CN111160483B - Network relation type prediction method based on multi-classifier fusion model - Google Patents
Network relation type prediction method based on multi-classifier fusion model Download PDFInfo
- Publication number
- CN111160483B CN111160483B CN201911414801.1A CN201911414801A CN111160483B CN 111160483 B CN111160483 B CN 111160483B CN 201911414801 A CN201911414801 A CN 201911414801A CN 111160483 B CN111160483 B CN 111160483B
- Authority
- CN
- China
- Prior art keywords
- node
- network
- data
- learner
- basemodel
- 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 43
- 230000004927 fusion Effects 0.000 title claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 35
- 239000013598 vector Substances 0.000 claims abstract description 23
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 238000012512 characterization method Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000005295 random walk Methods 0.000 claims description 12
- 238000002790 cross-validation Methods 0.000 claims description 9
- 238000007637 random forest analysis Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000000638 solvent extraction Methods 0.000 claims description 3
- 230000007704 transition Effects 0.000 claims description 3
- 230000001276 controlling effect Effects 0.000 claims 1
- 230000001105 regulatory effect Effects 0.000 claims 1
- 238000004458 analytical method Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 238000012407 engineering method Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000000813 microbial effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Medical Informatics (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a network relation type prediction method based on a multi-classifier fusion model. The method comprises the following steps: preprocessing network data, including extracting network structure from semi-structured data, and converting into an edge list or an adjacent list; performing Node characterization on the side list or the adjacent list by using a Node2Vec network embedding method to obtain k-dimensional feature vector representation of each Node in the network; obtaining a 3 xk-dimensional sample feature vector by splicing two adjacent node feature vectors and a difference vector of the two node feature vectors, and dividing a training set and a test set; taking the classification method after parameter adjustment as a base learner to make decisions; and fusing the classification result of the base learner as the input data of the meta learner, and outputting the final network relationship type prediction result. The method is suitable for common social network data, and improves the accuracy of prediction.
Description
Technical Field
The invention belongs to the technical field of computers, particularly relates to the technical field of link prediction in a complex network, and relates to a network relation type prediction method based on a multi-classifier fusion model.
Background
In the society with complex relationships, the structures of biomolecules are abstracted into network representations, most of real network data are missing, for example, the problem of sampling or hiding human relationships in the social relationship network exists, and the field of microbial medicine needs a lot of experiments to infer the interaction between cells or tissues. According to different research problems in different fields, a network structure is constructed, and 'possible connection' or 'missing connection' in the network structure is analyzed and mined, so that new findings can be brought to different fields, and the development of the corresponding fields is promoted.
The definition of link prediction in the network is that "the possibility that a connection exists between two nodes in the network is not generated through information such as network structure and network attributes prediction". The link prediction algorithm comprehensively uses multidisciplinary methods such as similarity analysis, network dynamics, bayesian models, machine learning, motif analysis, maximum likelihood analysis and the like.
In the field of social networks, most researchers tend to mine a connection relationship between two nodes in a network besides the possibility of whether the two nodes have a relationship, network relationship type prediction is applied to scenes such as recommendation, advertisement and community detection, and the connection relationship type in the social network is divided into "+" and "-" which respectively represent the quality of the relationship, the status or the skill level, and the like. There have been related researchers applying balance theory to network relationship type prediction, i.e., "friend of friend is my friend, friend of enemy is my enemy, friend of friend is my enemy, and enemy of enemy is my friend". The balance theory aims to explore the characteristics of a triangular network structure, capture the relation among three users and predict a positive chain and a negative chain. However, in reality, there are often a large number of "bridge" edges, i.e. the connecting edges between two adjacent nodes, and the two adjacent nodes do not have any common neighbors, whereas a "triangle" edge is composed of two adjacent nodes containing at least one common neighbor.
Although most edges in a directed network with relationships can be contained in a triangle, a "bridge" edge also exists globally. According to the balance theory, the nodes in the triangle are connected with each other, and the network relation type prediction of the edges in the triangle can be effectively carried out. The "bridge" edge lacks triangle information and cannot be modeled by equilibrium theory. Therefore, performing network relationship type prediction on "bridge" edges is a challenging task.
Disclosure of Invention
The invention aims to provide a network relation type prediction method based on a multi-classifier fusion model aiming at the defects of the prior art.
The method provided by the invention is characterized in that social network public data Slashdot, relationships and Wikirfa with relationships are utilized, a Node2Vec method in a network embedding method and the idea of a fusion model in machine learning are adopted, a set of social network-based relationship type prediction method is designed, and Roc _ auc, binary _ F1, macro _ F1 and Micro _ F1 are all expressed on a network data set with a bridge edge.
The method specifically comprises the following steps:
step (1), preprocessing network data, specifically:
(1-1) network data structured representation: the network data comprises structured network data and semi-structured network data, and the semi-structured network data is converted into the structured network data { (v) s1 ,v e1 ,flag 1 ),(v s2 ,v e2 ,flag 2 ),…,(v si ,v ei ,flag i ),…,(v sn ,v en ,flag n ) }; where n represents the number of network edges, v si 、v ei Respectively representing start and end nodes, sample tag flag i Equal to 1 or-1, indicating that the actual relationship type is "friendly" or "hostile", i =1,2, …, n.
(1-2) structured network data standardized representation: converting the structured network data into an edge list or an adjacency list;
list of edges: { (v) s1 ,v e1 ),(v s2 ,v e2 ),…,(v si ,v ei ),…,(v sn ,v en )}，(v si ,v ei ) Denotes v si 、v ei A connecting edge exists;
adjacency list:
{(v s1 ,v s1-e1 ,v s1-e2 ,…,v s1-ei ,…,v s1-en ),(v s2 ,v s2-e1 ,v s2-e2 ,…,v s2-ei ,…,v s2-en ),…,(v sk ,v sk-e1 ,v sk-e2 ,…,v sk-ei ,…,v sk-en ) }; k represents the dimensions of each node in the network.
Step (2), the edge list or the adjacent list carries out Node characterization by using a Node2Vec network embedding method to obtain k-dimensional feature vector representation of each Node in the network, and the method specifically comprises the following steps:
the biased random walk method is adopted to regulate the walk probability from one node to the next node through the parameters p (return parameter), q (in-out parameter).
Given node v, the probability of selecting the next node x by random walk is:
And (3) learning the sampled vertex sequence by using the Node2Vec to obtain a network characterization vector set of the nodes: { (f) v1-1 ,f v1-2 ,…,f v1-k ),(f v2-1 ,f v2-2 ,…,f v2-k ),…,(f vi-1 ,f vi-2 ,…,f vi-k ),…,(f vn-1 ,v vn-2 ,…,f vn-k )}；(f vi-1 ,f vi-2 ,…,f vi-k ) Representing a representation of the k-dimensional feature vector corresponding to node i in the network.
And (3) carrying out network characteristic engineering processing on the network characterization vector set, and dividing a training set and a test set, wherein the method specifically comprises the following steps:
(3-1) network feature engineering: splicing two adjacent node feature vectors and a difference vector of the two node feature vectors to obtain a 3 xk-dimensional sample feature directionVolume, i.e. structured network data conversion GraphData = { simple 1 ,…,simple i ,…simple n In which sample simple i ＝([f si-1 ,f si-2 ,…,f si-k ],[f ei-1 ,f ei-2 ,…,f ei-k ],[f si-1 -f ei-1 ,f si-2 -f ei-2 ,…,f si-k -f ei-k ],flag i ) Mapping (v) in the structured network data in step (1-1) si ,v ei ,flag i )；
(3-2) data set partitioning: and randomly extracting GraphData, wherein 70-80% of data is used as a training set, and 20-30% of data is used as a testing set.
And (4) respectively carrying out super-parameter optimization on ExtraTrees, gradienBoosting, lightGBM and XGBboost base learners on a training set by using a GridCV parameter adjusting method: respectively obtaining super-parameter combinations of ExtraTrees, gradienenBoosting, lightGBM and XGBboost through GridCV calculation, initializing each base learner by using the super-parameter combinations, and obtaining a tuning model BaseModel 1 、BaseModel 2 、BaseModel 3 、BaseModel 4 。
And (5) using the obtained tuning model as a base learner, predicting a training set by adopting a K-fold cross validation method, fusing a prediction result of the base learner as input data of a meta learner RandomForest, and outputting a final network relationship type prediction result, wherein the method specifically comprises the following steps:
(5-1). K-fold cross validation base learner: respectively to BaseModel on training set 1 、BaseModel 2 、BaseModel 3 、BaseModel 4 Performing K-fold cross validation, namely dividing the training set into K parts to generate training subsets, and obtaining the prediction sets R of the training subsets by the four base learners 1 、R 2 、R 3 、R 4 ；
(5-2) meta learner training: will predict set R 1 、R 2 、R 3 、R 4 Splicing to obtain a new training set, training by using RandomForest as a meta-learner, and combining the trained base learner and the meta-learner to obtain a final fusion model;
(5-3) predicting results and evaluating: and predicting the test set by using the Stackingmodel, and evaluating the effect of the prediction result by using Roc _ auc, binary _ F1, macro _ F1 and Micro _ F1 as model performance evaluation indexes.
The method effectively recommends a network relation type prediction method based on a multi-classifier fusion model, and improves the prediction accuracy. The invention creatively provides a network characteristic engineering method, which comprises the characteristics of node vector difference and the characteristics of nodes, and creatively utilizes the idea of a fusion model in a network relation type prediction task. The method can better serve the aspects of actual recommendation, advertisement, community discovery and the like.
Detailed Description
The technical solution of the present invention is further explained below.
A network relation type prediction method based on a multi-classifier fusion model comprises the following specific steps:
step (1), preprocessing network data, specifically:
(1-1) network data structured representation: the network data has structured network data and semi-structured network data, and the semi-structured network data is converted into the structured network data { (v) s1 ,v e1 ,flag 1 ),(v s2 ,v e2 ,flag 2 ),…,(v si ,v ei ,flag i ),…,(v sn ,v en ,flag n ) }; where n represents the number of network edges, v si 、v ei Respectively representing start and end nodes, sample label flag i Equal to 1 or-1, indicating that the actual relationship type is "friendly" or "hostile", i =1,2, …, n.
(1-2) structured network data standardized representation: converting the structured network data into an edge list or an adjacency list;
list of edges: { (v) s1 ,v e1 ),(v s2 ,v e2 ),…,(v si ,v ei ),…,(v sn ,v en )}，(v si ,v ei ) Denotes v si 、v ei A connecting edge exists;
adjacency list:
{(v s1 ,v s1-e1 ,v s1-e2 ,…,v s1-ei ,…,v s1-en ),(v s2 ,v s2-e1 ,v s2-e2 ,…,v s2-ei ,…,v s2-en ),…,(v sk ,v sk-e1 ,v sk-e2 ,…,v sk-ei ,…,v sk-en ) }; k denotes the dimension of each node in the network.
Step (2), the edge list or the adjacent list carries out Node characterization by using a Node2Vec network embedding method to obtain k-dimensional feature vector representation of each Node in the network, and the method specifically comprises the following steps:
the biased random walk method is adopted to regulate the walk probability from one node to the next node through the parameters p (return parameter), q (in-out parameter).
Given node v, the probability of selecting the next node x by random walk is:
And (3) learning the sampled vertex sequence by using the Node2Vec to obtain a network characterization vector set of the nodes: { (f) v1-1 ,f v1-2 ,…,f v1-k ),(f v2-1 ,f v2-2 ,…,f v2-k ),…,(f vi-1 ,f vi-2 ,…,f vi-k ),…,(f vn-1 ,v vn-2 ,…,f vn-k )}；(f vi-1 ,f vi-2 ,…,f vi-k ) Representing a representation of the k-dimensional feature vector corresponding to node i in the network.
Step (3), carrying out network characteristic engineering processing on the network characterization vector set, and dividing a training set and a test set, wherein the method specifically comprises the following steps:
(3-1) network feature engineering: splicing two adjacent node eigenvectors and the difference vector of the two node eigenvectors to obtain a 3 x k dimensional sample eigenvector, namely, the structured network data conversion GraphData = { simple = 1 ,…,simple i ,…simple n In which sample simple i ＝([f si-1 ,f si-2 ,…,f si-k ],[f ei-1 ,f ei-2 ,…,f ei-k ],[f si-1 -f ei-1 ,f si-2 -f ei-2 ,…,f si-k -f ei-k ],flag i ) Mapping (v) in the structured network data in step (1-1) si ,v ei ,flag i )；
(3-2) data set partitioning: the GraphData is randomly extracted, 70-80% of data is used as a training set, 20-30% of data is used as a test set, 80% of data is extracted as the training set and 20% of data is used as the test set in the embodiment.
And (4) respectively carrying out super-parameter optimization on ExtraTrees, gradienBoosting, lightGBM and XGBboost base learners on a training set by using a GridCV parameter adjusting method: respectively obtaining super-parameter combinations of ExtraTrees, gradienenBoosting, lightGBM and XGBboost through GridCV calculation, initializing each base learner by using the super-parameter combinations, and obtaining a tuning model BaseModel 1 、BaseModel 2 、BaseModel 3 、BaseModel 4 。
And (5) using the obtained tuning model as a base learner, predicting a training set by adopting a K-fold cross validation method, fusing a prediction result of the base learner as input data of a meta learner RandomForest, and outputting a final network relationship type prediction result, wherein the method specifically comprises the following steps:
(5-1). K-fold cross validation base learner: respectively to BaseModel on training set 1 、BaseModel 2 、BaseModel 3 、BaseModel 4 Performing K-fold cross validation, namely dividing the training set into K parts to generate a training subset, and obtaining a prediction set R of the training subset by the four base learners 1 、R 2 、R 3 、R 4 ；
(5-2) meta learner training: will predict set R 1 、R 2 、R 3 、R 4 Splicing to obtain a new training set, training by using RandomForest as a meta-learner, and combining the trained base learner and the meta-learner to obtain a final fusion model;
(5-3) predicting results and evaluating: and predicting the test set through the StackingModel, introducing Roc _ auc, binary _ F1, macro _ F1 and Micro _ F1 as model performance evaluation indexes, and evaluating the effect of a prediction result.
By applying the social network data Slashdot, epinions and Wikirfa, the task of predicting the network relationship type on a large-scale network can be verified, for example, evaluation indexes of Roc _ auc, binary _ F1, macro _ F1 and Micro _ F1 on a Slashdot data set are 91.22%,91.82%,80.13% and 87.01%, and four index scores are all superior to those of the existing method, and the prediction method is generally applicable to the network relationship type prediction problem.
Claims (1)
1. A network relation type prediction method based on a multi-classifier fusion model is characterized by comprising the following steps:
step (1), preprocessing network data, specifically:
(1-1) network data structured representation: the network data comprises structured network data and semi-structured network data, and the semi-structured network data is converted into the structured network data { (v) s1 ,v e1 ,flag 1 ),(v s2 ,v e2 ,flag 2 ),…,(v si ,v ei ,flag i ),…,(v sn ,v en ,flag n ) }; where n represents the number of network edges, v si 、v ei Respectively representing start and end nodes, sample tag flag i Equal to 1 or-1, indicating that the actual relationship type is friendly or hostile, i =1,2, …, n;
(1-2) structured network data standardized representation: converting the structured network data into an edge list or an adjacency list;
list of edges: { (v) s1 ,v e1 ),(v s2 ,v e2 ),…,(v si ,v ei ),…,(v sn ,v en )}，(v si ,v ei ) Denotes v si 、v ei A connecting edge exists;
adjacency list:
{(v s1 ,v s1-e1 ,v s1-e2 ,…,v s1-ei ,…,v s1-en ),(v s2 ,v s2-e1 ,v s2-e2 ,…,v s2-ei ,…,v s2-en ),…,(v sk ,v sk-e1 ,v sk-e2 ,…,v sk-ei ,…,v sk-en ) }; k represents the dimension of each node in the network;
step (2), the edge list or the adjacent list carries out Node characterization by using a Node2Vec network embedding method to obtain k-dimensional feature vector representation of each Node in the network, and the method specifically comprises the following steps:
adopting a biased random walk method, and regulating and controlling the walk probability from one node to the next node through parameters p and q; given node v, the probability of selecting the next node x by random walk is:
and (3) learning the sampled vertex sequence by using the Node2Vec to obtain a network characterization vector set of the nodes: { (f) v1-1 ,f v1-2 ,…,f v1-k ),(f v2-1 ,f v2-2 ,…,f v2-k ),…,(f vi-1 ,f vi-2 ,…,f vi-k ),…,(f vn-1 ,v vn-2 ,…,f vn-k )}；(f vi-1 ,f vi-2 ,…,f vi-k ) Representing k-dimensional feature vectors corresponding to nodes i in the network;
step (3), carrying out network characteristic engineering processing on the network characterization vector set, and dividing a training set and a test set, wherein the method specifically comprises the following steps:
(3-1) network feature engineering: splicing two adjacent node eigenvectors and the difference vector of the two node eigenvectors to obtain a 3 xk-dimensional sample eigenvector, namely structured network data conversion GraphData = { simple 1 ,…,simple i ,…simple n In which sample simple i ＝([f si-1 ,f si-2 ,…,f si-k ],[f ei-1 ,f ei-2 ,…,f ei-k ],[f si-1 -f ei-1, f si-2 -f ei-2 ,…,f si-k -f ei-k ],flag i ) Mapping (v) in the structured network data in step (1-1) si ,v ei ,flag i )；
(3-2) data set partitioning: randomly extracting GraphData, wherein 70-80% of data is used as a training set, and 20-30% of data is used as a testing set;
and (4) respectively carrying out super-parameter optimization on ExtraTrees, gradienBoosting, lightGBM and XGBboost base learners on a training set by using a GridCV parameter adjusting method: respectively obtaining super-parameter combinations of ExtraTrees, gradienenBoosting, lightGBM and XGBboost through GridCV calculation, initializing each base learner by using the super-parameter combinations, and obtaining a tuning model BaseModel 1 、BaseModel 2 、BaseModel 3 、BaseModel 4 ；
And (5) using the obtained tuning model as a base learner, predicting a training set by adopting a K-fold cross validation method, fusing a prediction result of the base learner as input data of a meta learner RandomForest, and outputting a final network relationship type prediction result, wherein the method specifically comprises the following steps:
(5-1). K-fold cross validation base learner: respectively to BaseModel on training set 1 、BaseModel 2 、BaseModel 3 、BaseModel 4 Performing K-fold cross validation, namely dividing the training set into K parts to generate a training subset, and obtaining a prediction set R of the training subset by the four base learners 1 、R 2 、R 3 、R 4 ；
(5-2) meta learner training: will predict set R 1 、R 2 、R 3 、R 4 Splicing to obtain a new training set, training by using RandomForest as a meta-learner, and combining the trained base learner and the meta-learner to obtain a final fusion model Stacking model;
(5-3) predicting results and evaluating: and predicting the test set by using a Stackingmodel, and evaluating the effect of the prediction result by using Roc _ auc, binary _ F1, macro _ F1 and Micro _ F1 as model performance evaluation indexes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911414801.1A CN111160483B (en) | 2019-12-31 | 2019-12-31 | Network relation type prediction method based on multi-classifier fusion model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911414801.1A CN111160483B (en) | 2019-12-31 | 2019-12-31 | Network relation type prediction method based on multi-classifier fusion model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111160483A CN111160483A (en) | 2020-05-15 |
| CN111160483B true CN111160483B (en) | 2023-03-17 |
Family
ID=70560071
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911414801.1A Active CN111160483B (en) | 2019-12-31 | 2019-12-31 | Network relation type prediction method based on multi-classifier fusion model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111160483B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111695042B (en) * | 2020-06-10 | 2023-04-18 | 湖南湖大金科科技发展有限公司 | User behavior prediction method and system based on deep walking and ensemble learning |
| CN113095440B (en) * | 2020-09-01 | 2022-05-17 | 电子科技大学 | Training data generation method based on metalearner and causal effect heterogeneous reaction difference estimation method |
| CN113869609A (en) * | 2021-10-29 | 2021-12-31 | 北京宝兰德软件股份有限公司 | Method and system for predicting confidence of frequent subgraph of root cause analysis |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8595167B1 (en) * | 2010-11-30 | 2013-11-26 | Google Inc. | Predicting likelihood of a successful connection between unconnected users within a social network using a learning network |
| CN107622307A (en) * | 2017-09-11 | 2018-01-23 | 浙江工业大学 | A kind of Undirected networks based on deep learning connect side right weight Forecasting Methodology |
| DE102018202875A1 (en) * | 2017-02-28 | 2018-08-30 | NEC Laboratories Europe GmbH | System and method for multimodal graph based personalization |
| CN108596774A (en) * | 2018-04-24 | 2018-09-28 | 山东师范大学 | Socialization information recommendation algorithm based on profound internet startup disk feature and system |
| CN108647800A (en) * | 2018-03-19 | 2018-10-12 | 浙江工业大学 | A kind of online social network user missing attribute forecast method based on node insertion |
| CN110489567A (en) * | 2019-08-26 | 2019-11-22 | 重庆邮电大学 | A kind of node information acquisition method and its device based on across a network Feature Mapping |
-
2019
- 2019-12-31 CN CN201911414801.1A patent/CN111160483B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8595167B1 (en) * | 2010-11-30 | 2013-11-26 | Google Inc. | Predicting likelihood of a successful connection between unconnected users within a social network using a learning network |
| DE102018202875A1 (en) * | 2017-02-28 | 2018-08-30 | NEC Laboratories Europe GmbH | System and method for multimodal graph based personalization |
| CN107622307A (en) * | 2017-09-11 | 2018-01-23 | 浙江工业大学 | A kind of Undirected networks based on deep learning connect side right weight Forecasting Methodology |
| CN108647800A (en) * | 2018-03-19 | 2018-10-12 | 浙江工业大学 | A kind of online social network user missing attribute forecast method based on node insertion |
| CN108596774A (en) * | 2018-04-24 | 2018-09-28 | 山东师范大学 | Socialization information recommendation algorithm based on profound internet startup disk feature and system |
| CN110489567A (en) * | 2019-08-26 | 2019-11-22 | 重庆邮电大学 | A kind of node information acquisition method and its device based on across a network Feature Mapping |
Non-Patent Citations (3)
| Title |
|---|
| Effective Model Integration Algorithm for Improving Link and Sign Prediction in Complex Networks;Chuang Liu et al.;《 IEEE Transactions on Network Science and Engineering》;20210730;第8卷(第3期);全文 * |
| node2vec: Scalable Feature Learning for Networks;Aditya Grover et al.;《In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining》;20160703;全文 * |
| 大规模复杂信息网络表示学习:概念、方法与挑战;齐金山等;《计算机学报》;20171220(第10期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111160483A (en) | 2020-05-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111160483B (en) | Network relation type prediction method based on multi-classifier fusion model | |
| CN106779087B (en) | A kind of general-purpose machinery learning data analysis platform | |
| He et al. | Modeling the urban landscape dynamics in a megalopolitan cluster area by incorporating a gravitational field model with cellular automata | |
| Do et al. | A neuro‐fuzzy approach in the classification of students’ academic performance | |
| Dyer | Is there such a thing as landscape genetics? | |
| CN117237559B (en) | Digital twin city-oriented three-dimensional model data intelligent analysis method and system | |
| Rustia et al. | Predicting student's board examination performance using classification algorithms | |
| He et al. | Deriving urban dynamic evolution rules from self-adaptive cellular automata with multi-temporal remote sensing images | |
| CN115019123B (en) | Self-distillation contrast learning method for remote sensing image scene classification | |
| CN111008337A (en) | Deep attention rumor identification method and device based on ternary characteristics | |
| CN114819056B (en) | Single-cell data integration method based on domain countermeasure and variation inference | |
| Liu et al. | Using the ART-MMAP neural network to model and predict urban growth: a spatiotemporal data mining approach | |
| CN114897085A (en) | Clustering method based on closed subgraph link prediction and computer equipment | |
| Rastogi et al. | GA based clustering of mixed data type of attributes (numeric, categorical, ordinal, binary and ratio-scaled) | |
| CN103559642A (en) | Financial data mining method based on cloud computing | |
| Wang et al. | Learning visual features from figure-ground maps for urban morphology discovery | |
| Li et al. | Zero-shot learning for intrusion detection via attribute representation | |
| Kashyap et al. | Multi-objective metaheuristic search algorithms for service composition in IoT | |
| CN109495316B (en) | Network characterization method fusing adjacency and node role similarity | |
| Zhang et al. | Wild plant data collection system based on distributed location | |
| CN115205606A (en) | Image multi-label classification method and device and related products | |
| Hanna | Spectral comparison of large urban graphs | |
| Laureano et al. | Affinity propagation SMOTE approach for imbalanced dataset used in predicting student at risk of low performance | |
| Belderbos et al. | GANs Based Conditional Aerial Images Generation for Imbalanced Learning | |
| Ouyang et al. | Prediction of graduation development Based on Hypergraph Contrastive Learning with Imbalanced Sampling |
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 |