CN109410082A - A kind of online sociodistance's estimation method based on user emotion distribution - Google Patents
A kind of online sociodistance's estimation method based on user emotion distribution Download PDFInfo
- Publication number
- CN109410082A CN109410082A CN201811282747.5A CN201811282747A CN109410082A CN 109410082 A CN109410082 A CN 109410082A CN 201811282747 A CN201811282747 A CN 201811282747A CN 109410082 A CN109410082 A CN 109410082A
- Authority
- CN
- China
- Prior art keywords
- sample
- user
- sociodistance
- online
- distance
- 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 39
- 230000008451 emotion Effects 0.000 title claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 34
- 230000036651 mood Effects 0.000 claims abstract description 13
- 238000005070 sampling Methods 0.000 claims abstract description 12
- 238000004364 calculation method Methods 0.000 claims abstract description 10
- 239000000284 extract Substances 0.000 claims abstract description 5
- 238000007637 random forest analysis Methods 0.000 claims description 17
- 238000003066 decision tree Methods 0.000 claims description 11
- 230000002996 emotional effect Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 8
- 238000005516 engineering process Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 3
- 238000012952 Resampling Methods 0.000 claims description 2
- 238000004422 calculation algorithm Methods 0.000 claims description 2
- 230000018199 S phase Effects 0.000 claims 1
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 4
- 244000097202 Rathbunia alamosensis Species 0.000 description 2
- 235000009776 Rathbunia alamosensis Nutrition 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 1
- 206010027940 Mood altered Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention proposes a kind of online sociodistance's estimation method based on user emotion distribution, including training set building, unbalanced processing, the estimation of online sociodistance；The training set building, according to the user social contact range distribution of entire social networks, equal proportion extracts user distance sample and the social mood distribution of corresponding user.The unbalanced processing improves rare effective identification apart from user couple by the combination sampling of statistical learning method and Nearest Neighbor with Weighted Voting etc., then trains and generate the estimation model to unknown subscriber's distance using sampling information.Online sociodistance's estimation, using estimation model, the online sociodistance being based only upon between mood distribution estimation user.The present invention is not necessarily to any social network structure information, it relies only on user emotion and is distributed the online sociodistance that can accurately estimate between any user, it can be avoided complicated calculations amount caused by directly inferring by network structure and high time cost, and can be widely applied in the scene of common structural information missing.
Description
Technical field
The present invention relates to online sociodistance's estimation method more particularly to a kind of online social activities based on user emotion distribution
Method for estimating distance.
Background technique
With the development of internet, online social platform is rapidly growing.By taking microblogging as an example, by March, 2018, the microblogging moon
Any active ues totally 4.11 hundred million, increase by 21% on a year-on-year basis.Compared with traditional social form, network social intercourse has virtual, diversity, wound
The features such as new property, freedom, user can issue any type of content to express personal mood in social platform.Together
When, with the fast development of online social platform in recent years, the scale of social networks expands rapidly, is in particular in number of users
And user interaction sharply increases.On the one hand the sociodistance's (i.e. shortest distances of two nodes in network) for calculating user is advantageous
It is the basis for further providing for recommending, marketing etc. services in better understanding customer relationship；On the other hand, in large-scale complex
Network in calculate user sociodistance's calculation amount and time cost it is huge, economy and operability are poor.Side of the present invention
Method realizes the accurate estimation to sociodistance merely with the mood distribution of user, compared with tradition distance calculates, in huge compression
While calculating cost, the dependence to structural information is also reduced, greatly extends the applied field of online sociodistance's estimation
Scape.
Summary of the invention
The present invention proposes a kind of online sociodistance's estimation method based on user emotion distribution, characterized by comprising:
Training set building, unbalanced processing, the estimation of online sociodistance；The training set building, according to the overall situation of training social networks
Structure, calculates user social contact range distribution, and equal proportion extracts user distance sample and the social mood distribution of corresponding user.Institute
State unbalanced processing, using the above sampling information, based in statistical learning combination sampling and the methods of Nearest Neighbor with Weighted Voting, training is simultaneously
It generates between the estimation model of distance unknown distance user.Online sociodistance's estimation is realized only sharp based on estimation model
The online sociodistance estimated between social user is distributed with user emotion.
This estimation method relies only on user emotion point under the common application scenario of social networks complete structure loss of learning
Cloth, can be to avoid directly passing through network query function user social contact apart from the huge calculation amount of bring and guarantee higher as mode input
Precision.
Detailed description of the invention
Fig. 1 is composition block diagram of the invention；
Fig. 2 is the user social contact distance estimations model realization flow chart in the present invention；
Fig. 3 is the implementation flow chart of the combination EUS technology classification device in the present invention；
Fig. 4 is the practical application flow chart in the present invention；
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
As shown in Figure 1, this method collect user emotion information, the distribution of different distance in the true social networks of statistics, from
Different user distance samples, the sample using extraction are extracted in social networks in proportion, is used for user social contact distance estimations mould
The training of type, using obtained model, calculate the several steps of sociodistance between the user of other unknown distances.
The online sociodistance is defined it and there is concern (such as in Sina weibo, including a to any two users a, b
The b and b concern behaviors such as a) are paid close attention to, note distance is 1；To any two users a and b, there is no above-mentioned behaviors, but there are another
User c, the distance that the distance for having a and c is 1, b and c are 1, then remember that distance is 2；Above-mentioned pass is not present to any two users a, b
System, and there are another two users c, d, so that the distance of a to c, the distance of d to c, the distance of b to d are 1, then remember that distance is 3；With
The adopted distance 4 of such presumption.The definition of distance is equal to the node being regarded as user in undirected no weight graph.
The user emotion information includes following a few class moos indexes: neutral, angry, glad, low, detest, value model
Enclosing is 0-1, and numerical value is bigger to indicate stronger relative to this kind of mood of other moods for the user.The moos index is based on using
The associated information calculation at family obtains, such as the text list or the favorites list of social media.To any user, all kinds of moos indexes
Summation is 1.
The extraction user distance sample carries out user distance based on true social network user sociodistance distribution proportion
Sampling.Specific implementation are as follows:
Step 4a: several users in social networks are randomly selected, particular number is according to the use in specific training social networks
Family sum is determining, usually about the 1% of whole network user.Calculate all distances of these users and other users.It obtains not
With the distributed number of user distance.
Step 4b: in the entire network, user is randomly selected using breadth first search method, obtains user distance letter
Breath, until the quantity of the sample of all distances reaches the quantity that sampling needs.
Step 4c: it according to the distributed number ratio of different user distance, is randomly selected from the sample moderate proportions obtained
To training data.Arranging is following format: one emotional information of user, two emotional information of user, distance.
The user social contact distance estimations model, as shown in Figure 2, it is characterised in that: secondary classification utilizes random forest knot
EUS technology and ballot method training are closed, the non-uniform sample of distributed pole is adapted to.Specific implementation are as follows:
Step 5a: being divided into two classes similar in the adjacent, quantity for sample category, i.e., distance 1,2 is regarded as one kind, 3,
4 are regarded as one kind, finely tune the sample size of two classes, and the ratio that is allowed to is 1:1.
Step 5b: construction random forest grader is used to distinguish two classes of neotectonics, is trained using two class samples, into
Row preliminary classification.Specific step is as follows for random forest grader:
Indicate that number of samples, M indicate number of features with N.
It determines the number of features m for inputting single sub-classifier, makes m much smaller than M.
Resampling (having duplicate) selects N number of sample from sample.
Decision tree (being used herein as CART decision tree) is established using m feature of this N number of sample as training data.
It repeats above step p times, obtains p decision tree.
Final classification result is determined using ballot method.
Wherein, the feature selecting standard of CART decision tree is Gini coefficient, and Gini coefficient is for indicating single decision herein
The purity of two classes obtained after the node branch of tree.Its calculation formula is:
Wherein, D indicates training set in the sample set of the branch, and c indicates the quantity of class in the set, piIt indicates in the sample
The probability (i.e. ratio) that the i-th class occurs in this set.The value range of Gini coefficient is 0-1, and numerical value is smaller, indicates set sample
This is purer, when Gini coefficient is 0, only comprising a kind of sample in set D.If branching selection attribute A as the criteria for classifying,
The Gini coefficient of data set after branch calculates are as follows:
Wherein k is branch node number.In random forests algorithm, when selection divides attribute A, to consider to be allowed to Gini coefficient
Gain is maximum, the calculation formula of Gini coefficient gain are as follows:
Δ Gini (A)=Gini (D)-GiniA(D)。
Step 5c: the random forest grader of the construction respective finer combination EUS technology of two classifications and method of voting,
Such as Fig. 3, it is trained using the respective sample of two classes.
The random forest grader of the combination EUS technology and ballot method, it is characterised in that: it is non-uniform to adapt to distributed pole
The differentiation of classification.Specific implementation are as follows:
Step 6a: calculate the unbalance rate IR of sample, it is assumed here that positive sample be a fairly large number of one kind, negative sample be quantity compared with
Few one kind
Step 6b: a fairly large number of sample is carried out IR times without the double sampling put back to, extracts n every time(negative)A sample, and it is negative
One group of sample of sample Compositional balance, is put into random forest grader and is trained.Similarly generate IR classifier.Training knot
Shu Hou verifies the accuracy rate of the classifier using verifying collection.If accuracy > 50%, then receive the classifier；Otherwise,
Refuse the classifier, and the positive sample of extraction is put back into training set, sample drawn is trained again.
Step 6c: integrating above-mentioned IR classifier using ballot method, and each classifier has a different weights, specific weight by
The accuracy rate of above-mentioned verifying collection determines.For i-th of classifier, weight are as follows:
Wherein, εiFor accuracy rate of i-th of classifier on verifying collection, final classification device is obtained.
Classifying step such as Fig. 4 of final practical application.
One embodiment is the concern relation distance between the mood distribution and its user of user in Sina weibo, it should be noted that
, due to similitude of the social networks on Forming Mechanism and structure, the estimation model that this analysis example obtains is in other societies
It hands over and is likewise supplied with similar estimated capacity on network, and present implementation can completely migrate to other social networks.It is specific real
Apply this method process such as Fig. 1.
The known users social networks defines a undirected no weight graph in this example, the concern relation of user is regarded as
A line.Whole users are regarded into vertex.I.e. entire social networks is using user as vertex, and concern relation is a net on side
Network.Since number of users is larger, it is believed that the cyberrelationship tallies with the actual situation.
The user emotion information utilizes the microblogging item number of the different moods of user and the ratio of microblogging total number in this example
Example, as the emotional information of user, the emotional information generation step are as follows:
Step 1, a certain user is selected, the item number of the microblogging of its different mood is counted.
Step 2, total with the microblogging quantity and microblogging of every kind of mood using the microblog number summation being in a bad mood as microblogging sum
Value of several ratios as each moos index.
The sum of the microblogging quantity of different emotions for paying attention to microblogging quantity used herein and not all, but counting.Mesh
Be make the sum of each moos index be 1, convenient for comparing.
It is described to utilize accurate sociodistance between network query function certain customers, it is therefore intended that determine all types of user in network
The ratio of distance.Implement step are as follows:
Step 1, a vertex (user) optionally in above-mentioned social networks, calculate the user and other all users away from
From.Being used herein as Dijkstra method can be realized.Count the quantity of the relationship of different distance.
Step 2, alternative vertex, repeats step 1 several times, until the ratio of all kinds of distances is basically unchanged.In this example, it selects
Fixed 400 points calculate it at a distance from other points.
It is described to sample (sampling) according to actual user apart from quantitative proportion, implement step are as follows:
Step 1, a vertex (user) of optional above-mentioned social networks searches all vertex for being 1 with its distance, by this
A little vertex are stored in another interim chained list, while remembering that these vertex are 1 at a distance from initial point.
Step 2, from temporary table select first point, which is put into described in queue step 1, search with the point away from
From the vertex for 1, remember the point at a distance from these points for 1, point in note initial point and these points not in interim chained list away from
From being 2.And these points are put into the end of interim chained list.
Step 3, with step 2, next vertex in chained list is selected to be calculated, until having traversed whole vertex.
Step 4, optional another point is calculated at the beginning from step, until the side of all distances reaches specified quantity.
Step 5: be following format by the sample record of extraction: (1 emotional information of user, 2 emotional information of user, user 1,2
The distance between): (e11, e12, e13, e14, e15, e21, e22, e23, e24, e25, y).
It should be noted that there are many method for calculating user distance in network, it is contemplated that sampling will meet really
Range distribution, while unnecessary calculation amount is reduced, therefore select breadth first search, it is ensured that oversampling ratio substantially conforms to
Distribution and guarantee stochastical sampling.
The online sociodistance's classifier of generation user, specific implementation are as follows:
Step 1, distance for 1,2 sample will be regarded as the first kind, distance is considered as the second class for 3,4 sample, and adjust away from
From the sample size for 2 and 3, keep first kind sample and the second class sample size identical.
Step 2, the distance sample of above-mentioned extraction is divided into test set, inspection set, training set in the ratio of 2:1:7, it is different
Each distance sample ratio in data set is identical with total data set distribution proportion.
Step 3, using training set data one random forest grader of training, for distinguishing two classifications.
Step 4, the first kind sample in training set data, one combination EUS method based on random forest of training are utilized
Classifier.It is implemented as follows:
Step 4a remembers that one kind of negligible amounts in training set sample is negative example, it is a fairly large number of it is a kind of be positive example, for this
, negative example is the sample that distance is 1 in the first kind, and positive example is the sample that distance is 2；Negative example is the sample that distance is 4 in second class
This, positive example is the sample that distance is 3.Calculate the unbalance rate IR of sample
Wherein n indicates the quantity of the type sample.
Step 4b extracts n with not putting back to from the positive example of training set(negative example)A sample, with existing n(negative example)A negative example composition
New data set, for training the random forest grader for distinguishing positive example and negative example.
Step 4c concentrates first kind sample to verify the step 4b classifier generated, calculates accuracy using verifying,
It is denoted as εi.If εi> 50%, then receive the classifier, remembers the weight α of the classifieriAre as follows:
Otherwise, refuse the classifier, and the positive sample of extraction is put back into training set, sample drawn is trained again.
Step 4d is repeated step b, c IR times, obtains IR effective classifiers, using the mode of Nearest Neighbor with Weighted Voting, is determined
Final classification results.It is inputted IR classifier, obtains IR classification results, calculates IR by the sample for treating judgement
It is the correspondence α of positive example in classification resultsiSummation and result be negative the correspondence α of exampleiSummation, both compare, access value is larger
It is final result.
Step 4e similarly generates the classifier of the positive example and negative example for the second class sample according to the method described above.
In this example, which estimates the user distance information of actual proportions, and it is as follows to obtain confusion matrix:
The estimation method overall estimation accuracy reaches 0.53, and has preferable accuracy of estimation to every one kind, in reality
During testing, we have been attempted using random forest, and decision tree, the simple classification method such as KNN, these methods are in all kinds of distances
Show preferable classifying quality when sample proportion is close, but the detection when facing the sample of actual proportions, to small sample
Rate is greatly reduced, and does not meet actual operation requirements.And this method can well adapt to actual distance ratio in extensive social networks
The data of example, and due to the structure proximate of extensive social networks, this estimation method has one in face of a variety of social networks
Determine effect.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify to technical solution documented by previous embodiment or equivalent replacement of some of the technical features；And
These are modified or replaceed, the spirit and model of technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution
It encloses.
Claims (7)
1. a kind of online sociodistance's estimation method based on user emotion distribution, specifically includes the following steps: step 1, is collected
User emotion information；Step 2, the distribution of different distance in social networks is counted；Step 3, it is extracted in proportion from social networks
Different user distance samples；Step 4, the training of user social contact distance estimations model is carried out using the sample of extraction, and is being instructed
The lack of uniformity of different distance sample is solved during practicing；Step 5, using obtained estimation model, estimate any social networks
In online sociodistance between any user couple.
2. the method according to claim 1, wherein to any two users a, b, if there is concern relation,
Then the online sociodistance is 1；To any two users a and b, there is no concern relations, but there are another user c, have a and
The online sociodistance that the online sociodistance of c is 1, b and c is 1, then the online sociodistance of a and c is 2；To any two users a,
Concern relation is not present in b, and there are another two users c, d so that the online sociodistance of a to c, d to c online sociodistance,
The online sociodistance of b to d is that the online sociodistance of 1, then a and d is 3.
3. the method according to claim 1, wherein the user emotion information includes following moos index: in
Property, indignation, happiness, it is low, detest, the value range of the moos index is 0-1, and the bigger expression of numerical value is for user's phase
Mood this kind of for other moods is stronger, and the moos index is obtained based on the associated information calculation of user, the related letter
Breath includes the text list or the favorites list of social media, and to any user, all kinds of moos index summations are 1.
4. the method according to claim 1, wherein the distribution of different distance is based in the statistics social networks
True social network user sociodistance distribution, specific implementation are as follows:
Step 4a randomly selects several users, and particular number is determined according to the total number of users in training set network, calculates these use
The distance at family and other users, and obtain the distributed number of different distance.
Step 4b, using breadth first search method, obtains user distance information in entire training set network, until all
The quantity of the sample of distance reaches the quantity that sampling needs.
Step 4c is randomly selected from the sample moderate proportions obtained and is instructed according to the distributed number ratio of different user distance
Practice data, and by building training set.
5. the method according to claim 1, wherein the training of the user social contact distance estimations model, specifically
It realizes are as follows:
Four class samples are recombinated and are divided into two classes similar in the adjacent, quantity by step 5a, adjust the distance for 1,2,3,4 four
Class is regarded as one kind for 1,2, and 3,4 are regarded as one kind, finely tune the sample size of two classes, and the ratio that is allowed to is 1:1.
Step 5b, construction random forest grader is used to distinguish two classes of neotectonics, and carries out model training using two class samples,
Sample input format are as follows: (1 emotional information of user, 2 emotional information of user, the distance between user 1 and 2) is denoted as (e11, e12,
e13, e14, e15, e21, e22, e23, e24, e25, y), to realize the preliminary classification of sociodistance.
Step 5c constructs the random forest grader of the respective combination EUS technology of two classifications respectively, utilizes the respective sample of two classes
Originally it is trained.
6. according to the method described in claim 5, the specific implementation steps are as follows for the random forest grader:
It indicates that number of samples, M indicate number of features with N, determines the number of features m for inputting single sub-classifier, be much smaller than m
M, resampling selects N number of sample from sample, establishes decision tree as training data using m feature of this N number of sample, described
Decision tree uses CART decision tree, repeats above step p times, obtains p decision tree, determines final classification knot using ballot method
The feature selecting standard of fruit, the CART decision tree is Gini coefficient, and the Gini coefficient is used to indicate the section of single decision tree
The purity of two classes obtained after point branch, its calculation formula is:
Wherein, D indicates training set in the sample set of the branch, and c indicates the quantity of class in the set, piIt indicates in the sample set
The probability that the i-th class occurs in conjunction, the value range of the Gini coefficient is 0-1, and numerical value is smaller, indicates that set sample is purer, when
When Gini coefficient is 0, only comprising one kind sample in set D, if branching selection attribute A as the criteria for classifying, after branch
The Gini coefficient of data set calculates are as follows:
Wherein k is the number of branch node.In random forests algorithm, when selection divides attribute A, to consider to be allowed to Gini coefficient
Gain is maximum, the calculation formula of Gini coefficient gain are as follows:
Δ Gini (A)=Gini (D)-GiniA(D), the N, M, m, p, i, j, k are positive integer.
7. according to the method described in claim 5, it is characterized in that, the random forest grader of the combination EUS technology, into
The specific implementation of row training are as follows:
Specific implementation are as follows:
Step 7a: the unbalance rate IR of sample is calculated, it is assumed here that positive sample is a fairly large number of one kind, and negative sample is negligible amounts
It is a kind of
Step 7b: a fairly large number of sample is carried out IR times without the double sampling put back to, extracts n every time(negative)A sample, with negative sample
One group of sample of Compositional balance, is put into random forest grader and is trained, and then generates IR classifier.Training terminates
Afterwards, the accuracy rate of the classifier is verified using verifying collection.If accuracy > 50%, then receive the classifier；Otherwise, it refuses
The exhausted classifier, and the positive sample of extraction is put back into training set, sample drawn is trained again.
Step 7c: above-mentioned IR classifier is integrated using ballot method, each classifier has different weights, and specific weight is by above-mentioned
The accuracy rate of verifying collection determines.For i-th of classifier, weight are as follows:Wherein, εiIt is being verified for i-th of classifier
Accuracy rate on collection obtains final classification device, and the IR, i are positive integer, the n(negative)For the number of negative sample, n(just)It is positive
The number of sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811282747.5A CN109410082A (en) | 2018-10-31 | 2018-10-31 | A kind of online sociodistance's estimation method based on user emotion distribution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811282747.5A CN109410082A (en) | 2018-10-31 | 2018-10-31 | A kind of online sociodistance's estimation method based on user emotion distribution |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109410082A true CN109410082A (en) | 2019-03-01 |
Family
ID=65470326
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811282747.5A Pending CN109410082A (en) | 2018-10-31 | 2018-10-31 | A kind of online sociodistance's estimation method based on user emotion distribution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109410082A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2020130528A (en) * | 2019-02-18 | 2020-08-31 | 沖電気工業株式会社 | Emotion estimation device, emotion estimation method, program, and emotion estimation system |
| CN112561705A (en) * | 2020-12-28 | 2021-03-26 | 杭州趣链科技有限公司 | Alliance link autonomous method, device, equipment and storage medium based on artificial intelligence |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080004941A1 (en) * | 2004-12-23 | 2008-01-03 | Hermann Calabria | Social-Network Enabled Review System With Social Distance Based Syndication |
| CN103136303A (en) * | 2011-11-24 | 2013-06-05 | 北京千橡网景科技发展有限公司 | Method and equipment of dividing user group in social network service website |
| US20130275352A1 (en) * | 2012-04-17 | 2013-10-17 | The Mitre Corporation | Identifying and Forecasting Shifts in the Mood of Social Media Users |
| US8595167B1 (en) * | 2010-11-30 | 2013-11-26 | Google Inc. | Predicting likelihood of a successful connection between unconnected users within a social network using a learning network |
| CN103562906A (en) * | 2011-06-02 | 2014-02-05 | 微软公司 | Emotion-based user identification for online experiences |
| CN103795613A (en) * | 2014-01-16 | 2014-05-14 | 西北工业大学 | Method for predicting friend relationships in online social network |
| CN103995909A (en) * | 2014-06-17 | 2014-08-20 | 东南大学成贤学院 | Online user relation measurement and classification method based on three-dimensional relation strength model |
| CN107145527A (en) * | 2017-04-14 | 2017-09-08 | 东南大学 | Link prediction method based on first path in alignment isomery social networks |
-
2018
- 2018-10-31 CN CN201811282747.5A patent/CN109410082A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080004941A1 (en) * | 2004-12-23 | 2008-01-03 | Hermann Calabria | Social-Network Enabled Review System With Social Distance Based Syndication |
| US8595167B1 (en) * | 2010-11-30 | 2013-11-26 | Google Inc. | Predicting likelihood of a successful connection between unconnected users within a social network using a learning network |
| CN103562906A (en) * | 2011-06-02 | 2014-02-05 | 微软公司 | Emotion-based user identification for online experiences |
| CN103136303A (en) * | 2011-11-24 | 2013-06-05 | 北京千橡网景科技发展有限公司 | Method and equipment of dividing user group in social network service website |
| US20130275352A1 (en) * | 2012-04-17 | 2013-10-17 | The Mitre Corporation | Identifying and Forecasting Shifts in the Mood of Social Media Users |
| CN103795613A (en) * | 2014-01-16 | 2014-05-14 | 西北工业大学 | Method for predicting friend relationships in online social network |
| CN103995909A (en) * | 2014-06-17 | 2014-08-20 | 东南大学成贤学院 | Online user relation measurement and classification method based on three-dimensional relation strength model |
| CN107145527A (en) * | 2017-04-14 | 2017-09-08 | 东南大学 | Link prediction method based on first path in alignment isomery social networks |
Non-Patent Citations (3)
| Title |
|---|
| JOHNJOHNGA: "https://stackoverflow.com/questions/7220232/compute-social-distance-between-two-users", 《COMPUTE SOCIAL DISTANCE BETWEEN TWO USERS》 * |
| 梁霄，赵吉昌，许可: "社交网络用户的社交关系和签到行为分析", 《科技导报》 * |
| 罗梁等: "跨社交网络的实体用户关联技术研究", 《信息网络安全》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2020130528A (en) * | 2019-02-18 | 2020-08-31 | 沖電気工業株式会社 | Emotion estimation device, emotion estimation method, program, and emotion estimation system |
| JP7172705B2 (en) | 2019-02-18 | 2022-11-16 | 沖電気工業株式会社 | Emotion estimation device, emotion estimation method, program, and emotion estimation system |
| CN112561705A (en) * | 2020-12-28 | 2021-03-26 | 杭州趣链科技有限公司 | Alliance link autonomous method, device, equipment and storage medium based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110221965B (en) | Test case generation method, test case generation device, test case testing method, test case testing device, test equipment and test system | |
| CN106651519B (en) | Personalized recommendation method and system based on label information | |
| CN104899267B (en) | A kind of integrated data method for digging of social network sites account similarity | |
| CN105787025B (en) | Network platform public account classification method and device | |
| CN105824813B (en) | A kind of method and device for excavating core customer | |
| CN109711422A (en) | Image real time transfer, the method for building up of model, device, computer equipment and storage medium | |
| CN108647800B (en) | Online social network user missing attribute prediction method based on node embedding | |
| Li et al. | A method of purchase prediction based on user behavior log | |
| CN106708953A (en) | Discrete particle swarm optimization based local community detection collaborative filtering recommendation method | |
| CN110334356A (en) | Article matter method for determination of amount, article screening technique and corresponding device | |
| CN107256241B (en) | Movie recommendation method for improving multi-target genetic algorithm based on grid and difference replacement | |
| CN107305545A (en) | A kind of recognition methods of the network opinion leader based on text tendency analysis | |
| CN109410082A (en) | A kind of online sociodistance's estimation method based on user emotion distribution | |
| CN104217088B (en) | The optimization method and system of operator's mobile service resource | |
| CN110069545A (en) | A kind of behavioral data appraisal procedure and device | |
| CN104965846B (en) | Visual human's method for building up in MapReduce platform | |
| CN115775026A (en) | Federated learning method based on organization similarity | |
| CN108563786A (en) | Text classification and methods of exhibiting, device, computer equipment and storage medium | |
| Sirotin et al. | Information and telecommunication aspect of Russian regions development | |
| Han et al. | Identifying different community members in complex networks based on topology potential | |
| CN112651790B (en) | OCPX self-adaptive learning method and system based on user touch in quick-elimination industry | |
| Gloor et al. | Identifying tribes on twitter through shared context | |
| Liu et al. | Worker recommendation for crowdsourced q&a services: A triple-factor aware approach | |
| CN110489660A (en) | A kind of user's economic situation portrait method of social media public data | |
| CN107203632B (en) | Topic Popularity prediction method based on similarity relation and cooccurrence relation |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190301 |