CN114548569A - Missing link prediction method, system and storage medium in heterogeneous social network - Google Patents
Missing link prediction method, system and storage medium in heterogeneous social network Download PDFInfo
- Publication number
- CN114548569A CN114548569A CN202210175814.3A CN202210175814A CN114548569A CN 114548569 A CN114548569 A CN 114548569A CN 202210175814 A CN202210175814 A CN 202210175814A CN 114548569 A CN114548569 A CN 114548569A
- Authority
- CN
- China
- Prior art keywords
- link
- prediction
- social network
- predictor
- loss
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 80
- 238000003860 storage Methods 0.000 title claims abstract description 17
- 239000013598 vector Substances 0.000 claims abstract description 62
- 230000006870 function Effects 0.000 claims abstract description 60
- 230000008569 process Effects 0.000 claims abstract description 24
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 19
- 238000005457 optimization Methods 0.000 claims abstract description 19
- 230000002776 aggregation Effects 0.000 claims abstract description 18
- 238000004220 aggregation Methods 0.000 claims abstract description 18
- 230000015654 memory Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 7
- 230000000694 effects Effects 0.000 abstract description 8
- 238000000605 extraction Methods 0.000 description 6
- 230000003042 antagnostic effect Effects 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- 230000009471 action Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000003012 network analysis Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- FNMKZDDKPDBYJM-UHFFFAOYSA-N 3-(1,3-benzodioxol-5-yl)-7-(3-methylbut-2-enoxy)chromen-4-one Chemical compound C1=C2OCOC2=CC(C2=COC=3C(C2=O)=CC=C(C=3)OCC=C(C)C)=C1 FNMKZDDKPDBYJM-UHFFFAOYSA-N 0.000 description 1
- 208000019901 Anxiety disease Diseases 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000036506 anxiety Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000593 degrading effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 230000007723 transport mechanism Effects 0.000 description 1
Images
Classifications
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- General Business, Economics & Management (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Entrepreneurship & Innovation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Development Economics (AREA)
- Evolutionary Biology (AREA)
- Primary Health Care (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a method, a system and a storage medium for predicting missing links in a heterogeneous social network, which comprises the following steps: converting links of the heterogeneous social network into structural feature vectors; inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization; performing preliminary link prediction on the link samples of the structural feature vectors by using a generated predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using a discriminant classifier; adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier; and acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discriminant classifier, and predicting the missing link of the new type according to the final public characteristics. The method can more accurately and comprehensively generate a good prediction effect on the link prediction of the new type.
Description
Technical Field
The invention relates to the technical field of prediction of missing links of heterogeneous social networks, in particular to a method, a system and a storage medium for predicting missing links in heterogeneous social networks.
Background
The missing link prediction means how to predict the possibility of generating a link between two nodes which are not connected in the network through known network nodes, network structures and other information in the social network. Such predictions include both predictions of unknown links and predictions of future links.
Social network analysis has had great success in various research areas, from population planning queries to information diffusion, from impact maximization to assessing public anxiety. Among them, the study of heterogeneous social network missing link prediction becomes a unique challenge. Heterogeneous social networks are typically represented as a general network graph, where nodes represent individuals belonging to different categories and links represent different types of interactions. In practical applications, it is difficult to construct a complete network diagram to represent the entire heterogeneous social network by observing all existing connections due to inaccurate information, unresponsiveness of individuals, sampling bias, and the like, and particularly, existing links that are not observed in the network structure are referred to as missing links. In practical application, missing links often exist in the collected network graph, which affects the integrity of the heterogeneous social network, and leads to wrong conclusion obtained by social network analysis.
Disclosure of Invention
The invention aims to overcome the technical defects, provides a method for predicting missing links in a heterogeneous social network, can more accurately and comprehensively generate a good prediction effect on the new link prediction, and has a wide action surface and stronger applicability on the new link prediction by learning of shared characteristics through an antagonistic neural network.
In order to achieve the above technical objective, a first aspect of the present invention provides a method for predicting a missing link in a heterogeneous social network, including the following steps:
converting the link of the heterogeneous social network into a structural feature vector;
inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization;
performing preliminary link prediction on the link samples of the structural feature vectors by using a generation predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using a discrimination classifier;
adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier;
and acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discrimination classifier, and predicting the missing link of the new type according to the final public characteristics.
Compared with the prior art, the invention has the beneficial effects that:
according to the method for predicting the missing link in the heterogeneous social network, provided by the technical scheme, the heterogeneous social network is subjected to primary feature extraction; then, carrying out preliminary aggregation and optimization on the extracted features through a convolutional neural network; and then respectively carrying out antagonistic training of two processes to obtain a final common characteristic for link prediction. The first process is a process of performing link prediction through training of a full connection layer, which is equivalent to a two-classification process, and the second process is a process of performing auxiliary link type classification through training of a gradient inversion layer and the full connection layer, which is equivalent to a multi-classification process, so that the auxiliary training is used for obtaining common characteristics to perform link prediction. By using the specific means of countertraining and the mode of transfer learning, the common features are extracted from the features of the historical links, and the method has extremely important significance for detecting the missing links of the new type in the heterogeneous social network.
Compared with other link prediction methods, the missing link prediction method in the heterogeneous social network is based on feature extraction, and predicts the new-appearing type link without prior type by paying attention to the difference of link types in the heterogeneous social network. The method can accurately and comprehensively generate a good prediction effect on the link prediction of the new type, and the antagonistic neural network is adopted for learning of shared characteristics, so that the method has a wide action range on the link prediction of the new type and has stronger applicability.
According to some embodiments of the invention, the heterogeneous social network comprises a plurality of nodes, links are formed between the nodes, the link sample e is a link between the node u and the node v, and the characteristic representation function of the link sample e is as follows:
r(e)＝f(u)*f(v)
wherein, f (u) is a feature expression function of the node u, and f (v) is a feature expression function of the node v.
According to some embodiments of the invention, the preliminary link prediction of the link samples of the structural feature vector using a generated predictor comprises the steps of:
inputting a set of link samples into the generative predictor;
calculating the possibility that the link samples in the link sample set are real links by using the generation predictor;
and calculating the prediction loss of the generation predictor according to the link sample set, the possibility that the link samples are real links and the attributes of the link samples.
According to some embodiments of the invention, the prediction penalty of the generated predictor is expressed as:
wherein S is the link sample set, P (e) is the probability that the link sample is a real link, meFor the attributes of the link samples, me∈{0,1}，me1 indicates that the link sample e is a positive sample, me0 means that the link sample e is a negative sample.
According to some embodiments of the invention, classifying the link samples of the structural feature vector using a discriminative classifier comprises:
calculating to obtain a feature representation set of all links in the link sample set according to the link sample set;
based on the feature representation sets of all the links, the discriminative classifier classifies the links in the link sample set into corresponding history types.
According to some embodiments of the invention, the classification penalty of the discriminative classifier is expressed as:
wherein, TpA plurality of link types are included and,
indicating that the predicted link sample e is at TpIn is (a)1,a2) E has a link type of (a)1,a2) Then n ise1 is ═ 1; e link type is not (a)1,a2),ne＝0。
In a second aspect, an embodiment of the present invention provides a system for predicting missing links in a heterogeneous social network, including:
the characteristic extractor is used for converting the links of the heterogeneous social network into structural characteristic vectors;
the convolutional neural network is used for carrying out feature aggregation and feature optimization on the structural feature vector;
generating a predictor for performing preliminary link prediction on the link samples of the structural feature vector;
the discrimination classifier is used for classifying the link samples of the structural feature vectors;
the parameter adjusting module is used for adjusting parameters of a loss function of the generated predictor so as to reduce the prediction loss of the generated predictor and adjusting parameters of a loss function of the discriminant classifier so as to reduce the classification loss of the discriminant classifier;
and the missing link prediction module is used for acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discrimination classifier and predicting the missing link of a new type according to the final public characteristics.
In a third aspect, a technical solution of the present invention provides a missing link prediction system in a heterogeneous social network, including: a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method for missing link prediction in a heterogeneous social network according to any one of the first aspect when executing the computer program.
In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions for causing a computer to perform the method for predicting a missing link in a heterogeneous social network according to any one of the first aspect.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which the abstract is to be fully consistent with one of the figures of the specification:
FIG. 1 is a flowchart of a method for missing link prediction in a heterogeneous social network according to an embodiment of the present invention;
FIG. 2 is a flowchart of a method for missing link prediction in a heterogeneous social network according to another embodiment of the present invention;
fig. 3 is a flowchart of a method for predicting a missing link in a heterogeneous social network according to another embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
It should be noted that although functional block divisions are provided in the system drawings and logical orders are shown in the flowcharts, in some cases, the steps shown and described may be performed in different orders than the block divisions in the systems or in the flowcharts. The terms first, second and the like in the description and in the claims, and the drawings described above, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
The invention provides a method for predicting missing links in a heterogeneous social network, which comprises the steps of carrying out primary feature extraction on the heterogeneous social network; then, carrying out preliminary aggregation and optimization on the extracted features through a convolutional neural network; and then performing two processes of countertraining respectively to obtain a final common characteristic for link prediction. Based on feature extraction, the link of the new occurrence type without prior type is predicted by paying attention to the difference of the link types in the heterogeneous social network. The method can accurately and comprehensively generate a good prediction effect on the link prediction of the new type, and the antagonistic neural network is adopted for learning of shared characteristics, so that the method has a wide action range on the link prediction of the new type and has stronger applicability.
The embodiments of the present invention will be further explained with reference to the drawings.
Referring to fig. 1, fig. 1 is a flowchart of a method for predicting a missing link in a heterogeneous social network according to an embodiment of the present invention, where the method for predicting a missing link in a heterogeneous social network includes, but is not limited to, steps S110 to S150.
Step S110, converting links of the heterogeneous social network into structural feature vectors;
step S120, inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization;
step S130, performing preliminary link prediction on the link samples of the structural feature vectors by using the generated predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using a discrimination classifier;
step S140, adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier;
and S150, acquiring final public characteristics in the parameter adjustment process of the loss functions of the generated predictor and the discriminant classifier, and predicting the missing link of the new type according to the final public characteristics.
In one embodiment, the missing link prediction method in the heterogeneous social network comprises the following steps: converting links of the heterogeneous social network into structural feature vectors; inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization; performing preliminary link prediction on the link samples of the structural feature vectors by using a generated predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using a discriminant classifier; adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier; and acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discriminant classifier, and predicting the missing link of the new type according to the final public characteristics.
According to the method for predicting the missing link in the heterogeneous social network, the heterogeneous social network is subjected to primary feature extraction; then, carrying out preliminary aggregation and optimization on the extracted features through a convolutional neural network; and then performing two processes of countertraining respectively to obtain a final common characteristic for link prediction. Compared with other link prediction methods, the missing link prediction method in the heterogeneous social network is based on feature extraction, and predicts the new-appearing type link without prior type by paying attention to the difference of link types in the heterogeneous social network. The method can accurately and comprehensively generate a good prediction effect on the link prediction of the new type, and the antagonistic neural network is adopted for learning of shared characteristics, so that the method has a wide action range on the link prediction of the new type and has stronger applicability.
In one embodiment, the missing link prediction method in the heterogeneous social network comprises the following steps: converting links of the heterogeneous social network into structural feature vectors; inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization; performing preliminary link prediction on the link samples of the structural feature vectors by using a generated predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using a discriminant classifier; adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier; and acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discriminant classifier, and predicting the missing link of the new type according to the final public characteristics.
Step one, extracting a general heterogeneous social network, wherein the heterogeneous social network is generally a graph, each node is arranged in the graph, and links are generated among the nodes. The graph input node2vec feature extractor extracts preliminary structural features in the heterogeneous social network, namely, links in the graph are represented as general vectors, and edges in the graph are represented as 128-dimensional vectors.
To explore the feature representation of an edge, we must consider the mapping from nodes to feature representations. Let f (u) be the mapping function of node u to the feature representation. For each node u in the heterogeneous social network G, we obtain its sampled set of neighbor nodes n (u). We try to optimize the objective function that maximizes the logarithmic probability of observing the network in the vicinity of the set of neighbor nodes n (u) for a node u represented by its characteristics.
maxf∑u∈Vlog Pr(N(u)|f(u)) (1)
And (3) optimizing the formula (1) to obtain a characteristic representation function of the node u, which is expressed as f (u). For link sample e, we first get the feature representations f (u) and f (v) of node u and node v. Then, the link sample e is characterized as:
r(e)＝f(u)*f(v) (2)
and secondly, performing feature aggregation and optimization through a convolutional neural network based on the preliminary structural features, transmitting the optimized features into a first full-connection layer to perform preliminary link prediction, judging the truth of the link, and judging the quality of the link prediction through a loss function of an output unit of a predictor. Wherein, the node2vec feature extractor, the convolutional neural network, the first fully-connected layer and the predictor output unit are together called as a generation predictor.
Using Gp(S；θr，θp) Representing a generative predictor, where S represents the input link sample set, θpRepresenting the relevant parameters. RFThe set of feature representations for all links in S is defined as follows:
RF＝{r(e)|e∈S} (3)
to predict whether a link sample is a missing link, three fully-connected layers with softmax functions, i.e., the first fully-connected layer, are used. For a given link sample e, the output of the generator predictor can be expressed as:
P(e)＝Gp({e}；θr，θp) (4)
where p (e) represents the likelihood of the presence of the link sample e. The goal of the generative predictor is to predict whether a particular link sample is a missing link. For a given set of link samples S, the prediction penalty for generating the predictor is defined by the cross entropy function as:
Lp(θr，θp)＝-∑e∈S[melog(P(e))+(1-me)log(1-P(e))] (5)
here, meM in e {0, 1}e1 denotes that link sample e is a positive sample; otherwise, me0 means that the link sample e is a negative sample. To better predict missing links, the primary task is to minimize prediction loss. Calculating optimal parameters
And
can be expressed as:
and thirdly, based on the preliminary structural features, performing feature aggregation and optimization through a convolutional neural network, transmitting the optimized features into a gradient inversion layer, then transmitting the optimized features into a second full-connection layer, and judging the quality degree of the type through a loss function of a classifier output unit. Wherein, the gradient inversion layer, the second fully-connected layer and the classifier output unit are together called a discriminant classifier.
The discriminant classifier is a neural network consisting of two fully connected layers and corresponding activation functions. Based on RFThe captured features in (c) represent that the links in the link sample set S are correctly classified as corresponding history types. Using Gc(RF；θc) To represent a discriminative classifier, wherein RFFrom the generative predictor, θcRepresenting the parameters to be learned.
Gc(RF；θc) By measuring RFThe difference between the link types is indirectly estimated by the difference of the representations of the corresponding links. For the link sample set S, we define the classification penalty of the discriminating classifier with the cross-entropy function as follows:
wherein, TpIncluding the different possible types of links that may be,
indicating that the predicted link sample e is at TpIn is (a)1,a2) The probability of the link type of (c). If the link type of e is (a)1,a2) Then n ise1 is ═ 1; otherwise, ne0. When L isc(θr,θc) The smaller the value, the better the performance of the discriminative classifier in classifying the link samples in S into the correct type. To distinguish between different link sample types, the loss L is minimizedc(θr,θc) Post discriminant classifier parametric representationThe following:
the above loss Lc(θr,θc) Can be used for calculating RFThe representation of different link types in (c). The larger the loss, the similar distribution of different types of characteristics is shown, and the learning characteristics are type-invariant.
Step four, the loss function in step two is required to be minimum finally to obtain the best link prediction effect, and the loss function of the discrimination classifier in step three is required to be maximum to make the features more common. During the solving process of the maximum and minimum loss functions, a very large and very small countermeasure effect is generated, and finally balance is achieved, so that the method can obtain the best effect.
To remove features unique to each link type during the training phase, it is desirable to maximize
To find the optimum parameter thetar. Thus, a very small maximum two-player game between the generative predictor and the discriminative classifier is constructed. Generating a predictor Gp(S；θr,θp) Attempting to fool the discriminant classifier G by degrading classification performancec(RF；θc) To determine the classifier Gc(RF；θc) It is attempted not to be spoofed by discovering certain type features that identify the type of link.
Based on the two-man strategy of maximin, the integrated loss is defined as follows:
Lfinal(θr,θp,θc)＝Lp(θr,θp)-Lc(θr,θc) (9)
finding the optimal parameter θrTo help RFThe feature in (1) indicates compliance with a criterion for sharing the feature. The optimization process of the saddle points of the relevant parameters is expressed as:
in one embodiment, the missing link prediction method in the heterogeneous social network comprises the following steps: converting links of the heterogeneous social network into structural feature vectors; inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization; performing preliminary link prediction on the link samples of the structural feature vectors by using the generated predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using the judgment classifier; adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier; and acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discriminant classifier, and predicting the missing link of the new type according to the final public characteristics.
The heterogeneous social network comprises a plurality of nodes, links are formed among the nodes, a link sample e is a link between a node u and a node v, and a characteristic representation function of the link sample e is as follows:
r(e)＝f(u)*f(v)
wherein f (u) is a feature expression function of the node u, and f (v) is a feature expression function of the node v.
Referring to fig. 2, fig. 2 is a flowchart of a method for predicting a missing link in a heterogeneous social network according to another embodiment of the present invention, where the method for predicting a missing link in a heterogeneous social network includes, but is not limited to, steps S210 to S230.
Step S210, inputting a link sample set to a generation predictor;
step S220, calculating the possibility that the link samples in the link sample set are real links by using a generated predictor;
and step S230, calculating the prediction loss of the generated predictor according to the link sample set, the possibility that the link sample is a real link and the attribute of the link sample.
In one embodiment, the missing link prediction method in the heterogeneous social network comprises the following steps: converting links of the heterogeneous social network into structural feature vectors; inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization; performing preliminary link prediction on the link samples of the structural feature vectors by using a generated predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using a discriminant classifier; adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier; and acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discriminant classifier, and predicting the missing link of the new type according to the final public characteristics.
The method for performing preliminary link prediction on the link samples of the structural feature vector by using the generated predictor comprises the following steps: inputting a link sample set to a generation predictor; calculating the possibility that the link samples in the link sample set are real links by using a generation predictor; and calculating the prediction loss of the generated predictor according to the link sample set, the possibility that the link sample is a real link and the attribute of the link sample.
In one embodiment, the missing link prediction method in the heterogeneous social network comprises the following steps: converting links of the heterogeneous social network into structural feature vectors; inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization; performing preliminary link prediction on the link samples of the structural feature vectors by using a generated predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using a discriminant classifier; adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier; and acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discriminant classifier, and predicting the missing link of the new type according to the final public characteristics.
The method for performing preliminary link prediction on the link samples of the structural feature vector by using the generated predictor comprises the following steps: inputting a link sample set to a generation predictor; calculating the possibility that the link samples in the link sample set are real links by using a generation predictor; and calculating the prediction loss of the generated predictor according to the link sample set, the possibility that the link samples are real links and the attributes of the link samples.
The prediction loss of the generative predictor is expressed as:
wherein S is a link sample set, P (e) is a probability that a link sample is a real link, meIs an attribute of the link sample, me∈{0,1}，me1 indicates that the link sample e is a positive sample, me0 means that the link sample e is a negative sample.
Referring to fig. 3, fig. 3 is a flowchart of a method for predicting a missing link in a heterogeneous social network according to another embodiment of the present invention, where the method for predicting a missing link in a heterogeneous social network includes, but is not limited to, steps S310 to S320.
Step S310, calculating to obtain a feature representation set of all links in the link sample set according to the link sample set;
step S320, based on the feature representation sets of all links, the discriminant classifier classifies the links in the link sample set into corresponding history types.
In one embodiment, the missing link prediction method in the heterogeneous social network comprises the following steps: converting links of the heterogeneous social network into structural feature vectors; inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization; performing preliminary link prediction on the link samples of the structural feature vectors by using a generated predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using a discriminant classifier; adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier; and acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discriminant classifier, and predicting the missing link of the new type according to the final public characteristics.
The method for performing preliminary link prediction on the link samples of the structural feature vector by using the generated predictor comprises the following steps: inputting a link sample set to a generation predictor; calculating the possibility that the link samples in the link sample set are real links by using a generation predictor; and calculating the prediction loss of the generated predictor according to the link sample set, the possibility that the link sample is a real link and the attribute of the link sample. The method for classifying the link samples of the structure feature vector by using the discriminant classifier comprises the following steps: calculating to obtain a feature representation set of all links in the link sample set according to the link sample set; based on the feature representation sets of all links, the discriminative classifier classifies the links in the link sample set as corresponding history types.
The classification penalty of the discriminant classifier is expressed as:
wherein, TpA plurality of link types are included and,
indicating that the predicted link sample e is at TpIn is (a)1,a2) E has a link type of (a)1,a2) Then n ise1 is ═ 1; e link type is not (a)1,a2),ne＝0。
The invention also provides a system for predicting missing links in the heterogeneous social network, which comprises the following steps: the characteristic extractor is used for converting the links of the heterogeneous social network into structural characteristic vectors;
the convolutional neural network is used for carrying out feature aggregation and feature optimization on the structural feature vector;
generating a predictor for performing preliminary link prediction on the link samples of the structural feature vectors;
the judging classifier is used for classifying the link samples of the structural feature vectors;
the parameter adjusting module is used for adjusting the parameters of the loss function of the generated predictor so as to reduce the prediction loss of the generated predictor and adjusting the parameters of the loss function of the discriminant classifier so as to reduce the classification loss of the discriminant classifier;
and the missing link prediction module is used for acquiring the final public characteristic in the parameter adjustment process of the loss function of the generation predictor and the discriminant classifier and predicting the missing link of the new type according to the final public characteristic.
The invention also provides a system for predicting missing links in the heterogeneous social network, which comprises the following steps: a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor when executing the computer program implementing a method of missing link prediction in a heterogeneous social network as described above.
The processor and memory may be connected by a bus or other means.
The memory, which is a non-transitory computer readable storage medium, may be used to store non-transitory software programs as well as non-transitory computer executable programs. Further, the memory may include high speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory located remotely from the processor, and these remote memories may be connected to the processor through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
It should be noted that, the missing link prediction system in the heterogeneous social network in this embodiment may include a service processing module, an edge database, a server version information register, and a data synchronization module, and when the processor executes the computer program, the missing link prediction method in the heterogeneous social network that is applied to the missing link prediction system in the heterogeneous social network is implemented as described above.
The above-described embodiments of the apparatus are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may also be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.
Furthermore, an embodiment of the present invention further provides a computer-readable storage medium, where computer-executable instructions are stored, and executed by a processor or a controller, for example, by a processor in the terminal embodiment, so that the processor may execute the missing link prediction method in the heterogeneous social network in the foregoing embodiment.
One of ordinary skill in the art will appreciate that all or some of the steps, systems, and methods disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those of ordinary skill in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, Digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can accessed by a computer. In addition, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media as known to those skilled in the art.
While the preferred embodiments of the present invention have been described in detail, it will be understood by those skilled in the art that the foregoing and various other changes, omissions and deviations in the form and detail thereof may be made without departing from the scope of this invention.
The above-described embodiments of the present invention should not be construed as limiting the scope of the present invention. Any other corresponding changes and modifications made according to the technical idea of the present invention should be included in the protection scope of the claims of the present invention.
Claims (9)
1. A method for predicting a missing link in a heterogeneous social network is characterized by comprising the following steps:
converting the links of the heterogeneous social network into structural feature vectors;
inputting the structural feature vector into a convolutional neural network for feature aggregation and feature optimization;
performing preliminary link prediction on the link samples of the structural feature vectors by using a generation predictor to judge the truth of the link samples, and classifying the link samples of the structural feature vectors by using a discrimination classifier;
adjusting parameters of a loss function of the generated predictor to reduce the prediction loss of the generated predictor, and adjusting parameters of a loss function of the discriminant classifier to reduce the classification loss of the discriminant classifier;
and acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discrimination classifier, and predicting the missing link of the new type according to the final public characteristics.
2. The method for predicting the missing link in the heterogeneous social network according to claim 1, wherein the heterogeneous social network comprises a plurality of nodes, links are formed among the nodes, a link sample e is a link between a node u and a node v, and a characteristic representation function of the link sample e is as follows:
r(e)＝f(u)*f(v)
wherein, f (u) is a feature expression function of the node u, and f (v) is a feature expression function of the node v.
3. The method for predicting the missing link in the heterogeneous social network as claimed in claim 1, wherein the preliminary link prediction of the link sample of the structural feature vector by using the generated predictor comprises the steps of:
inputting a set of link samples into the generative predictor;
calculating the possibility that the link samples in the link sample set are real links by using the generation predictor;
and calculating the prediction loss of the generation predictor according to the link sample set, the possibility that the link sample is a real link and the attribute of the link sample.
4. The method for predicting the missing link in the heterogeneous social network according to claim 3, wherein the prediction loss of the generated predictor is represented as:
wherein S is the link sample set, P (e) is the probability that the link sample is a real link, meFor the attributes of the link samples, me∈{0，1}，me1 indicates that the link sample e is a positive sample, me0 means that the link sample e is a negative sample.
5. The method for predicting the missing link in the heterogeneous social network according to claim 3, wherein the step of classifying the link samples of the structural feature vector by using a discriminant classifier comprises the steps of:
calculating to obtain a feature representation set of all links in the link sample set according to the link sample set;
based on the feature representation sets of all the links, the discriminative classifier classifies the links in the link sample set into corresponding history types.
6. The method of claim 5, wherein the classification loss of the discriminant classifier is expressed as:
wherein, TpA plurality of link types are included and,
indicating that the predicted link sample e is at TpIn is (a)1，a2) E has a link type of (a)1，a2) Then n ise1 is ═ 1; e link type is not (a)1，a2)，ne＝0。
7. A system for missing link prediction in a heterogeneous social network, comprising:
the characteristic extractor is used for converting the links of the heterogeneous social network into structural characteristic vectors;
the convolutional neural network is used for carrying out feature aggregation and feature optimization on the structural feature vector;
generating a predictor for performing preliminary link prediction on the link samples of the structural feature vector;
the discrimination classifier is used for classifying the link samples of the structural feature vectors;
the parameter adjusting module is used for adjusting parameters of a loss function of the generated predictor so as to reduce the prediction loss of the generated predictor and adjusting parameters of a loss function of the discriminant classifier so as to reduce the classification loss of the discriminant classifier;
and the missing link prediction module is used for acquiring final public characteristics in the parameter adjustment process of the loss functions of the generation predictor and the discrimination classifier and predicting the missing link of a new type according to the final public characteristics.
8. The system of claim 1, wherein the system comprises: memory, a processor and a computer program stored on the memory and executable on the processor, the processor when executing the computer program implementing the method of missing link prediction in a heterogeneous social network according to any of claims 1 to 6.
9. A computer-readable storage medium storing computer-executable instructions for causing a computer to perform the method of missing link prediction in a heterogeneous social network of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210175814.3A CN114548569A (en) | 2022-02-24 | 2022-02-24 | Missing link prediction method, system and storage medium in heterogeneous social network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210175814.3A CN114548569A (en) | 2022-02-24 | 2022-02-24 | Missing link prediction method, system and storage medium in heterogeneous social network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114548569A true CN114548569A (en) | 2022-05-27 |
Family
ID=81678643
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210175814.3A Pending CN114548569A (en) | 2022-02-24 | 2022-02-24 | Missing link prediction method, system and storage medium in heterogeneous social network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114548569A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117151279A (en) * | 2023-08-15 | 2023-12-01 | 哈尔滨工业大学 | Isomorphic network link prediction method and system based on line graph neural network |
-
2022
- 2022-02-24 CN CN202210175814.3A patent/CN114548569A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117151279A (en) * | 2023-08-15 | 2023-12-01 | 哈尔滨工业大学 | Isomorphic network link prediction method and system based on line graph neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111709028B (en) | Network security state evaluation and attack prediction method | |
| CN113378990B (en) | Flow data anomaly detection method based on deep learning | |
| Idrissi et al. | Accelerating the update of a DL-based IDS for IoT using deep transfer learning | |
| CN113364751A (en) | Network attack prediction method, computer-readable storage medium, and electronic device | |
| CN116522263A (en) | Abnormal cell detection method and device | |
| CN115987552A (en) | Network intrusion detection method based on deep learning | |
| CN115801374A (en) | Network intrusion data classification method and device, electronic equipment and storage medium | |
| KR102488789B1 (en) | Prediction and classification method, apparatus and program using one class anomaly detection model based on artificial intelligence | |
| CN115063664A (en) | Model learning method, training method and system for industrial vision detection | |
| CN115600194A (en) | Intrusion detection method, storage medium and device based on XGboost and LGBM | |
| CN114548569A (en) | Missing link prediction method, system and storage medium in heterogeneous social network | |
| CN113343123B (en) | Training method and detection method for generating confrontation multiple relation graph network | |
| Liu et al. | Learning multiple gaussian prototypes for open-set recognition | |
| CN111291078B (en) | Domain name matching detection method and device | |
| Puzanov et al. | Deep reinforcement one-shot learning for change point detection | |
| CN115410250A (en) | Array type human face beauty prediction method, equipment and storage medium | |
| CN117523218A (en) | Label generation, training of image classification model and image classification method and device | |
| CN115620083A (en) | Model training method, face image quality evaluation method, device and medium | |
| CN115564709A (en) | Evaluation method and system for robustness of power algorithm model in confrontation scene | |
| CN115170838A (en) | Data screening method and device | |
| Liu et al. | Scene-Specialized Multitarget Detector with an SMC-PHD Filter and a YOLO Network | |
| Gyamfi et al. | Automated system-level anomaly detection and classification using modified random forest | |
| Saffarzadeh et al. | Identifying Spam Tweets in Social Networks with Combined Approaches of Feature Selection and Deep Learning | |
| CN117312976A (en) | Internet of things equipment fingerprint identification system and method based on small sample learning | |
| Asyaev et al. | Anomaly Detection Model in APCS Using AutoML |
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 |