CN106202488B - Method of the estimation user to physical event distance - Google Patents
Method of the estimation user to physical event distance Download PDFInfo
- Publication number
- CN106202488B CN106202488B CN201610569794.2A CN201610569794A CN106202488B CN 106202488 B CN106202488 B CN 106202488B CN 201610569794 A CN201610569794 A CN 201610569794A CN 106202488 B CN106202488 B CN 106202488B
- Authority
- CN
- China
- Prior art keywords
- event
- user
- information
- formula
- social
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Abstract
The invention discloses a kind of method that estimation user arrives physical event distance, for solve it is existing predict user to the method for physical event distance the practicability is poor the technical issues of.Technical solution is that user is associated with event, the position occurred based on event, by comparing event and user in physics, the similitude of information and social three feature spaces, and combine the user data of known position information, Gaussian process regression model is constructed, the distance of user to the event unknown to location information is estimated.It can be in physics, information and social three dimensions, explain the mobility of user, and by the way that user and event to be associated, the Behavior law and mode of user's deeper are excavated, there is very strong realistic meaning in the application scenarios such as public security and social security management.
Description
Technical field
The invention belongs to social network data excavate and analysis field, in particular to a kind of estimation user to physical event away from
From method.
Background technique
With the development of social networks, the thing occurred at one's side is published in social media by more and more users
(such as: Sina weibo, Facebook and Twitter etc.).However, due to being related to individual privacy, most users are unwilling
Share the location information of oneself, in order to solve the problems, such as that user information is unknown, extensive work user oriented action trail study with
Position prediction expansion.
Document 1 " number of patent application is 201410104399.8 Chinese invention patent " discloses a kind of mobile based on user
The position predicting method of rule excavates user's Move Mode, from historical movement path by the space-time data of research user
User's movement rule is excavated, solves the problems, such as the quick response and high-precision of location of mobile users prediction.
Document 2 " number of patent application is 201510073153.3 Chinese invention patent " discloses a kind of position prediction system
System, the system are divided into three big modules, and input, position prediction and output module: input module is used to receive the number of registering of user
According to or position prediction request;Position prediction module predicts user location by combining prediction model and user data;It is defeated
Module is then used to show the position prediction result out.This method is predicted using position of the probabilistic model to user.
In addition to this, number of patent application is 201310518476.X, 201110308289.X and 200810218368.X
Chinese invention patent all discloses position predicting method and system model based on user's history trace information, however, existing
Method in, do not have and propose the method that is associated the place of the position of user and event, only by historical track to user
Position is predicted that obtained result is mostly data value (GPS coordinate), can not application-oriented offer user and reality
The incidence relation of event;Since the behavioural characteristic of people is often mutually related with social event, how user and event are explained
Relationship, physics when being occurred by event, society and the feature on information space, the incidence relation for excavating user and event exist
The application scenarios such as public security, social security management have extremely important meaning.
Summary of the invention
In order to overcome the shortcomings of existing prediction user, to the method for physical event distance, the practicability is poor, and the present invention provides one kind
Method of the estimation user to physical event distance.This method is associated with event by user, based on the position that event occurs, passes through
Compare event and user in physics, the similitude of information and social three feature spaces, and combines the user of known position information
Data, construct Gaussian process regression model, and the distance of user to the event unknown to location information is estimated.It can be in object
Reason, information and social three dimensions explain the mobility of user, and by the way that user and event to be associated, excavate user
The Behavior law and mode of deeper have very strong reality meaning in the application scenarios such as public security and social security management
Justice.
A kind of the technical solution adopted by the present invention to solve the technical problems: estimation user to the side of physical event distance
Method, its main feature is that the following steps are included:
Step 1: screening user in social networks according to the subject key words of event, and extract use relevant to event
User data;
Assuming that the keyword of event is EW, the period of generation is ETP, then all that keyword EW is referred in time ETP
User will be screened as user relevant to event;For these users, its history number is obtained using web crawlers tool
According to the data model of building user's individual is expressed as formula (1)
RU=< L, C, F > (1)
Wherein, L indicates that the history of the user is registered data, and C indicates the status information issued in user's history, and F indicates to use
Friend information of the family in social networks.
Step 2: the subject content of location information, participant's information when being occurred according to event and event, constructs event
Character representation model;
Geographical location for event in physics, the feature of three aspects of information and social space, when binding events occur
Information, participant's information and event topic construct the feature representation model of event, are expressed as formula (2)
EF=<CM, ET, EA>(2)
Wherein, CM indicates event in the feature of physical space, by the history moving rail for extracting all event participants
Mark sets up the physical space attribute that One-male unit feature is used to indicate event, is in the user's history of all participation events to thing
The probability-distribution function of part positional distance;Assuming that the history of each participant is registered, sequence is PLS, and the position of current event is
EL, pdiIndicate any one PLSiTo the distance of EL, then calculated all pdiIndicate that some participant arrives event location
Range distribution, then, the pd of all participantsiThe set of composition is exactly range distribution of the participant group to event location, i.e.,
It is a kind of probability-distribution function about distance for group mobility feature CM.
ET indicates that event in the feature of information space, is obtained by the topic and descriptor of extraction event;Every participation
Person can issue the state in relation to episode topic in event generation period, by extracting the content of text of all participants' publications,
And keyword is extracted, keyword feature vector is constituted, each dimension indicates a kind of keyword, finally constitutes event in information space
Feature ET.
EA indicates event in the feature of social space, by the user information and their social activity of extracting participation event
Relationship obtains.
Step 3: extracting user's individual in physics, information and society based on user data relevant to event in step 1
The character representation model in space;
User's individual passes through the user established in step 1 in the feature representation model of physics, information and social three spaces
Data model export, is expressed as formula (3)
UF=<IM, HC, RF>(3)
Wherein, IM is measured by register place and current event distance location of the history in user data, and expression is to use
Probability-distribution function of the family to incident distance;Assuming that the history of user is registered, location sequence is LS, and the position of current event is EL,
diIndicate any LS in location sequenceiTo the distance of EL, then, all d being calculatediProbability distribution be then IM, i.e.,
A kind of probability-distribution function about distance.
HC is obtained by extracting history text of the user in social media, and expression is often referred in user's history
Text information;By the extraction to history text information key, keyword feature vector is constructed, each dimension indicates a kind of and closes
Keyword information can compare the topic keyword feature of event, calculate the similitude of the two.
RF is obtained by extracting in the every text message of user with the interactive information of good friend, and expression is that user is frequent in the recent period
The social friend information of communication.
Step 4: be directed to Step 2: the event that constructs in step 3 and user's personal feature indicate model, define user with
Event is in physics, the similitude of information and social three feature spaces, the degree of association based on this building user and event;
For Step 2: the event and user's personal feature model that step 3 constructs respectively, define the two feature representations
Model and then measures the degree of association between user and event in the similitude of physics, information and social space;For user u with
And event e, physical space similitude are expressed as formula (4)
Wherein, IM (d) indicates the probability-distribution function in user's individual historical track with respect to event location distance, CM (d)
Probability-distribution function of the expression event participant group relative to event location distance respectively corresponds formula (3), in formula (2)
IM and CM.The measured value is smaller, shows that designated user is higher with corresponding event correlation in physical space feature.
Secondly, the two information space similitude is expressed as formula (5)
Wherein, C (u) and C (e) respectively refer to alternative family u and event e in the feature of information space, i.e. formula (3) and formula
(2) theme ET when user's history text information HC and event in occur, the formula calculate the cosine similarity of the two, value
It is bigger, show that designated user is higher with corresponding event correlation in information space feature;Wherein, due to the history of user's publication
Text has chronological order, and the period occurred closer to current event, content can more show the current text of user
Interest preference, therefore time factor is considered when constructing user keyword feature vector C (u), expression such as formula (6)
With formula (7)
C (u)=< w1,w2,......,wn> (6)
Wherein, wiIt indicates certain one-dimensional keyword weight, is calculated by formula (7);Wherein, wi,jIndicate i-th dimension
Keyword is in tjThe number that moment occurs, TeAt the time of when expression event occurs, by calculating, finally obtain per one-dimensional keyword
Weight size, the text key word issued when occurring closer to event, weight is higher.
Again, in social space, the similitude of user u and event e are expressed as formula (8)
Wherein, S (u) and S (e) respectively refer to alternative family u and event e in the feature of social space, i.e. formula (3) and formula
(2) the participant information EA when friend information RF and event that the user in often interacts in the recent period occur, the measured value is bigger,
It is bigger to illustrate that the good friend of designated user participates in number ratio shared by corresponding event, i.e., the user in social space's feature with it is corresponding
Event correlation is higher.
Finally, the association table of user and social event is shown as formula (9)
Correlation (u, e)=M (u, e)-1+C(u,e)+S(u,e) (9)
Wherein, M (u, e), C (u, e) and S (u, e) are respectively above-mentioned user and event in physics, information and social characteristic
The similitude in space;Since the value of M (u, e) is smaller, show that user is higher with event correlation, thus it is inverted to M (u, e), make
It obtains during constructing user and event correlation degree, final result Correlation (u, e) value is bigger, shows user and thing
The degree of association of part is higher.
Step 5: the degree of association based on user and event, in conjunction with the user data of known position information, training Gaussian process
Regression model, the distance of the unknown user of estimated position information to event;
By fusion user and social event in physics, information and the similitude of social space, i.e. merging formula (4), public affairs
The calculated result of formula (5) and formula (6), and the user data of known position information is combined to establish Gaussian process regression model GPR,
It is input with user and the degree of association Correlation (u, e) of event, is output, training with the distance of user to location of incident
Gaussian process regression model.And pass through model, the degree of association based on user and event, the unknown user of estimated position information and thing
The distance of part.Model calculation formula such as (10)
Distance(uestimate, e) and=GPRTrained(Correlation(uestimate,e)) (10)
Wherein, GPRTrainedExpression is trained using the user and the corresponding customer incident degree of association of known position information
Gaussian process regression model, Correlation (uestimate, e) and indicate to need the user that estimates, the degree of association with event,
Distance(uestimate, e) indicate the user to be estimated to incident distance value size.
The beneficial effects of the present invention are: this method is associated with event by user, based on the position that event occurs, pass through ratio
Compared with event and user in physics, the similitude of information and social three feature spaces, and combine the number of users of known position information
According to building Gaussian process regression model, the distance of user to the event unknown to location information is estimated.Can physics,
Information and social three dimensions explain the mobility of user, and by the way that user and event to be associated, it is deeper to excavate user
The Behavior law and mode of layer, have very strong realistic meaning in the application scenarios such as public security and social security management.
It elaborates with reference to the accompanying drawings and detailed description to the present invention.
Detailed description of the invention
Fig. 1 is the flow chart of method of the estimation user of the invention to physical event distance.
Specific embodiment
Referring to Fig.1.The present invention estimates user, and to the method for physical event distance, specific step is as follows:
Step 1: screening user in social networks according to the subject key words of event, and extract use relevant to event
User data;
Theme and topic word when event being used to occur compare the ownness of social network user publication as keyword
Whether the content of information includes event keyword according to its content, and user is divided into two parts: it is relevant to event and
It is unrelated with event.This method is unfolded to study just for user relevant to event, and the user unrelated with event is not as we
The research object of method.Assuming that the keyword of event is EW (Event Words), the period of generation is ETP (Event Time
Period), then all users that keyword EW is referred in time ETP will be screened as user relevant to event;For
These users obtain its historical data using web crawlers tool, construct the data model of user's individual, be expressed as formula (1)
RU=<L, C, F>(1)
Wherein, L indicates that the history of the user is registered data (Location), and C indicates the state issued in user's history letter
Ceasing (Contents) and F indicates friend information (Friends) of the user in social networks.
Step 2: the subject content of location information, participant's information when being occurred according to event and event, constructs event
Character representation model;
Geographical location for event in physics, the feature of three aspects of information and social space, when binding events occur
Information, participant's information and event topic construct the feature representation model of event, are expressed as formula (2)
EF=<CM, ET, EA>(2)
Wherein, CM indicates event in the feature of physical space, by the history moving rail for extracting all event participants
Mark sets up the physical space attribute that One-male unit feature (Collective Mobility) is used to indicate event, is all participations
The probability-distribution function of event location distance is arrived in the user's history of event;Assuming that the history of each participant is registered, sequence is
PLS (Participant Location Sequence), the position of current event are EL (Event Location), pdiIt indicates
Any one PLSiTo the distance of EL, then calculated all pdiIndicate some participant to event location range distribution,
So, the pd of all participantsiThe set of composition is exactly range distribution of the participant group to event location, as group mobility
Feature CM is a kind of probability-distribution function about distance.
ET indicates that event in the feature of information space, passes through the topic (Event Topic) and descriptor of extraction event
It obtains;Every participant can issue the state in relation to episode topic in event generation period, pass through and extract all participants' hairs
The content of text of cloth, and keyword is extracted, keyword feature vector is constituted, each dimension indicates a kind of keyword, finally constitutes
Feature ET of the event in information space.
EA indicates event in the feature of social space, by the user information (Event for extracting participation event
Attendees) and their social networks obtain.
Step 3: extracting user's individual in physics, information and society based on user data relevant to event in step 1
The character representation model in space;
User's individual passes through the user established in step 1 in the feature representation model of physics, information and social three spaces
Data model export, is expressed as formula (3)
UF=<IM, HC, RF>(3)
Wherein, IM (Individual Mobility) by the history in user data register place (Locations) with
Current event distance location measurement, expression is probability-distribution function of the user to incident distance;Assuming that the history of user is registered
Location sequence is LS (Location Sequence), and the position of current event is EL (Event Location), diIndicate place
Any LS in sequenceiTo the distance of EL, then, all d being calculatediProbability distribution be then IM, i.e., it is a kind of about away from
From probability-distribution function.
HC is obtained by extracting history text of the user in social media, and expression is often referred in user's history
Text information (Historical Contents);By the extraction to history text information key, construct keyword feature to
Amount, each dimension indicates a kind of key word information, can compare the topic keyword feature of event, calculates the similitude of the two.
RF is obtained by extracting in the every text message of user with the interactive information of good friend, and expression is that user is frequent in the recent period
The social friend information (Recent Friends) of communication.
Step 4: be directed to Step 2: the event that constructs in step 3 and user's personal feature indicate model, define user with
Event is in physics, the similitude of information and social three feature spaces, the degree of association based on this building user and event;
For Step 2: the event and user's personal feature model that step 3 constructs respectively, define the two feature representations
Model and then measures the degree of association between user and event in the similitude of physics, information and social space;For user u with
And event e, physical space similitude are expressed as formula (4)
Wherein, IM (d) indicates the probability-distribution function in user's individual historical track with respect to event location distance, CM (d)
Probability-distribution function of the expression event participant group relative to event location distance respectively corresponds formula (3), the IM in (2)
With CM.The measured value is smaller, shows that designated user is higher with corresponding event correlation in physical space feature.
Secondly, the two information space similitude is expressed as formula (5)
Wherein, C (u) and C (e) respectively refer to alternative family u and event e in the feature of information space (Content), i.e. formula
(3) theme ET when user's history text information HC and event and in (2) occur, the cosine which calculates the two are similar
Degree, value is bigger, shows that designated user is higher with corresponding event correlation in information space feature;Wherein, since user issues
History text have chronological order, closer to current event occur period, it is current that content can more show user
Text interest preference, therefore time factor is considered when constructing user keyword feature vector C (u), expression is such as
Formula (6) and (7)
C (u)=< w1,w2,......,wn〉 (6)
Wherein wiIt indicates certain one-dimensional keyword weight, can be calculated by formula (7);Wherein wi,jIndicate that i-th dimension is closed
Keyword is in tjThe number that moment occurs, TeAt the time of when expression event occurs, by calculating, it can finally obtain per one-dimensional keyword
Weight size, the text key word issued when occurring closer to event, weight are higher.
Again, in social space, the similitude of user u and event e are expressed as formula (8)
Wherein, S (u) and S (e) respectively refer to alternative family u and event e in the feature of social space (Social), i.e. formula (3)
(2) the participant information EA when friend information RF and event that the user in often interacts in the recent period occur, the measured value are got over
Greatly, it is bigger to illustrate that the good friend of designated user participates in number ratio shared by corresponding event, i.e., the user in social space's feature with
Corresponding event correlation is higher.
Finally, the degree of association of user and social event can be expressed as formula (9)
Correlation (u, e)=M (u, e)-1+C(u,e)+S(u,e) (9)
Wherein, M (u, e), C (u, e) and S (u, e) are respectively above-mentioned user and event in physics, information and social characteristic
The similitude in space;Since the value of M (u, e) is smaller, show that user is higher with event correlation, thus it is inverted to M (u, e), make
It obtains during constructing user and event correlation degree, final result Correlation (u, e) value is bigger, shows user and thing
The degree of association of part is higher.
Step 5: the degree of association based on user and event, in conjunction with the user data of known position information, training Gaussian process
Regression model, the distance of the unknown user of estimated position information to event;
By fusion user and social event in physics, information and the similitude of social space, i.e. merging formula (4), (5)
(6) calculated result, and the user data of known position information is combined to establish Gaussian process regression model (GPR, Gaussian
Process Regression), it is input with user and the degree of association Correlation (u, e) of event, with user to event
The distance of point is output, training Gaussian process regression model.And by model, the degree of association based on user and event estimates position
Confidence ceases unknown user at a distance from event.Model calculation formula such as (10)
Distance(uestimate, e) and=GPRTrained(Correlation(uestimate,e)) (10)
Wherein, GPRTrainedExpression is trained using the user and the corresponding customer incident degree of association of known position information
Gaussian process regression model, Correlation (uestimate, e) and indicate to need the user that estimates, the degree of association with event,
Distance(uestimate, e) indicate the user to be estimated to incident distance value size.
Claims (1)
1. it is a kind of estimation user to physical event distance method, it is characterised in that the following steps are included:
Step 1: screening user in social networks according to the subject key words of event, and extract number of users relevant to event
According to;
Assuming that the keyword of event is EW, the period of generation is ETP, then all use that keyword EW is referred in time ETP
Family will be screened as user relevant to event;For these users, its historical data, structure are obtained using web crawlers tool
The data model for building user's individual is expressed as formula (1)
RU=<L, C, F>(1)
Wherein, L indicates that the history of the user is registered data, and C indicates the status information issued in user's history, and F indicates that user exists
Friend information in social networks;
Step 2: the subject content of location information, participant's information when being occurred according to event and event, constructs the spy of event
Sign indicates model;
For event in physics, the feature of information and the aspect of social space three, geographical location information when binding events occur,
Participant's information and event topic construct the feature representation model of event, are expressed as formula (2)
EF=<CM, ET, EA>(2)
Wherein, CM indicates event in the feature of physical space, by extracting the historical movement path of all event participants, group
One-male unit feature is built for indicating the physical space attribute of event, is in the user's history of all participation events to event location
The probability-distribution function of distance;Assuming that the history of each participant is registered, sequence is PLS, and the position of current event is EL, pdiTable
Show any one PLSiTo the distance of EL, then calculated all pdiIndicate that some participant divides to the distance of event location
Cloth, then, the pd of all participantsiThe set of composition is exactly range distribution of the participant group to event location, as group
Moving characteristic CM is a kind of probability-distribution function about distance;
ET indicates that event in the feature of information space, is obtained by the topic and descriptor of extraction event;Every participant exists
Event generation period can all issue the state in relation to episode topic, by extracting the content of text of all participants' publications, and mention
Keyword is taken, keyword feature vector is constituted, each dimension indicates a kind of keyword, finally constitutes event in the spy of information space
Levy ET;
EA indicates event in the feature of social space, by the user information and their social networks of extracting participation event
It obtains;
Step 3: extracting user's individual in physics, information and social space based on user data relevant to event in step 1
Character representation model;
User's individual passes through the user data established in step 1 in the feature representation model of physics, information and social three spaces
Model export, is expressed as formula (3)
UF=<IM, HC, RF>(3)
Wherein, IM is measured by register place and current event distance location of the history in user data, and expression is that user arrives
The probability-distribution function of incident distance;Assuming that the history of user is registered, location sequence is LS, and the position of current event is EL, diTable
Show any LS in location sequenceiTo the distance of EL, then, all d being calculatediProbability distribution be then IM, i.e., it is a kind of
Probability-distribution function about distance;
HC is obtained by extracting history text of the user in social media, and expression is the text often referred in user's history
Information;By the extraction to history text information key, keyword feature vector is constructed, each dimension indicates a kind of keyword
Information can compare the topic keyword feature of event, calculate the similitude of the two;
RF is obtained by extracting in the every text message of user with the interactive information of good friend, and expression is that user often links up in the recent period
Social friend information;
Step 4: being directed to Step 2: the event and user's personal feature that construct in step 3 indicate model, definition user and event
In physics, the similitude of information and social three feature spaces, the degree of association based on this building user and event;
For Step 2: the event and user's personal feature model that step 3 constructs respectively, define the two feature representation models
In the similitude of physics, information and social space, and then measure the degree of association between user and event;For user u and thing
Part e, physical space similitude are expressed as formula (4)
Wherein, IM (d) indicates that the probability-distribution function in user's individual historical track with respect to event location distance, CM (d) indicate
Probability-distribution function of the event participant group relative to event location distance respectively corresponds formula (3), the IM in formula (2)
With CM;Measured value M (u, e) is smaller, shows that designated user is higher with corresponding event correlation in physical space feature;
Secondly, the two information space similitude is expressed as formula (5)
Wherein, C (u) and C (e) respectively refer to alternative family u and event e in the feature of information space, i.e., in formula (3) and formula (2)
User's history text information HC and event occur when theme ET, the formula calculate both cosine similarity, be worth it is bigger,
Show that designated user is higher with corresponding event correlation in information space feature;Wherein, due to the history text of user's publication
With chronological order, the period occurred closer to current event, content can more show the current text interest of user
Preference, therefore time factor is considered when constructing user keyword feature vector C (u), expression is such as formula (6) and public
Formula (7)
C (u)=< w1,w2,......,wn>(6)
Wherein, wiIt indicates certain one-dimensional keyword weight, is calculated by formula (7);Wherein, wi,jIndicate that i-th dimension is crucial
Word is in tjThe number that moment occurs, TeAt the time of when expression event occurs, by calculating, the power per one-dimensional keyword is finally obtained
Text key word that is great small, issuing when occurring closer to event, weight are higher;
Again, in social space, the similitude of user u and event e are expressed as formula (8)
Wherein, S (u) and S (e) respectively refer to alternative family u and event e in the feature of social space, i.e., in formula (3) and formula (2)
Participant information EA, measured value S (u, e) of the user when often interactive friend information RF and event occurs in the recent period more
Greatly, it is bigger to illustrate that the good friend of designated user participates in number ratio shared by corresponding event, i.e., the user in social space's feature with
Corresponding event correlation is higher;
Finally, the association table of user and social event is shown as formula (9)
Correlation (u, e)=M (u, e)-1+C(u,e)+S(u,e)(9)
Wherein, M (u, e), C (u, e) and S (u, e) are respectively above-mentioned user and event in physics, information and social characteristic space
Similitude;Since the value of M (u, e) is smaller, show that user is higher with event correlation, thus it is inverted to M (u, e), so that
During constructing user and event correlation degree, final result Correlation (u, e) value is bigger, shows user and event
The degree of association is higher;
Step 5: the degree of association based on user and event, in conjunction with the user data of known position information, training Gaussian process is returned
Model, the distance of the unknown user of estimated position information to event;
By fusion user and social event in physics, information and the similitude of social space, i.e. merging formula (4), formula (5)
With the calculated result of formula (6), and combine known position information user data establish Gaussian process regression model GPR, with
The degree of association Correlation (u, e) of family and event is input, is output, training Gauss with the distance of user to location of incident
Process regression model;And by model, the degree of association based on user and event, the unknown user of estimated position information and event
Distance;Model calculation formula such as (10)
Distance(uestimate, e) and=GPRTrained(Correlation(uestimate,e))(10)
Wherein, GPRTrainedIndicate the height trained using the user and the corresponding customer incident degree of association of known position information
This process regression model, Correlation (uestimate, e) and indicate to need the user that estimates, the degree of association with event,
Distance(uestimate, e) indicate the user to be estimated to incident distance value size.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610569794.2A CN106202488B (en) | 2016-07-19 | 2016-07-19 | Method of the estimation user to physical event distance |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610569794.2A CN106202488B (en) | 2016-07-19 | 2016-07-19 | Method of the estimation user to physical event distance |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106202488A CN106202488A (en) | 2016-12-07 |
CN106202488B true CN106202488B (en) | 2019-04-19 |
Family
ID=57494172
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610569794.2A Active CN106202488B (en) | 2016-07-19 | 2016-07-19 | Method of the estimation user to physical event distance |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106202488B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106709047B (en) * | 2017-01-04 | 2019-12-13 | 浙江宇视科技有限公司 | Object searching method and device |
CN107290797B (en) * | 2017-06-23 | 2019-01-18 | 西北工业大学 | A kind of obstacle detection system and method based on quorum-sensing system |
CN109600761B (en) * | 2017-09-30 | 2021-01-15 | 华为技术有限公司 | Method of wireless communication |
CN109974699B (en) * | 2017-12-27 | 2021-08-27 | 深圳市优必选科技有限公司 | Robot and map autonomous exploration method and device thereof |
CN108268762B (en) * | 2018-01-17 | 2021-04-30 | 同济大学 | Mobile social network user identity identification method based on behavior modeling |
CN112148979B (en) * | 2020-09-27 | 2023-08-01 | 北京百度网讯科技有限公司 | Event-associated user identification method, device, electronic equipment and storage medium |
CN113656686A (en) * | 2021-07-26 | 2021-11-16 | 深圳市中元产教融合科技有限公司 | Task report generation method based on birth teaching fusion and service system |
CN115455130B (en) * | 2022-11-10 | 2023-01-31 | 中国测绘科学研究院 | Fusion method of social media data and movement track data |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102664819A (en) * | 2012-03-27 | 2012-09-12 | 陈奕龙 | Event publishing network system based on social mechanism and geographical position |
CN104838673A (en) * | 2012-06-25 | 2015-08-12 | 谷歌公司 | Providing route recommendations |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2014089583A (en) * | 2012-10-30 | 2014-05-15 | International Business Maschines Corporation | Method, computer/program and computer for estimating location based on basis of social media |
US10242412B2 (en) * | 2012-11-20 | 2019-03-26 | Facebook, Inc. | Ambient-location-push notification |
-
2016
- 2016-07-19 CN CN201610569794.2A patent/CN106202488B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102664819A (en) * | 2012-03-27 | 2012-09-12 | 陈奕龙 | Event publishing network system based on social mechanism and geographical position |
CN104838673A (en) * | 2012-06-25 | 2015-08-12 | 谷歌公司 | Providing route recommendations |
Non-Patent Citations (2)
Title |
---|
Find me if you can: improving geographical prediction with social and spatial proximity;Lars Backstrom 等;《WWW "10 Proceedings of the 19th international conference on World wide web》;20100430;第61-70页 * |
Toward Estimating User-Social Event Distance:Mobility, Content, and Social Relationship;Fei Yi 等;《UbiComp "16 Proceedings of the 2016 ACM International Joint Conference on Pervasive and Ubiquitous Computing: Adjunct》;20160916;第233-236页 * |
Also Published As
Publication number | Publication date |
---|---|
CN106202488A (en) | 2016-12-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106202488B (en) | Method of the estimation user to physical event distance | |
Feng et al. | Deepmove: Predicting human mobility with attentional recurrent networks | |
Toole et al. | Tracking employment shocks using mobile phone data | |
Wu et al. | Location prediction on trajectory data: A review | |
Wang et al. | PGT: Measuring mobility relationship using personal, global and temporal factors | |
Schneider et al. | Unravelling daily human mobility motifs | |
Quercia et al. | Friendsensing: recommending friends using mobile phones | |
Chen et al. | Fine-grained urban event detection and characterization based on tensor cofactorization | |
CN106682770A (en) | Friend circle-based dynamic microblog forwarding behavior prediction system and method | |
Cheng et al. | A content-driven framework for geolocating microblog users | |
Cho et al. | Latent self-exciting point process model for spatial-temporal networks | |
CN110162717A (en) | A kind of method and apparatus of commending friends | |
Yu et al. | Identifying on-site users for social events: Mobility, content, and social relationship | |
CN102646122B (en) | Automatic building method of academic social network | |
Matamalas et al. | Assessing reliable human mobility patterns from higher order memory in mobile communications | |
Zhu et al. | FineRoute: Personalized and time-aware route recommendation based on check-ins | |
Zhang et al. | CRATS: An LDA-based model for jointly mining latent communities, regions, activities, topics, and sentiments from geosocial network data | |
Caverlee et al. | Towards Geo-Social Intelligence: Mining, Analyzing, and Leveraging Geospatial Footprints in Social Media. | |
Qiao et al. | Recommending nearby strangers instantly based on similar check-in behaviors | |
Zhang et al. | SEABIG: A deep learning-based method for location prediction in pedestrian semantic trajectories | |
Njoo et al. | Distinguishing friends from strangers in location-based social networks using co-location | |
Dario et al. | The point break effect: an examination of surf, crime, and transitory opportunities | |
Liu et al. | Dynamic metric embedding model for point-of-interest prediction | |
CN108513262A (en) | Location privacy protection method based on synthesis likelihood path | |
Miyazawa et al. | City-scale human mobility prediction model by integrating GNSS trajectories and SNS data using long short-term memory |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |