the emerging topics are considered, communities are divided according to whether the emerging topics belong to the historical topics, and the accuracy of network topic evolution detection is improved; the resolution limit dilemma can be prevented from occurring as compared with the conventional modularization method.
Drawings
FIG. 1 is a topic discovery diagram of an embodiment of a network topic structure evolution discovery method of the present invention;
fig. 2 is an evolution process discovery diagram of an embodiment of the network topic structure evolution discovery method of the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
A network topic structure evolution discovery method comprises the following steps:
step 1, establishing a interest function B before t time period ≤t Interest function B ≤t The evaluation function is used for representing that posts appearing in the time period t belong to emerging topics and historical topics before the time period t-delta t, and the time period t refers to the time period from the time period t-delta t to the time period t;
in the above formula, B ≤t-Δt As a function of interest before the t- Δ t period, B t As a function of interest before the t period, I i(≤t)j(≤t) Representing the attribute of text overlap degree between posts, if the overlap degree of keywords of i post and j post is more than 40 percent of threshold valueOtherwiseRespectively represent the i labelJ keyword vector of post, F i(t)j(t) Representing a forward comment relationship between the i post and the j post, and if so, F i(t)j(t) 1, otherwise F i(t)j(t) =0;Indicating whether the online post i and the online post j are indicative functions of posts before the t-delta t period, and if so, indicating that the online posts i and j are indicative functions of posts before the t-delta t periodOtherwise, the value is zero;an indicative function representing that online post i belongs to posts occurring during the t period and that online post j is a post occurring before the t- Δ t period; if so, thenOtherwise, the value is zero,indicating whether the online post i and the online post j are indicative functions of posts occurring during the t period, and if so, indicating that the posts occur during the t periodOtherwise, the value is zero; b i(≤t) Representing the number of posts that belong to the same topic as online post i,an indicative function indicating that post i and post j belong to the same topic, and if so, thenOtherwise it is zero.
Step 2, obtaining a interest function B before a t-delta t time period according to the formula (1) ≤t-Δt :
Obtaining historical topics before a t-delta t time interval according to a formula (2);
step 3, simultaneously obtaining a interest function B of the t time period t Comprises the following steps:
and (4) acquiring newly added posts of the historical topics before the t period and newly-added topics during the t period according to the formula (3), and completing dynamic structure evolution detection of the topic network during the t period.
Step 4, according to the step 2, the step 3 and the quadratic form theory, the interest function B is obtained t Conversion to a form comprising a positive definite matrix Z:
equation (4) is converted to the following form according to equation (6):
in the above formula, L 1 =Z -1 L;
By aligning the above symmetric matrixThe characteristic value and the characteristic vector of the topic are calculated and analyzed, and a new post of the historical topic in the t time period and a new topic (namely the new topic in the t time period) in the t time period before the t-delta t time period are obtained;
and 5, converting the formula (2) into:
analyzing the characteristic value and the characteristic vector in the formula (11), and if posts before each t-delta t time period belong to the same topic, the positive and negative characteristics of the characteristic value are consistent, so that the classification of historical topics before the t-delta t time period is realized;
step 6, establishing a interest function B of the historical topics before the t-delta t time period and the topics at the t time period t + :
In the above formula, when both the online post i and the online post j belong to posts appearing in the period t, θ i,j 1, otherwise zero;
using the t- Δ t obtained in step 5Topic classification result before segment and interest function B t + Set omega of online posts that occur during t periods t Is divided into setsAnd collectionsCollectionIncluding posts, collections, relating to historical topicsIncluding posts that are unrelated to historical topics. Specifically, by a pairFeature value and feature vector analysis is performed for posts in each t-period if they all belong toThe positive and negative of the eigenvalues are consistent; if they all belong toThe signs of their corresponding feature values are consistent.
Step 7, utilizing the interest function B 1 t To the collectionEach post in the system is clustered with the belonged historical topic category by utilizing a interest function B 2 t Pair setClustering emerging topics in each post, and performing a interest function B 1 t Interest function B 2 t Is represented as follows:
in the above formula, L 2 Is composed ofThe vector of the nodes in (1) is,is L 2 Of each post, L 3 Is composed ofThe vector of the nodes in (1) is,is L 3 The vector entries of each post.
Specifically, by applying the above toAndcarrying out eigenvalue and eigenvector analysis, and carrying out analysis on each time period tThe post of (1), if they all belong to the same topic, the feature value is positiveNegativity is uniform; if they all belong toIf the same topic is found, the corresponding feature values are consistent in positive and negative.
Through the mode, the network topic structure evolution discovery method provided by the invention has the advantages that the evaluation functions of posts and emerging topics appearing in the time period t and historical topics before the time period t-delta t, namely the interest function B before the time period t are established ≤t Further capturing the historical topics before the t-delta t period, newly added posts of the historical topics before the t period and newly added topics after the t period, thereby realizing dynamic structure evolution detection of the topic network during the t period; obtaining new posts of the historical topics before the t-delta t period in the t period and emerging topics before the t period through matrix transformation, classifying the historical topics before the t-delta t period, and obtaining a interest function B according to a classification result t + Dividing topics into two communities according to whether the topics are related to historical topics or not, and then obtaining historical categories to which posts related to the historical topics belong and emerging topic categories of the posts which are not related to the historical topics; meanwhile, social attribute inheritance (historical topics and emerging topics) and semantic similarity (the relationship between emerging topics and the historical topics and the emerging topics are considered, communities are divided according to whether the emerging topics belong to the historical topics, and the accuracy of network topic evolution detection is improved; the resolution limit dilemma can be prevented from occurring, compared to the conventional modularization method.
Example one
Nearly 8 hundred million online netizen information records are used as test data, a dynamic structure evolution detection method of a topic network in a t period is constructed by defining a benefit function before the t period, new emerging nodes related to historical topics can be explored, all topics before epochs are detected by using a matrix, online posts in the t period are divided into two online post communities, and the matrix B is used for detecting all topics before epochs t + Obtaining new topics related to historical topics and new online posts B 1 t And B 2 t 。
The effectiveness and feasibility of the method are proved by comparing the method with the traditional topic structure evolution discovery method without considering the time-varying condition. The method can accurately divide online posts on the Twitter into ten topics, but the traditional method cannot be competent for the task, the traditional method for discovering the topic structure evolution by analyzing the coincidence degree of the posts keywords divides online post sets of the Twitter into 34 topics by mistake, the wrong division rate is very high, in the real online social network situation and scene, a plurality of online posts belonging to the same topic do not have overlapped keyword sets, but the attributes of the posts belonging to the same topic are shown by forwarding comments among the online posts. The topic discovery of the data by using the topic structure evolution discovery method not only can give consideration to the coincidence of keywords (through I in the network topic structure evolution discovery method) i(≤t)j(≤t) Showing function to reflect contact degree between posts) and forwarding comment relation (social attribute) between posts (through F in network topic structure evolution discovery method i(≤t),j(≤t) An indicative function is used for reflecting the forwarding comment relation among posts), and the result shows that the topic structure evolution discovery method can accurately discover ten topics into ten topics, as shown in fig. 1. Moreover, the network topic structure evolution discovery method can also realize the discovery of the evolution process of the network topic in different time periods from step 1 to step 7, as shown in fig. 2.
Claims (5)
1. A network topic structure evolution discovery method is characterized by comprising the following steps:
step 1, establishing a interest function B before t time period ≤t Said interest function B ≤t The evaluation function is used for representing that posts appearing in a time period t belong to emerging topics and historical topics before the time period t-delta t, and the time period t refers to the time period from the time period t-delta t to the time period t;
in the above formula, B ≤t-Δt As a function of interest before the t- Δ t period, B t As a function of interest before the t period, I i(≤t)j(≤t) An attribute representing text overlap between posts, keyword vectors representing i and j posts, respectively, F i(t)j(t) Representing a forward comment relationship between the i and j posts,indicative function of whether online post i and online post j are posts before the t-deltat period,a post that indicates that online post i belongs to the occurrence of the t period and online post j is an indicative function of the occurrence of the post prior to the t-deltat period,indicative function of whether or not the online post i and the online post j are posts occurring in the t period, b i(≤t) Representing the number of posts that belong to the same topic as the online post i, b j(≤t) Representing the number of other posts that belong to the same topic as online post j before the t time period; n is ≤t Representing the total number of degrees of all nodes in the network before the t period,an indicative function indicating that post i and post j belong to the same topic;
step 2, obtaining a interest function B before a t-delta t time period according to the formula (1) ≤t-Δt :
Obtaining historical topics before a t-delta t time period according to a formula (2);
step 3, simultaneously obtaining a interest function B of the t time period t Comprises the following steps:
acquiring newly added posts of the historical topics in the t period and newly-added topics in the t period according to a formula (3);
step 4, according to the step 2, the step 3 and the quadratic form theory, the interest function B is obtained t Conversion to a form comprising a positive definite matrix Z:
equation (4) is converted to the following form according to equation (6):
in the above formula, L 1 =Z -1 L;
By aligning the symmetric matrixCarrying out calculation and analysis on the characteristic value and the characteristic vector of the topic to obtain a new post of the historical topic in the t time period and a new topic in the t time period before the t-delta t time period;
and 5, converting the formula (2) into:
to formula (11)Analyzing the characteristic value and the characteristic vector to further realize the classification of the historical topics before the t-delta t time interval;
step 6, establishing a interest function B of the historical topics before the t-delta t time period and the topics at the t time period t + :
In the above formula, when both the online post i and the online post j belong to posts appearing in the period t, θ i,j 1, otherwise zero;
utilizing the topic classification result before the t-delta t time interval obtained in the step 5 and a profit function B t + Set omega of online posts that occur during t periods t Is divided into setsAnd collectionsThe collectionIncluding posts relating to historical topics, said collectionPosts that are unrelated to historical topics;
step 7, utilizing the interest function B 1 t For the setEach post in the system is clustered with the belonged historical topic category by utilizing a interest function B 2 t For the setEach post in the system carries out clustering of emerging topics, and the interest function B 1 t Interest function B 2 t Is represented as follows:
2. The method for discovering structure evolution of network topics according to claim 1, wherein in step 1: if the online post i and the online post j are posts before the t-delta t time period, thenOtherwise, the value is zero; if the online post i belongs to a post occurring at a time period t and the online post j is a post occurring before the time period t- Δ t, thenOtherwise, the value is zero; if it isIf the online post i and the online post j are posts appearing in the time period t, thenOtherwise it is zero.
4. The method for discovering evolution of network topic structure according to claim 1, wherein in step 1: if the i post and the j post have a forwarding comment relation, F i(t)j(t) 1 is ═ 1; otherwise F i(t)j(t) Is zero.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110211155.XA CN112818125B (en) | 2021-02-25 | 2021-02-25 | Network topic structure evolution discovery method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110211155.XA CN112818125B (en) | 2021-02-25 | 2021-02-25 | Network topic structure evolution discovery method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112818125A CN112818125A (en) | 2021-05-18 |
CN112818125B true CN112818125B (en) | 2022-09-09 |
Family
ID=75865649
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110211155.XA Active CN112818125B (en) | 2021-02-25 | 2021-02-25 | Network topic structure evolution discovery method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112818125B (en) |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU2019100854A4 (en) * | 2019-08-02 | 2019-09-05 | Xi’an University of Technology | Long-term trend prediction method based on network hotspot single-peak topic propagation model |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104408108B (en) * | 2014-11-18 | 2018-08-14 | 重庆邮电大学 | Much-talked-about topic group influence power analysis system based on gray system theory and method |
CN104516961A (en) * | 2014-12-18 | 2015-04-15 | 北京牡丹电子集团有限责任公司数字电视技术中心 | Topic digging and topic trend analysis method and system based on region |
US20170083817A1 (en) * | 2015-09-23 | 2017-03-23 | Isentium, Llc | Topic detection in a social media sentiment extraction system |
CN108399241B (en) * | 2018-02-28 | 2021-08-31 | 福州大学 | Emerging hot topic detection system based on multi-class feature fusion |
-
2021
- 2021-02-25 CN CN202110211155.XA patent/CN112818125B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU2019100854A4 (en) * | 2019-08-02 | 2019-09-05 | Xi’an University of Technology | Long-term trend prediction method based on network hotspot single-peak topic propagation model |
Non-Patent Citations (1)
Title |
---|
在线社交网络中的新兴话题检测技术综述;笱程成等;《中文信息学报》;20160915(第05期);13-22页 * |
Also Published As
Publication number | Publication date |
---|---|
CN112818125A (en) | 2021-05-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Xue et al. | Detecting fake news by exploring the consistency of multimodal data | |
Boididou et al. | Verifying information with multimedia content on twitter: a comparative study of automated approaches | |
Spinde et al. | Neural Media Bias Detection Using Distant Supervision With BABE--Bias Annotations By Experts | |
Abbasi et al. | Descriptive analytics: Examining expert hackers in web forums | |
Wang et al. | Burst time prediction in cascades | |
CN107835113A (en) | Abnormal user detection method in a kind of social networks based on network mapping | |
CN108228853A (en) | A kind of microblogging rumour recognition methods and system | |
Feng et al. | Computational social indicators: a case study of chinese university ranking | |
CN114138968B (en) | Network hotspot mining method, device, equipment and storage medium | |
CN115277888B (en) | Method and system for analyzing message type of mobile application encryption protocol | |
CN117874662A (en) | Micro-service log anomaly detection method based on graph mode | |
CN117828029A (en) | Multi-domain false news detection method based on emotion-style depolarization | |
Zhao et al. | Anomaly detection of aircraft lead‐acid battery | |
CN112818125B (en) | Network topic structure evolution discovery method | |
CN116541792A (en) | Method for carrying out group partner identification based on graph neural network node classification | |
Dong et al. | Online Burst Events Detection Oriented Real-Time Microblog Message Stream. | |
Fang et al. | Log Anomaly Detection Based on Hierarchical Graph Neural Network and Label Contrastive Coding. | |
Han et al. | Distributed system anomaly detection using deep learning‐based log analysis | |
CN114969761A (en) | Log anomaly detection method based on LDA theme characteristics | |
Fazelnia et al. | Lessons from the Use of Natural Language Inference (NLI) in Requirements Engineering Tasks | |
Guo et al. | Detecting spammers in E-commerce website via spectrum features of user relation graph | |
CN115964478A (en) | Network attack detection method, model training method and device, equipment and medium | |
Roy Chowdhury et al. | D-sieve: a novel data processing engine for efficient handling of crises-related social messages | |
Lv et al. | HGE2MED: Heterogeneous graph embedding for multi-domain event detection | |
CN112445939A (en) | Social network group discovery system, method and storage medium |
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 |