CN106951471A - A kind of construction method of the label prediction of the development trend model based on SVM - Google Patents
A kind of construction method of the label prediction of the development trend model based on SVM Download PDFInfo
- Publication number
- CN106951471A CN106951471A CN201710127478.4A CN201710127478A CN106951471A CN 106951471 A CN106951471 A CN 106951471A CN 201710127478 A CN201710127478 A CN 201710127478A CN 106951471 A CN106951471 A CN 106951471A
- Authority
- CN
- China
- Prior art keywords
- label
- days
- data
- svm
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/955—Retrieval from the web using information identifiers, e.g. uniform resource locators [URL]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2216/00—Indexing scheme relating to additional aspects of information retrieval not explicitly covered by G06F16/00 and subgroups
- G06F2216/03—Data mining
Abstract
A kind of construction method of the label prediction of the development trend model based on SVM, comprises the following steps:(1) data set is pre-processed, and counts the model data of website, removes non-relevant data information;(2) sample label is chosen, the frequency after label newly occurs 2 years is counted, popular tag set and non-streaming row label set is extracted;(3) Undirected networks of having the right of label are built;(4) label characteristics data are extracted, include the network characterization and association attributes feature of label, training test data is used as;(5) data are trained using the method for support vector machines, and build label fashion trend forecast model.The present invention considers correlation between label, and classification is predicted come the future developing trend to label by attributive character combination network characterization, for predicting that potential popular label has higher precision.Not only improve guiding user and select rational label, being also beneficial to Web Hosting, person provides higher-quality label.
Description
Technical field
The present invention relates to data mining, data analysis technique, more particularly to a kind of label development trend based on SVM is pre-
Survey the construction method of model.
Background technology
With the fast development of network, the exchange of row information is entered in increasing people's selection by network, but greatly
The information of amount is poured in simultaneously so that user is difficult quickly to carry out high frequency zone to information, therefore, occurs in that web tab.Net
The appearance of network label, greatly solves this problem.Label is by the closely related crucial phrase of some and content into it can
To help people easily to describe and categorised content, while also allowing for the retrieval of information with sharing.
At the same time, the development trend of label and classification prediction are also just increasingly paid close attention to by people, new label quilt
How is fashion trend after proposition, often represents the fashion trend in this field focus or direction, is that Web community is very big
The problem of concern.For website, trend prediction and the label recommendations of new label are effectively carried out, topic can be promoted or new
The development in emerging field.For a user, current area can be correctly found by content being searched according to the fashion trend of label
Development trend.
The Main Basiss for entering row label selection to information at present are word degree of correlation and the information hair of information and label
Play self attributes of person etc..But there are some shortcomings, it is mainly manifested in:(1) it have ignored the potential fashion trend of new label;
(2) it have ignored the correlation between label and label;(3) cold content causes unexpected winner label so that information can not be effective
Search;(4) a few features are only taken into account so that the selections of part labels tend to it is unilateral.
Therefore, in order to allow the user preferably to select label when releasing news, selection as far as possible has potential popularity
Label.The present invention proposes that a kind of construction method solution of label prediction of the development trend model based on SVM is following two and asked substantially
Topic:(1) extraction label forms the network characterization and association attributes feature at initial stage and the development trend of label is quantitatively portrayed;
(2) future developing trend of new label is predicted.
The content of the invention
In order to improve the prediction of development trend of the website to the management of Web Community's label and to new label, overcome right at present
The deficiency of label popularity prediction.The invention provides a kind of construction method of the label prediction of the development trend model based on SVM,
The network characterization between label is not only combined, while also extracting to be trained together by the attributive character that early stage occurs in label
Prediction.
The technical solution adopted for the present invention to solve the technical problems is as follows:
A kind of construction method of the label prediction of the development trend model based on SVM, comprises the following steps:
Step 1:Data prediction, collects information content label data corresponding with its of Web community, in its data
Hold according to time sequence, take community formed N days after data, preliminarily formed with the label network for ensuring community;
Step 2:Sample label is chosen, data set is counted, acquisition community's label frequency simultaneously sorts, is taken before ratio is
α % label is designated as U as popular label, its setpop;Choose what is contrasted with the popular label time in remaining label
Label is non-streaming row label;
Step 3:Build label network, to the multiple labels occurred in the same information content, then it is assumed that these labels it
Between there is relation, make its company of being formed side between any two.All information in Web community are traveled through, the label for the Undirected networks that obtain having the right
Network GTag, its interior joint is emerging label, and even side is the relation between label, and the weight of network goes out jointly for both
Existing number of times;
Step 4:Characteristic is extracted, to sample label set U={ Upop,Uunpop, extract its interior label and create it first
The network characterization of M days and attributive character, set up sample training data set afterwards;
Step 5:Using Machine learning classifiers model supports vector machine SVM, kernel function is chosen, training generation is based on SVM
Popular Tag Estimation model, and carry out ten folding cross validations, draw measuring accuracy.
Further, in the step 1, the data after choosing N days are used as the data of pretreatment, wherein N selection, it then follows rule
It is then:Ensure that in website preceding 10% label data has been generated in N days, i.e. the preliminary shape of label network in website
Into.
Further, in the step 2, the arrangement of line frequency descending, set note are entered in the selection of sample label data to label
For, selectionMiddle ratio for preceding α % label as popular label, its set is designated as Upop.All label ratios are taken to be
β % label is as non-streaming row label set afterwards, and its set is designated as Qunpop.To each popular label tpop∈Upop, search with
Label tpopThe nearest label of creation time, be designated as tunpop, while meeting tunpop∈Qunpop, as non-streaming row label, with stream
Row label data formation control, its set is designated as Uunpop;
Further, in the step 4, the extraction to label network feature, M takes 30, and network characterization mainly includes:
1) the relative degree centrality after new label is proposed in 30 days:Label tiAngle value DiCalculating using removing isolated section
The mode of point, computing formula is as follows:
Wherein, N represents the total number of labels in network;aijThe element of network adjacent matrix is represented, if label tiAnd tjHave
Lian Bian, then aij=1;Otherwise aij=0;
Label tiThe central feature calculation of degree, take the label t in networkiRelative degree centrality:
Wherein, DiRepresent label tiAngle value;
2) neighbours' average degree centrality after new label is proposed in 30 days, label tiNeighbours' average degree NCiCalculating such as
Under:
Wherein, NneighborRepresent label tiNeighbor node number,Represent label tiNeighbor node degree
It is worth sum;
3) after new label is proposed in 30 days close to centrality, label tiClose central metric calculation, then
Equally take label tiClose to centrad:
Wherein, dijRepresent label tiWith label tjDistance,Represent label tiTo neighbours' label node
Average geodesic distance;
4) eigenvector centrality after new label is proposed in 30 days, label tiEigenvector centrality be calculated as follows:
Wherein, η is a proportionality constant, A=(aijwij) it is the network adjacent matrix weighted, wherein wijRepresent label ti
With tjBetween weight, and have wij=wji.Remember x=[x1 x2 … xN]T, then formula (5) matrix form can be written as:
X=η Ax, (6)
X is the eigenvalue of maximum η of the mould of matrix A-1Characteristic vector under correspondence, also referred to as eigenvector centrality;
5) the node clustering coefficient after new label is proposed in 30 days, label tiCluster coefficients be calculated as follows:
Wherein, EiRepresent label tiKiThe side number of physical presence, k between individual neighbours' label nodei(ki- 1) mark/2 is represented
Sign tiKiThe Maximum edge numbers that there may exist between individual neighbor node.
In the step 4, attributive character, which is extracted, to be included:4.1) the interior institute for including the problem of 30 days after new label is proposed
There is answer number;4.2) the average answer number of the presenter and answerer of the label before this is participated in all 30 days and average
Number of questions and average time experience;4.3) in 30 days the label the average problem answers response time;4.4) should in 30 days
All participation user numbers of label, the i.e. presenter of the problem and answerer's sum;4.5) institute of the label is included in 30 days
Problematic average word quantity;4.6) flat spot of all the problems praises number in statistics label 30 days;
Wherein, the calculation of the average problem answers response time of label is as follows:
If including label t in 30 daysiThe problem of number be, 30 days interior label tiThe answer number of s-th of problem beLabel tiS-th of problem creation timeCount the creation time of its v-th of answerCalculate its response time
Difference, then to it is all the problem of and the difference of answer be averagedComputing formula is as follows:
In the step 5, the structure of the disaggregated model of support vector machines two, process is as follows:
First, the selection of kernel function is determined, Gaussian kernel RBF, i.e. sample t is usediAnd tjBetween feature space inner product use
They pass through function k (t in original sample spacei,tj) calculate, its expression formula is as follows:
Wherein, δ represents the bandwidth of Gaussian kernel.
Then carry out finding the parameter optimal value of SVM models by trellis algorithm, then carry out ten folding cross validations, carry out many
Secondary test is averaged, and draws the precision index of the label fashion trend forecast model based on SVM.
Beneficial effects of the present invention are:Compared with prior art, the label prediction of the development trend model based on SVM can be right
Emerging label carries out the prediction of development trend, solves in label recommendations and ignores problem to emerging unexpected winner label,
So that label recommendations are more rationally effective.
Brief description of the drawings
Fig. 1 is intermediate range programming flow diagram of the present invention;
Fig. 2 is the label trend prediction model building process based on SVM in the present invention.
Embodiment
The embodiment of the present invention is described in further detail with reference to Figure of description.
Referring to Figures 1 and 2, a kind of construction method of the label prediction of the development trend model based on SVM, the present invention exists
Carry out including each model creation time, model ID, user in instance analysis, initial data on Stackoverflow data sets
ID, the information such as model label.In this patent by taking the label of problem as an example, we extract label creation time first, and label is carried
The person of going out ID, and its neighbours' label information etc..
The present invention is specifically divided into following five steps:
Step 1:Data set is screened and pre-processed;
Step 2:Choose sample label data;
Step 3:Build label network;
Step 4:Extract the characteristic of sample label;
Step 5:Build and training is based on SVM label fashion trend forecast models.
In the step 1, specific operation process is as follows:Select the information content and corresponding label data of website, selection
Website starting for 3 months after beginning setting up so that the label network of website is preliminarily formed, and counts the frequency of emerging label
Rate, is then ranked up;
In the step 2, the screening of sample label data, specific operation process is as follows:
First, popular exemplar is chosen, descending sort is carried out to label frequency, set is designated asSelectionIn
The label that ratio is preceding 5% is as popular label, and its set is designated as Upop;
Secondly, non-popular exemplar is chosen, set is takenMiddle ratio is used as non-streaming row label for rear 85% label
Set, its set is designated as Qunpop.To each popular label tpop∈Upop, search for and label tpopThe nearest mark of creation time
Sign tunpop, while meeting tunpop∈Qunpop, as non-streaming row label, that is, temporal control is formed, its set is designated as Uunpop;
Finally take U={ Upop,UunpopIt is used as sample label data;
In the step 3, the structure of label network, specific operation process is as follows:In all information for traveling through community data
Hold, if label is appeared in same information record simultaneously, then it represents that the two labels have connection, i.e., two labels have company
Side, the Undirected networks G that has the right of label is built with thisTag, weight represents two labels while the number of times occurred.
In the step 4, the extraction of label network characteristic, as shown in Figure 1, to sample label set U={ Upop,
Uunpop, in the Undirected networks G that has the rightTagOn, extract its interior label tiThe time is being proposed firstThe network characterization of M days afterwards, M
Take 30,.Specific operation process is as follows:
1) the relative degree centrality after new label is proposed in 30 days:Label tiAngle value DiCalculating using removing isolated section
The mode of point, computing formula is as follows:
Wherein, N represents the total number of labels of network;aijThe element of network adjacent matrix is represented, if label tiAnd tjThere is company
Side, then aij=1;Otherwise aij=0;
Label tiThe central feature calculation of degree, take the label t in networkiRelative degree centrality:
Wherein, DiRepresent label tiAngle value;
2) neighbours' average degree centrality after new label is proposed in 30 days, label tiNeighbours' average degree NCiCalculating such as
Under:
Wherein, NneighborRepresent label tiNeighbor node number,Represent label tiNeighbor node degree
It is worth sum.
3) after new label is proposed in 30 days close to centrality, label tiClose central metric calculation, then
Equally take label tiClose to centrad:
Wherein, dijRepresent label tiWith label tjDistance,Represent label tiTo neighbours' label node
Average geodesic distance.
4) eigenvector centrality after new label is proposed in 30 days, label tiEigenvector centrality be calculated as follows:
Wherein, η is a proportionality constant, A=(aijwij) it is the network adjacent matrix weighted, wherein wijRepresent label ti
With tjBetween weight, and have wij=wji.Remember x=[x1 x2 … xN]T, then formula (14) matrix form can be written as:
X=η Ax, (6)
X is the eigenvalue of maximum η of the mould of matrix A-1Characteristic vector under correspondence, also referred to as eigenvector centrality.
5) the node clustering coefficient after new label is proposed in 30 days, label tiCluster coefficients be calculated as follows:
Wherein, EiRepresent label tiKiThe side number of physical presence, k between individual neighbor nodei(ki- 1) label t/2 is representedi
KiThe Maximum edge numbers that there may exist between individual neighbor node.
In the step 4, the extraction of sample label data attribute feature, to label ti∈ U, extract it and propose the time firstThe characterization step in 30 days afterwards is as follows:
4.1) extract and label t is included in 30 daysiAll problems, its set is designated as
4.2) search problem setIn presenter of all the problems, set is designated asIt is of all the problems to answer
Person, set is designated asThumb up number of all the problems, set is designated as
4.3) statistics includes label tiPresenter of all the problemsAnd answererCurrent time it
Preceding average answer number, average problem data;
4.4) label t is countediCorresponding average problem thumb up number, and the label average participation number, that is, answer
Person's number of users and presenter's number of users sum.
4.5) the average problem answers response time for the problem of counting 30 days interior label correspondences, if including label t in 30 daysi
The problem of number be30 days interior label tiThe answer number of s-th of problem beLabel tiS-th of problem establishment
TimeCount the creation time of its v-th of answerCalculate that its response time is poor, then to it is all the problem of and answer
Difference is averagedComputing formula is as follows:
In the step 5, the structure of the label fashion trend forecast model based on SVM and training, specific operation process is such as
Under:First, the selection of kernel function is determined, Gaussian kernel RBF, i.e. sample t is usediAnd tjBetween use them in the inner product of feature space
Pass through function k (t in original sample spacei,tj) calculate, its expression formula is as follows:
Wherein, δ represents the bandwidth of Gaussian kernel;
Then carry out finding the parameter optimal value of SVM models by trellis algorithm, the side of ten folding cross validations is carried out afterwards
Formula, i.e., to data are randomly divided into 10 parts, take 1 part therein to make test sample, remaining 9 parts are made training sample, are drawn successively
Label fashion trend forecast model based on SVM.
The popular classification embodiment introduction of label as described above for the present invention in question and answer website Stackoverflow communities,
By building label network, network characterization and attributive character of the label in 30 days after proposing first, structure are then extracted
The forecast model of the label future developing trend based on SVM is built, reasonable prediction, Yi Jijin are given to emerging label in website
Label recommendations and Information Transmission afterwards are all significant.
Claims (6)
1. a kind of construction method of the label prediction of the development trend model based on SVM, it is characterised in that methods described includes as follows
Step:
Step 1:Data prediction, collects information content label data corresponding with its of Web community, its data content is pressed
Time-sequencing, take community formed N days after data, preliminarily formed with the label network for ensuring community;
Step 2:Sample label is chosen, data set is counted, community's label frequency is obtained and sorts, it is preceding α % to take ratio
Label as popular label, its set is designated as Upop;The label contrasted with the popular label time is chosen in remaining label
For non-streaming row label;
Step 3:Label network is built, to the label occurred in the same information content, that is, thinks there is relation between these labels,
Make its company of being formed side between any two;All information are traveled through, the label network figure G for the Undirected networks that obtain having the rightTag, its interior joint
For emerging label, even side is the relation between label, and the weight of network is the number of times that both occur jointly;
Step 4:Characteristic is extracted, to sample label set U={ Upop,Uunpop, extract M after its interior label is created first
Its network characterization and attributive character, set up sample training data set;
Step 5:Using Machine learning classifiers model supports vector machine SVM, kernel function, mark of the training generation based on SVM are chosen
Fashion trend forecast model is signed, and carries out ten folding cross validations, model result is drawn.
2. a kind of construction method of the label prediction of the development trend model based on SVM as claimed in claim 1, its feature exists
In:In the step 1, the data after choosing N days are used as the data of pretreatment, wherein N selection, it then follows rule is:Ensure
Preceding 10% label data has been generated in N days in website, i.e., the label network in website has been preliminarily formed.
3. a kind of construction method of the label prediction of the development trend model based on SVM as claimed in claim 1 or 2, its feature
It is:In the step 2, the arrangement of line frequency descending is entered in the selection of sample label data to label, and set is designated asSelectionMiddle ratio for preceding α % label as popular label, its set is designated as Upop;Take the label that all label ratios are rear β %
As non-streaming row label set, its set is designated as Qunpop.To each popular label tpop∈Upop, search for and label tpopWound
Time nearest label is built, t is designated asunpop, while meeting tunpop∈Qunpop, as non-streaming row label, to be used as popular label
Control, its set is designated as Uunpop。
4. a kind of construction method of the label prediction of the development trend model based on SVM as claimed in claim 1 or 2, its feature
It is:In the step 4, the network characterization of sample label is extracted, M takes 30, and network characterization includes in the following manner:
1) the relative degree centrality after new label is proposed in 30 days:Label tiAngle value DiCalculating using removing isolated node
Mode, computing formula is as follows:
Wherein, N represents the total number of labels in network;aijThe element of network adjacent matrix is represented, if i.e. label tiAnd tjThere is company
Side, then aij=1, otherwise aij=0;
Label tiThe central feature calculation of degree, take the label t in networkiRelative degree centrality:
Wherein, DiRepresent label tiAngle value;
2) neighbours' average degree centrality after new label is proposed in 30 days, label tiNeighbours' average degree NCiBe calculated as follows:
Wherein, NneighborRepresent label tiNeighbor node number,Represent label tiNeighbor node angle value it
With;
3) after new label is proposed in 30 days close to centrality, label tiClose central metric calculation, then equally take
Label tiClose to centrad:
Wherein, dijRepresent label tiWith label tjDistance,Represent label tiTo the flat of neighbours' label node
Equal geodesic distance;
4) eigenvector centrality after new label is proposed in 30 days, label tiEigenvector centrality be calculated as follows:
Wherein, η is a proportionality constant, A=(aijwij) it is the network adjacent matrix weighted, wherein wijRepresent label tiWith tjIt
Between weight, and have wij=wji.Remember x=[x1 x2 … xN]T, then formula (5) matrix form can be written as:
X=η Ax, (6)
X is that matrix A is characteristic value η-1Characteristic vector under correspondence, also referred to as eigenvector centrality;
5) the node clustering coefficient after new label is proposed in 30 days, label tiCluster coefficients be calculated as follows:
Wherein, EiRepresent label tiKiThe side number of physical presence, k between individual neighbours' label nodei(ki- 1) label t/2 is representedi
KiThe Maximum edge numbers that there may exist between individual neighbor node.
5. a kind of construction method of the label prediction of the development trend model based on SVM as claimed in claim 1 or 2, its feature
It is:In the step 4, the attributive character of sample label is extracted, and the extraction of attributive character includes:
4.1) the interior all answer numbers for including the problem of 30 days after new label is proposed;
4.2) the average answer number and average problem of the presenter and answerer of the label before this are participated in all 30 days
Number and average time experience;
4.3) in 30 days the label the average problem answers response timeIts calculation is as follows:
If including label t in 30 daysiThe problem of number be30 days interior label tiThe answer number of s-th of problem be
Label tiS-th of problem creation timeCount the creation time of its v-th of answerIt is poor to calculate its response time,
Then to it is all the problem of and the difference of answer be averagedComputing formula is as follows:
4.4) in 30 days all participation user numbers, the i.e. problem of the label presenter answerer's sum;
4.5) the average word length of all the problems comprising the label in 30 days;
4.6) flat spot of all the problems comprising the label praises number in 30 days.
6. a kind of construction method of the label prediction of the development trend model based on SVM as claimed in claim 1 or 2, its feature
It is:In the step 5, the structure of the disaggregated model of support vector machines two, process is as follows:
First, the selection of kernel function is determined, Gaussian kernel RBF, i.e. sample t is usediAnd tjBetween use them in the inner product of feature space
Pass through function k (t in original sample spacei,tj) calculate, its expression formula is as follows:
Wherein, δ represents the bandwidth of Gaussian kernel;
Then carry out finding the parameter optimal value of SVM models by trellis algorithm, then carry out ten folding cross validations, repeatedly surveyed
Examination is averaged, and draws the precision index of the label fashion trend forecast model based on SVM.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710127478.4A CN106951471B (en) | 2017-03-06 | 2017-03-06 | SVM-based label development trend prediction model construction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710127478.4A CN106951471B (en) | 2017-03-06 | 2017-03-06 | SVM-based label development trend prediction model construction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106951471A true CN106951471A (en) | 2017-07-14 |
CN106951471B CN106951471B (en) | 2020-05-05 |
Family
ID=59466669
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710127478.4A Active CN106951471B (en) | 2017-03-06 | 2017-03-06 | SVM-based label development trend prediction model construction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106951471B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107544944A (en) * | 2017-09-04 | 2018-01-05 | 江西理工大学 | A kind of SVMs Selection of kernel function method and its application based on graph theory |
CN107644268A (en) * | 2017-09-11 | 2018-01-30 | 浙江工业大学 | A kind of open source software project hatching trend prediction method based on multiple features |
CN108681585A (en) * | 2018-05-14 | 2018-10-19 | 浙江工业大学 | A kind of construction method of the multi-source transfer learning label popularity prediction model based on NetSim-TL |
CN108764537A (en) * | 2018-05-14 | 2018-11-06 | 浙江工业大学 | A kind of multi-source community label prediction of the development trend method based on A-TrAdaboost algorithms |
CN110413657A (en) * | 2019-07-11 | 2019-11-05 | 东北大学 | Average response time appraisal procedure towards seasonal form non-stationary concurrency |
CN112988978A (en) * | 2021-04-27 | 2021-06-18 | 河南金明源信息技术有限公司 | Case trend analysis system in key field of public welfare litigation |
CN113220855A (en) * | 2021-05-27 | 2021-08-06 | 浙江大学 | Development trend analysis method for computer technology field based on IT technical question and answer website |
CN114580588A (en) * | 2022-05-06 | 2022-06-03 | 江苏省质量和标准化研究院 | UHF RFID group tag type selection method based on probability matrix model |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140278355A1 (en) * | 2013-03-14 | 2014-09-18 | Microsoft Corporation | Using human perception in building language understanding models |
CN105183887A (en) * | 2015-09-28 | 2015-12-23 | 北京奇虎科技有限公司 | Data processing method based on browser and browser device |
CN105550275A (en) * | 2015-12-09 | 2016-05-04 | 中国科学院重庆绿色智能技术研究院 | Microblog forwarding quantity prediction method |
CN105654122A (en) * | 2015-12-28 | 2016-06-08 | 江南大学 | Spatial pyramid object identification method based on kernel function matching |
CN105787049A (en) * | 2016-02-26 | 2016-07-20 | 浙江大学 | Network video hotspot event finding method based on multi-source information fusion analysis |
CN106446191A (en) * | 2016-09-30 | 2017-02-22 | 浙江工业大学 | Logistic regression based multi-feature network popular tag prediction method |
CN106447505A (en) * | 2016-09-26 | 2017-02-22 | 浙江工业大学 | Implementation method for effective friend relationship discovery in social network |
-
2017
- 2017-03-06 CN CN201710127478.4A patent/CN106951471B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140278355A1 (en) * | 2013-03-14 | 2014-09-18 | Microsoft Corporation | Using human perception in building language understanding models |
CN105183887A (en) * | 2015-09-28 | 2015-12-23 | 北京奇虎科技有限公司 | Data processing method based on browser and browser device |
CN105550275A (en) * | 2015-12-09 | 2016-05-04 | 中国科学院重庆绿色智能技术研究院 | Microblog forwarding quantity prediction method |
CN105654122A (en) * | 2015-12-28 | 2016-06-08 | 江南大学 | Spatial pyramid object identification method based on kernel function matching |
CN105787049A (en) * | 2016-02-26 | 2016-07-20 | 浙江大学 | Network video hotspot event finding method based on multi-source information fusion analysis |
CN106447505A (en) * | 2016-09-26 | 2017-02-22 | 浙江工业大学 | Implementation method for effective friend relationship discovery in social network |
CN106446191A (en) * | 2016-09-30 | 2017-02-22 | 浙江工业大学 | Logistic regression based multi-feature network popular tag prediction method |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107544944B (en) * | 2017-09-04 | 2020-06-02 | 江西理工大学 | Graph theory-based support vector machine kernel function selection method and application thereof |
CN107544944A (en) * | 2017-09-04 | 2018-01-05 | 江西理工大学 | A kind of SVMs Selection of kernel function method and its application based on graph theory |
CN107644268B (en) * | 2017-09-11 | 2021-08-03 | 浙江工业大学 | Open source software project incubation state prediction method based on multiple features |
CN107644268A (en) * | 2017-09-11 | 2018-01-30 | 浙江工业大学 | A kind of open source software project hatching trend prediction method based on multiple features |
CN108681585A (en) * | 2018-05-14 | 2018-10-19 | 浙江工业大学 | A kind of construction method of the multi-source transfer learning label popularity prediction model based on NetSim-TL |
CN108764537A (en) * | 2018-05-14 | 2018-11-06 | 浙江工业大学 | A kind of multi-source community label prediction of the development trend method based on A-TrAdaboost algorithms |
CN108764537B (en) * | 2018-05-14 | 2021-11-23 | 浙江工业大学 | A-TrAdaboost algorithm-based multi-source community label development trend prediction method |
CN110413657B (en) * | 2019-07-11 | 2021-08-17 | 东北大学 | Average response time evaluation method for seasonal non-stationary concurrency |
CN110413657A (en) * | 2019-07-11 | 2019-11-05 | 东北大学 | Average response time appraisal procedure towards seasonal form non-stationary concurrency |
CN112988978A (en) * | 2021-04-27 | 2021-06-18 | 河南金明源信息技术有限公司 | Case trend analysis system in key field of public welfare litigation |
CN112988978B (en) * | 2021-04-27 | 2024-03-26 | 河南金明源信息技术有限公司 | Case trend analysis system in important field of public service litigation |
CN113220855A (en) * | 2021-05-27 | 2021-08-06 | 浙江大学 | Development trend analysis method for computer technology field based on IT technical question and answer website |
CN113220855B (en) * | 2021-05-27 | 2022-07-22 | 浙江大学 | Computer technology field development trend analysis method based on IT technical question-answering website |
CN114580588A (en) * | 2022-05-06 | 2022-06-03 | 江苏省质量和标准化研究院 | UHF RFID group tag type selection method based on probability matrix model |
Also Published As
Publication number | Publication date |
---|---|
CN106951471B (en) | 2020-05-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106951471A (en) | A kind of construction method of the label prediction of the development trend model based on SVM | |
CN110222267A (en) | A kind of gaming platform information-pushing method, system, storage medium and equipment | |
CN110532379B (en) | Electronic information recommendation method based on LSTM (least Square TM) user comment sentiment analysis | |
CN105095219B (en) | Micro-blog recommendation method and terminal | |
CN112214685A (en) | Knowledge graph-based personalized recommendation method | |
CN108804689A (en) | The label recommendation method of the fusion hidden connection relation of user towards answer platform | |
CN110674407A (en) | Hybrid recommendation method based on graph convolution neural network | |
CN108665323A (en) | A kind of integrated approach for finance product commending system | |
Huang et al. | A multi-source integration framework for user occupation inference in social media systems | |
CN103034963B (en) | A kind of service selection system and system of selection based on correlation | |
CN106934071A (en) | Recommendation method and device based on Heterogeneous Information network and Bayes's personalized ordering | |
WO2020135642A1 (en) | Model training method and apparatus employing generative adversarial network | |
CN111949885B (en) | Personalized recommendation method for scenic spots | |
CN111241394A (en) | Data processing method and device, computer readable storage medium and electronic equipment | |
CN111523055A (en) | Collaborative recommendation method and system based on agricultural product characteristic attribute comment tendency | |
Gu et al. | Application of fuzzy decision tree algorithm based on mobile computing in sports fitness member management | |
CN112749330B (en) | Information pushing method, device, computer equipment and storage medium | |
CN110334278A (en) | A kind of web services recommended method based on improvement deep learning | |
CN111723973A (en) | Learning effect optimization method based on user behavior causal relationship in MOOC log data | |
Fagni et al. | Fine-grained prediction of political leaning on social media with unsupervised deep learning | |
CN115510322A (en) | Multi-objective optimization recommendation method based on deep learning | |
CN111611432A (en) | Singer classification method based on Labeled LDA model | |
CN115936939A (en) | Project recommendation method and system based on multiple preference degrees | |
Zhu | A book recommendation algorithm based on collaborative filtering | |
CN115329215A (en) | Recommendation method and system based on self-adaptive dynamic knowledge graph in heterogeneous network |
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 |