CN110493045A - A kind of directed networks link prediction method merging multimode body information - Google Patents
A kind of directed networks link prediction method merging multimode body information Download PDFInfo
- Publication number
- CN110493045A CN110493045A CN201910764256.2A CN201910764256A CN110493045A CN 110493045 A CN110493045 A CN 110493045A CN 201910764256 A CN201910764256 A CN 201910764256A CN 110493045 A CN110493045 A CN 110493045A
- Authority
- CN
- China
- Prior art keywords
- node
- list
- neighbours
- link prediction
- network
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/147—Network analysis or design for predicting network behaviour
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of directed networks link prediction methods for merging multimode body information, which comprises the steps of: S1. constructs initial network, and obtains the node on the company of being not present side to list；S2. in initial network data 10% company side is randomly selected as test set positive sample, and the company side of residue 90% is used as training set, chooses and big company's line sets such as test set positive sample are as test set negative sample；S3. the corresponding role's functional value of individual in initial network is obtained；S4. each node is obtained to role's function R of corresponding common neighbourswList；S5. common neighbours' quantity of node pair is obtained；S6. the r' of node pair is obtainedxyList；S7. according to single die body r'xyList obtains the r of bimodulus body by way of superpositionxyList；Or the r' for obtaining different single die bodysxyList obtains new score list using the method XGBoost of machine learning.The application sufficiently applies the structure feature of directed networks, is greatly improved link prediction accuracy.
Description
Technical field
The present invention relates to a kind of link prediction method, specifically a kind of directed networks link for merging multimode body information is pre-
Survey method.
Background technique
Link prediction is the important research direction in one, complex network field, and basic problem to be dealt with is exactly to pass through
The information such as the network node known and network structure predict there is a possibility that link in network between any two node.It is logical
Cross side future that link prediction is not only not present in available network there may be a possibility that, can also find out in network
Whether already present be even when falseness the connects or company of missing side.
In the link prediction method based on network structure, the most frequently used is the method for common neighbours' similitude.Liben-
The method of common neighbours of the Nowellhe and Kleinberg discovery based on node is one of forecasting accuracy the best way.But
It is that common neighbours' index between node does not consider the link direction between node, does not can be used directly in directed networks.Prediction
Side and common neighbours constitute the triangular structure of closure, consider the problems of direction on the basis of triangular structure, just constitute
The partial structurtes of directed networks, currently, research a part to directed networks structure is carried out based on this kind of partial structurtes.
In the link prediction of directed networks based on partial structurtes progress, most common is the prediction based on local message similitude
Method.
In true complex network, the relationship between individual is not often reciprocity, is made of this not peer-to-peer
Network is exactly directed networks.Link prediction is carried out using die body in the link prediction of directed networks.Die body is referred to true
Correspondence subgraph of the number occurred in real network much higher than the number occurred in random network.Zhang Qianming et al. proposes direct net
Network gesture is theoretical, and finds that the die body structure for meeting gesture theory has better link prediction effect.But in they study only
The link prediction situation for considering single die body, does not account for the link prediction of multimode body.
Link prediction method based on die body structure simply thinks: the contribution margin that each node constitutes die body is identical.
But in true network system, this idea is often incorrect.Liu Zhen et al. is in Undirected networks based on common adjacent
It occupies method and has counted influence of the node contribution margin to the accuracy of network using model-naive Bayesian, discovery is based on simple pattra leaves
Other than PB network, the predictive ability of other networks is all improved this forecasting accuracy；Wu et al. is to undirected
WLNBCN model is proposed in the research of weighted network, and the weight of the common neighbours of the node pair of prediction is examined as role's function
Consider in link prediction, improves the accuracy of link prediction to varying degrees.But in the chain of existing directed networks
Influence of the node contribution margin except the prediction side of die body to link prediction is not accounted in the prediction of road.
Summary of the invention
The application proposes a kind of directed networks link prediction method for merging multimode body information, by the way that role's function is added,
So that forecasting accuracy is got a promotion, the joint effect of multiple die bodys is subjected to link prediction using XGBoost later, is further mentioned
The accuracy of link prediction is risen.
To achieve the above object, the technical solution of the application are as follows: a kind of directed networks link for merging multimode body information is pre-
Survey method, includes the following steps:
S1. initial network is constructed by original directed networks data, and obtains the node on the company of being not present side to list；
S2. in initial network data 10% company side is randomly selected as test set positive sample, the Lian Bianzuo of residue 90%
For training set, from there is no the nodes on even side to being chosen in list with big company's line sets such as test set positive samples as test set
Negative sample；
S3. the corresponding role's functional value of individual in initial network is obtained using model-naive Bayesian algorithm；
S4. since the first node in initial network to, according to the common neighbours of node pair, each node is obtained to institute
Role's function R of corresponding common neighbourswList；
S5. since the first node in initial network to, common neighbours' similarity indices of each pair of node are successively calculated
CN obtains common neighbours' quantity of node pair；
S6. node is obtained to role's function of corresponding common neighbours according to common neighbours' quantity of node pair and node
Pair r'xyList；
S7. according to single die body r'xyList obtains the r of bimodulus body by way of superpositionxyList；Or it will be different
The obtained r' of single die bodyxyList obtains new score list using the method XGBoost of machine learning.
Further, the correlation between all die bodys is calculated using the method for XGBoost, according to the correlation between die body
Property carry out the die body selection of bimodulus body link prediction, obtained result stability is preferable.
Further, the calculation method of common neighbours' similarity indices CN are as follows: undirected and unweighted network is directed to, with Γ (a) table
Show that the neighbours of node a, Γ (b) indicate the neighbours of node b, then:
CN (a, b)=| Γ (a) ∩ Γ (b) |
The similarity indices of node a and node b are equal to the neighbor node quantity Chong Die with the neighbor node of node b of node a,
That is common neighbours' quantity of node a and node b；
And then obtain, have no right network for oriented, Γ (x) indicates that the neighbours on node x assigned direction, Γ (y) indicate
Neighbours on node y assigned direction；
Oxy=| Γ (x) ∩ Γ (y) |
Oxy indicates neighbours' quantity Chong Die with the neighbor node on node y assigned direction on node x assigned direction.
Further, role's function RwThe calculation method of value are as follows:
VwIndicate node to all common neighbor nodes of (x, y), N in formulaΔwIt is node vwNeighbours in be connected with each other
Node pair number, NΛwIt is node vwNeighbours in mutually discrete node pair number.
Further, the r' of single die body node pairxyThe calculation method of list value are as follows:
In formulaWherein MF=V (V-1)/2 indicates the number on all even sides that may be present in network,
M=| ET| indicate the number on the company side of necessary being in network, V is node total number, and E is the set on all even sides in network.
Further, the r of bimodulus body node pairxyThe calculation method of list value are as follows:
Wherein, O1xyIndicate node x1Neighbours and node y on assigned direction1Neighbor node on assigned direction is overlapped number
Amount, O2xy| indicate node x2Neighbours and node y on assigned direction2Neighbor node on assigned direction is overlapped quantity, RvIt indicates
Role's functional value of one of die body, RwIndicate role's functional value of another die body.
The present invention due to using the technology described above, can obtain following technical effect:
1. the link prediction of bimodulus body is on the whole to a certain extent compared to the link prediction accuracy of single mode body
It improves.
2. carrying out link prediction for multiple die bodys, considers the joint effect of multiple die bodys, it is pre- further to improve link
The accuracy of survey.
3. the application sufficiently applies the structure feature of directed networks, it is greatly improved link prediction accuracy.
Detailed description of the invention
Fig. 1 is the link prediction based on die body quantity compared with the AUC of the single mode body link prediction based on naive Bayesian
Schematic diagram；
Fig. 2 is the structure chart of role's function calculating of three rank die bodys and quadravalence die body；
Link prediction figure of the Fig. 3 based on bimodulus body；
Fig. 4 shows figure based on the correlation between the die body of XGBoost.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, carries out to the technical solution in present invention implementation clear, complete
Description, it is to be understood that described example is only a part of example of the invention, instead of all the embodiments.
Based on the embodiment of the present invention, those skilled in the art without making creative work it is obtained it is all its
His embodiment, belongs to protection scope of the present invention.
The present embodiment provides a kind of directed networks link prediction method for merging multimode body information, network G (V, E) tables
Show, wherein V indicates nodes set, and E indicates to connect line set in network.Usually E is divided for two parts: training set ETAnd survey
Examination collection EP, haveAnd ET∪EP=E.10% company side has been randomly selected as test set positive sample EP, residue 90%
Company side as training set ET, and choose with big company's line sets such as test set positive samples from there is no even side as test set
Negative sampleSpecific step is as follows:
S1. original directed networks data are obtained and construct initial network, and obtain the node on the company of being not present side to list；
S2. in network data 10% company side is randomly selected as test set positive sample, and the company side of residue 90% is as instruction
Practice collection, from there is no the nodes on even side to being chosen in list with big company's line set such as test set positive sample as the negative sample of test set
This；
S3. the corresponding role's functional value of individual in network is obtained using model-naive Bayesian algorithm；
S4. since the first node in network to, according to the common neighbours of node pair, each node is obtained to corresponding
Common neighbours role's function RwList；
S5. since the first node in network to, according to common neighbours' quantity of node pair, each pair of node is successively calculated
Common neighbours' similarity indices CN,
S6. node is obtained to role's function of corresponding common neighbours according to common neighbours' quantity of node pair and node
Pair r'xyList；
Then it is divided into two kinds of approach and obtains list used in link prediction:
S71. it is directed to bimodulus body, according to single die body r'xyList obtains the r of bimodulus body by way of superpositionxyColumn
Table；The correlation between all die bodys is calculated using the method for XGBoost, bimodulus body chain is carried out according to the correlation between die body
The die body selection of road prediction, obtained result stability are preferable.
S72. multimode body, the r' that different fallout predictors is obtained are directed toxyList is obtained using the method XGBoost of machine learning
To new score list；
The similarity indices of calculate node x and node y, measure its there are a possibility that, score is higher to mean connection
Possibility is bigger.By all there is no connecting side to arrange according to score descending, then the company side for coming front most possibly exists.
The calculation R of role's functionw: the similarity algorithm based on common neighbours carries out link prediction and assumes that all sections
Point is consistent the contribution weight of prediction node pair, does not distinguish role and its contribution of the common neighbours of different attribute.But
In directed networks, due to the presence of the not peer-to-peer in network between individual, so that node has the structure of overall network
Very big influence, RwEach node can be measured or influence that each edge (for four node die bodys) constitutes die body.Then
The role's function R calculated based on model-naive BayesianwIs defined as:
VwIndicate node to all common neighbor nodes of (x, y), N in formulaΔwIt is that a node in network can be with
EPIn number of the node to the node pair for constituting specified die body, NΛwIt is that a node in network can be withIn node pair
Constitute the number of the node pair of specified die body.
r'xyThe calculation method of value are as follows: role's functional value from available each node or side, later according to each edge energy
It is enough with node or while constitute the set of node for the die body specified or while set it is available for each side, it
Total role's functional value.Each edge corresponding angle is added on the basis of common neighbours' quantity of the corresponding assigned direction of each edge
Color function sum, so that it may as the corresponding r' of each edgexyValue
In formulaWherein MF=V (V-1)/2 indicates the number on all even sides that may be present in network,
M=| ET| indicate the number on the company side of necessary being in network.
The r of bimodulus bodyxyThe calculation method of value are as follows:
Formula is divided into two parts, is that two single mode bodies obtain r' respectivelyxyCalculating, two single mode bodies are synthesized one later
The whole result being calculated for carrying out bimodulus body is equal to the result for being directly added the result of two die bodys.
The calculation of the prediction score of multimode body is carried out by way of machine learning:
The r' for being obtained the test set of obtained single die body using XGBoostxyIt list and is obtained by training set
R'xyList is brought into the frame of machine learning, passes through the r' to obtained training training setxyThe study of list can obtain
One new test set score list.
Meanwhile by the correlation between the available different die body of XGBoost model, according to the correlation between die body
Property can choose different multimode body combinations, so that it may obtain the score list for various combination.
In order to probe into influence of the die body node of directed networks to link prediction, model-naive Bayesian calculate node is used
Role's functional value, role's function of node is added in traditional link prediction algorithm.Traditional node role's function
It is to be proposed based on three node die bodys, does not account for the calculated case of four node die bodys.It is three nodes as shown in a figure in Fig. 2
Die body, it is envisaged that the influence that the node C other than predicting side AB generates die body.It is expanded,
Calculate role's function of four node die bodys.As shown in the b figure in Fig. 2, due to being gone back other than predicting side AB in four node die bodys
There are two node C and D, so there are three types of may situation when role's function to four node die bodys accounts for.First
Kind is the role's function for only considering node C, and second is the role's function for only considering node D, the third is by node C and node
The entirety that company side between D and node C D is constituted is as role's function.Because first two mode all only considers prediction
Part-structure except side, it is more unilateral for integrally-built prediction.So in the mistake for carrying out the calculating of four node role's functions
Cheng Zhong calculates the structure in role's function, that is, consideration figure b in the frame of four node die bodys for die body using the third mode
The influence of generation.
The full name in English of AUC is area under the receiver operating characteristic
Curve refers to the area under ROC curve (receiver operating characteristic curve).
AUC can measure the accuracy of link prediction on the whole.AUC refers to randomly choosing in test set positive sample
The fractional value of a line, the high probability than randomly choosing the fractional value on side of a test set negative sample.That is, every time from
EPWithMiddle random selection one, if EPThe fractional value on side be greater thanSide fractional value, then just plus 1 point, if EP's
Fractional value is equal toSide fractional value, just plus 0.5 point, otherwise just not bonus point.This process is independently subjected to n times, if there is X
Secondary EPThe score on side be greater thanSide score, have Y EPThe score on side be equal toSide score, have Z EPSide
Score is less thanSide score, then AUC can be with is defined as:
It as AUC=0.5, indicates what all scores were equivalent to be randomly generated, as AUC=1, indicates that algorithm is completely correct
The company of predicting side situation of change.AUC is bigger to illustrate that prediction result is more accurate, and the size of AUC reflects algorithm phase used
For the height of random algorithm accuracy.
The link prediction result of single mode body and multimode body joint effect of the table 1 based on naive Bayesian
Table one refers to that multiple die body features have been carried out the common shadow that common prediction obtains by the method for XGBoost
It rings, as can be seen from the table, the prediction result obtained based on all die bodys is than the prediction that any one individual fallout predictor obtains
As a result it to get well；
Bimodulus body link prediction result of the table 2 based on naive Bayesian
Table two refers to the link prediction of bimodulus body, using the bimodulus body link prediction method based on naive Bayesian, from
It can be seen that the accuracy of link prediction can be improved in obtained bimodulus body link prediction result to a certain extent in table.But
Such die body selection randomness is relatively high, the calculating of correlation between die body is carried out using the method for XGBoost, according to die body
Between correlation select inhomogeneous die body to carry out the link prediction of bimodulus body, for die body selection provide selection according to
According to, while also increasing the stability of bimodulus body link prediction.
The above is only a preferred embodiment of the present invention, it is not intended to restrict the invention, it is noted that for this skill
The those of ordinary skill in art field can also make several improvements and modifications without departing from the technical principles of the invention,
These improvements and modifications also should be regarded as protection scope of the present invention.
Claims (6)
1. a kind of directed networks link prediction method for merging multimode body information, which comprises the steps of:
S1. initial network is constructed by original directed networks data, and obtains the node on the company of being not present side to list；
S2. in initial network data 10% company side is randomly selected as test set positive sample, and the company side of residue 90% is as instruction
Practice collection, from there is no the nodes on even side to being chosen in list with big company's line set such as test set positive sample as the negative sample of test set
This；
S3. the corresponding role's functional value of individual in initial network is obtained using model-naive Bayesian algorithm；
S4. since the first node in initial network to, according to the common neighbours of node pair, each node is obtained to corresponding
Common neighbours role's function RwList；
S5. since the first node in initial network to, common neighbours' similarity indices CN of each pair of node is successively calculated, is obtained
To common neighbours' quantity of node pair；
S6. node pair is obtained to role's function of corresponding common neighbours according to common neighbours' quantity of node pair and node
r'xyList；
S7. according to single die body r'xyList obtains the r of bimodulus body by way of superpositionxyList；Or it will be different single
The r' that die body obtainsxyList obtains new score list using the method XGBoost of machine learning.
2. a kind of directed networks link prediction method for merging multimode body information according to claim 1, which is characterized in that make
The correlation between all die bodys is calculated with the method for XGBoost, it is pre- to carry out bimodulus body link according to the correlation between die body
The die body of survey selects.
3. a kind of directed networks link prediction method for merging multimode body information according to claim 1, which is characterized in that altogether
With the calculation method of neighbours' similarity indices CN are as follows: be directed to undirected and unweighted network, the neighbours of node a, Γ (b) are indicated with Γ (a)
Indicate the neighbours of node b, then:
CN (a, b)=| Γ (a) ∩ Γ (b) |
The similarity indices of node a and node b are equal to the neighbor node quantity Chong Die with the neighbor node of node b of node a, that is, save
Common neighbours' quantity of point a and node b；
And then obtain, have no right network for oriented, Γ (x) indicates that the neighbours on node x assigned direction, Γ (y) indicate node y
Neighbours on assigned direction；
Oxy=| Γ (x) ∩ Γ (y) |
Oxy indicates neighbours' quantity Chong Die with the neighbor node on node y assigned direction on node x assigned direction.
4. a kind of directed networks link prediction method for merging multimode body information according to claim 1, which is characterized in that angle
Color function RwThe calculation method of value are as follows:
VwIndicate node to all common neighbor nodes of (x, y), N in formulaΔwIt is node vwNeighbours in section interconnected
The number of point pair, NΛwIt is node vwNeighbours in mutually discrete node pair number.
5. a kind of directed networks link prediction method for merging multimode body information according to claim 1, which is characterized in that single
The r' of a die body node pairxyThe calculation method of list value are as follows:
In formulaWherein MF=| V | (| V | -1)/2 indicate the numbers on all even sides that may be present in networks, M
=| ET| indicate the number on the company side of necessary being in network, V is node total number, and E is the set on all even sides in network.
6. a kind of directed networks link prediction method for merging multimode body information according to claim 5, which is characterized in that double
The r of die body node pairxyThe calculation method of list value are as follows:
Wherein, | O1xy| indicate node x1Neighbours and node y on assigned direction1Neighbor node on assigned direction is overlapped number
Amount, | O2xy| indicate node x2Neighbours and node y on assigned direction2Neighbor node on assigned direction is overlapped quantity, RvTable
Show role's functional value of one of die body, RwIndicate role's functional value of another die body.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910764256.2A CN110493045A (en) | 2019-08-19 | 2019-08-19 | A kind of directed networks link prediction method merging multimode body information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910764256.2A CN110493045A (en) | 2019-08-19 | 2019-08-19 | A kind of directed networks link prediction method merging multimode body information |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110493045A true CN110493045A (en) | 2019-11-22 |
Family
ID=68552125
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910764256.2A Pending CN110493045A (en) | 2019-08-19 | 2019-08-19 | A kind of directed networks link prediction method merging multimode body information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110493045A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111461440A (en) * | 2020-04-02 | 2020-07-28 | 河北工程大学 | Link prediction method, system and terminal equipment |
| CN111669288A (en) * | 2020-05-25 | 2020-09-15 | 中国人民解放军战略支援部队信息工程大学 | Directional network link prediction method and device based on directional heterogeneous neighbor |
| CN112819645A (en) * | 2021-03-23 | 2021-05-18 | 大连民族大学 | Social network false information propagation detection method based on motif degree |
| CN112862082A (en) * | 2021-03-18 | 2021-05-28 | 杭州师范大学 | Link prediction method based on support vector machine |
-
2019
- 2019-08-19 CN CN201910764256.2A patent/CN110493045A/en active Pending
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111461440A (en) * | 2020-04-02 | 2020-07-28 | 河北工程大学 | Link prediction method, system and terminal equipment |
| CN111461440B (en) * | 2020-04-02 | 2022-05-31 | 河北工程大学 | Link prediction method, system and terminal equipment |
| CN111669288A (en) * | 2020-05-25 | 2020-09-15 | 中国人民解放军战略支援部队信息工程大学 | Directional network link prediction method and device based on directional heterogeneous neighbor |
| CN111669288B (en) * | 2020-05-25 | 2023-02-14 | 中国人民解放军战略支援部队信息工程大学 | Directional network link prediction method and device based on directional heterogeneous neighbor |
| CN112862082A (en) * | 2021-03-18 | 2021-05-28 | 杭州师范大学 | Link prediction method based on support vector machine |
| CN112862082B (en) * | 2021-03-18 | 2023-09-29 | 杭州师范大学 | Link prediction method based on support vector machine |
| CN112819645A (en) * | 2021-03-23 | 2021-05-18 | 大连民族大学 | Social network false information propagation detection method based on motif degree |
| CN112819645B (en) * | 2021-03-23 | 2024-03-29 | 大连民族大学 | Social network false information propagation detection method based on degree of motif |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110493045A (en) | A kind of directed networks link prediction method merging multimode body information | |
| Xian et al. | Zero-shot learning-the good, the bad and the ugly | |
| Draelos et al. | Use HiResCAM instead of Grad-CAM for faithful explanations of convolutional neural networks | |
| Li et al. | Remaining useful life prediction of aero-engine enabled by fusing knowledge and deep learning models | |
| Hoffman et al. | A note on using the adjusted Rand index for link prediction in networks | |
| Wu et al. | Predicting top-L missing links with node and link clustering information in large-scale networks | |
| CN104166731A (en) | Discovering system for social network overlapped community and method thereof | |
| CN106548206A (en) | Multi-modal nuclear magnetic resonance image data classification method based on minimum spanning tree | |
| WO2022205502A1 (en) | Image classification model construction method, image classification method, and storage medium | |
| CN108304380A (en) | A method of scholar's name disambiguation of fusion academic | |
| CN108346474A (en) | The electronic health record feature selection approach of distribution within class and distribution between class based on word | |
| CN110084423A (en) | A kind of link prediction method based on local similarity | |
| Šubelj et al. | Convexity in scientific collaboration networks | |
| Cao et al. | Specification of a performance indicator using the evidential-reasoning approach | |
| CN105843744B (en) | Transformation relation preference grade sort method for concurrent program metamorphic testing | |
| Soós et al. | Beyond the basemap of science: mapping multiple structures in research portfolios: evidence from Hungary | |
| Bai et al. | Measure the impact of institution and paper via institution-citation network | |
| Xu et al. | Reconstruction of international energy trade networks with given marginal data: A comparative analysis | |
| Roschewitz et al. | IFedAvg: Interpretable data-interoperability for federated learning | |
| Junttila | Patterns in permuted binary matrices | |
| CN107688647A (en) | A kind of study based on collaborative filtering reviews exam pool and recommends method | |
| Lee et al. | Benchmarking community detection methods on social media data | |
| Singh et al. | Ftl‐CoV19: A Transfer Learning Approach to Detect COVID‐19 | |
| Niemeijer et al. | Constructing and predicting school advice for academic achievement: A comparison of item response theory and machine learning techniques | |
| Khare et al. | COVID Pneumonia Prediction Based on Chest X-Ray Images Using Deep Learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191122 |
|
| RJ01 | Rejection of invention patent application after publication |