CN107657034A - A kind of event social networks proposed algorithm of social information enhancing - Google Patents
A kind of event social networks proposed algorithm of social information enhancing Download PDFInfo
- Publication number
- CN107657034A CN107657034A CN201710898707.2A CN201710898707A CN107657034A CN 107657034 A CN107657034 A CN 107657034A CN 201710898707 A CN201710898707 A CN 201710898707A CN 107657034 A CN107657034 A CN 107657034A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- event
- msup
- user
- 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
- 230000002708 enhancing effect Effects 0.000 title claims abstract description 30
- 230000006870 function Effects 0.000 claims description 52
- 239000013598 vector Substances 0.000 claims description 33
- 239000011159 matrix material Substances 0.000 claims description 13
- 238000000034 method Methods 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 4
- 230000003993 interaction Effects 0.000 claims description 4
- 238000005315 distribution function Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 238000003058 natural language processing Methods 0.000 claims description 3
- 230000008447 perception Effects 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- YBJHBAHKTGYVGT-ZKWXMUAHSA-N (+)-Biotin Chemical compound N1C(=O)N[C@@H]2[C@H](CCCCC(=O)O)SC[C@@H]21 YBJHBAHKTGYVGT-ZKWXMUAHSA-N 0.000 claims description 2
- 230000002123 temporal effect Effects 0.000 claims description 2
- FEPMHVLSLDOMQC-UHFFFAOYSA-N virginiamycin-S1 Natural products CC1OC(=O)C(C=2C=CC=CC=2)NC(=O)C2CC(=O)CCN2C(=O)C(CC=2C=CC=CC=2)N(C)C(=O)C2CCCN2C(=O)C(CC)NC(=O)C1NC(=O)C1=NC=CC=C1O FEPMHVLSLDOMQC-UHFFFAOYSA-N 0.000 claims description 2
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 244000046052 Phaseolus vulgaris Species 0.000 claims 1
- 235000010627 Phaseolus vulgaris Nutrition 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 8
- 230000010354 integration Effects 0.000 abstract 1
- 230000008859 change Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 239000000843 powder Substances 0.000 description 2
- 238000007418 data mining Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003973 paint Substances 0.000 description 1
- 238000010422 painting Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Primary Health Care (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
Abstract
The present invention relates to a kind of event social networks proposed algorithm of social information enhancing.This algorithm attempts the information by combining traditional social networks and event social networks, recommends problem to solve the cold start-up of event recommendation.First, this algorithm utilizes the social information of event sponsor, introduces event sponsor's assessment algorithm, strengthens the effect of proposed algorithm.Meanwhile this algorithm calculates the call guiding function of event participant, quantify influence of the participant to user, and as the factor of the event of recommendation.Then, the influence of the sponsor of this Algorithms Integration event and participant, and each side information such as the time of event, place, content, group, the selection preference of user is excavated.Searching order algorithm is finally combined, recommendation list is calculated for each user, realizes personalized recommendation.
Description
Technical field
The present invention relates to a kind of social networks proposed algorithm based on social information enhancing, belong to Data Mining.
Background technology
With the development of science and technology, social networks is also gradually ripe in recent years, and this gives and carries out data digging for social networks
The research of pick and analysis brings chance.And to be a kind of user can create event social networks thereon, publicize, share social activity
Social networks on the Novel wire of activity.Such as Meetup.com instantly all the fashion.On Meetup websites, user can create
A groups of users for having also interested hobby is built or added, interesting event is participated in and is interacted with group.Event society
Network is handed over to attract substantial amounts of user, so as to obtain fast development.But the social networks of more voluminous can produce it is very huge
Information flow, user is difficult the activity for therefrom quickly finding oneself to like, so as to waste the substantial amounts of time screen it is interested
Information on.In order to which user can more easily find activity interested, by computerized algorithm, automatically used for each
Activity interested is found at family, so as to solve the problems, such as information content overload.Such computer system, referred to as commending system.
Now, the proposed algorithm research based on social networks also turns into research and development focus both domestic and external, and algorithm research master
It is divided into four major classes：Event recommendation, group are recommended, label recommendations and entirety are recommended.Wherein, centered on event recommendation
Algorithm in, because event social networks does not have historic user scoring, pass through user to solve this problem, Minkov et al.
The similitude for participating in historical events is recommended, and Liu et al. then recommends new by using the interactions of user online and offline
Event, and Qiao et al. increases the social regularization factor on the basis of Liu, it is proposed that using Bayes's matrix disassembling method as
The model on basis is recommended to do, and Zhang et al. utilizes event geographical position, social factors information, it is proposed that linear decision model,
Macedo et al. is learnt using context, it is proposed that a proposed algorithm that can solve the problem that event social networks problem.The algorithm
Respectively the place of event, time, text presentation and social group attribute establish assessment models, and this comprehensive four models, structure
The proposed algorithm of an event ordering is made.
However, algorithm above all have ignored this information of event sponsor, and in current conditions, event sponsor with
Also played a role in the decision process at family.Meanwhile Bayes's Poisson decomposition model in Zhang et al. using set
In, event sponsor is introduced into assessment models as a stochastic variable, comprehensive other information, forms whole probability mould
Type, but such probabilistic model does not recommend foundation for the cold-start event of no historical record, so as to influence
The accuracy rate of event recommendation.And in event social networks, occasionally there are social networks between event sponsor and user, and
Sponsor promotes event using social account.The present invention is directed to this phenomenon, for cold-start event, utilizes its extra social network
The social information of network builds social assessment models as foundation is recommended, to event sponsor, it is intended to the social influence of analog subscriber
Power, and be incorporated among proposed algorithm.
It is considered herein that not solving cold start-up problem or not considering the probabilistic model Shortcomings of event sponsor, need to build
Found a kind of more perfect social assessment models.
The content of the invention
Present disclosure is to improve traditional commending system algorithm, there is provided a kind of social network based on social information enhancing
Network proposed algorithm.
1st, a kind of event social networks proposed algorithm of social information enhancing, it is characterised in that comprise the following steps：
Step 1, using the influence of event sponsor and event participant carry out event recommendation；Because event is held
Person usually promotes its event that will be held using its social account, therefore its social influence degree can have influence on other users and be
The no decision-making for participating in the event；
Step 2, the information with reference to traditional social networks and event social networks, quantify the influence power of event sponsor and carry
Go out sponsor's assessment models, specifically recommended using the similitude of group member；Because each event belongs to one
Individual group, and the user with group has bigger possibility once to participate in identical event, and then have bigger possibility again
Participate in a certain event simultaneously；Therefore it is similar to the event that current user to be recommended participated in using the member of group where event
Degree, to assess influence power of the group member to user's decision-making；Therefore group's recommender score is as follows：
Wherein GeThe group belonging to event e to be recommended is represented, u ' represents the other users in the group,Represent current to use
The event vector for the group that family is participated in, sim () is Jaccard coefficient of similarity；
Step 3, the recommended models of new perception of content are proposed to excavate the preference of user；Calculated using topic model algorithm
The correlation of event word content and the word content of event to be recommended that user participates in, so as to realize according to event content to enter
Row is recommended；
Step 4, propose the location of incident assessment models based on place popularity；In view of user in selection location of incident
When can take into account factors；When designing proposed algorithm, the preference for selecting place to user is modeled, and is recommended for user
Place sets rational event, is specifically recommended using the time cycle property of user；Event holds the time is to user
The no event that can participate in plays very big influence；And the time that generally user participates in event all has periodically；For example, weekend is empty
It is more between idle, therefore the possibility for participating in event is bigger；All event times that user was participated in form the time of user
Vector；The vector is the vector of 24*7 dimensions, represents 24 hours in 7 days and one day in one week respectively；
Time recommender score is calculated by calculating the time arrow of user and the cosine similarity of event time vector；Meter
Calculation method is as follows：
Step 5, established and commented according to temporal regularity, geographic constraint, event group and the content-preference of user's participation event
Estimate model, binding events sponsor's assessment models, the recommender score of event is calculated using searching order algorithm, obtains recommending row
Table.
In the event social networks proposed algorithm that a kind of above-mentioned social information strengthens, carried out using location of incident attribute
Recommend；Because event social networks is not only related to interaction on the line of user, can also be related to actually exchanging under the line of user；Cause
This user can also take the place of event and distance into account while event interested is selected；For example, some users
Footprint spread all over where city many corners, and some users then only haunted in the place near from company or family；Profit herein
The range distribution of event is participated in come analog subscriber with the method for Density Estimator, and is finally obtained in place side according to the distribution
Face, user select to participate in the probability of the event；Then, L is useduTo represent the ground point set of event that user u participated in；By to
The historical events at family point set carry out Density Estimator, obtain LuCorresponding distribution function fGApproximationIt is as follows：
Wherein l be a certain place l coordinate, KH() is gaussian kernel function, shaped like：
Wherein H=diag (h1, h2) it is 2 × 2 symmetrical matrixes tieed up, represent the wave peak width of kernel function；Then u pairs of user
Current event e place lePreference probability can useShow；Place recommender score is as follows：
In the event social networks proposed algorithm that a kind of above-mentioned social information strengthens, entered using the content introduction of event
Row is recommended；Content is the factor that can most show user interest that user selects event；Therefore text data is carried out first basic
Natural language processing, such as remove to stop word, stem reduction etc.；Then, treated data are established with classical word bag model
(bags of words), and TFIDF weightings are done to the content introduction vector of each event；Then because the interest of user passes through
It often can over time change and change, such as once was elapsed by interest over time and is changed by the user A interested that paints,
Like music now；Therefore, usage time attenuation function is weighted to all historical events vectors of user, forms user's
Content matrix is as follows：
WhereinExpression event e ' TFIDF vectors, δ are time attenuation parameters, and τ (e ') represents that user preengages event e ' and arrived
The now time interval；Matrix ultimately forms the content matrix of performance user interest hobby；Pass through calculatingMatrix is with currently treating
The cosine similarity of recommendation event e TFIDF vectors assesses effectiveness of the event content to user's decision-making；Commending contents fraction is such as
Under：
In the event social networks proposed algorithm that a kind of above-mentioned social information strengthens, propose that one kind utilizes event sponsor
Social information algorithm and recommended；It was found that in the event social network environment of reality, event sponsor and user it
Between occasionally there are social networks, and sponsor generally carries out the popularization of event using its social account；Therefore analysis attempts
The recommendation effectiveness of event sponsor is assessed in terms of these three from social influence power, historical events scoring, historical events content；Hair
Sponsor h social influence power factor S A (h) now is assessed using the concern number in the Twitter accounts of event sponsor,
It is as follows：
Wherein μ and σ does not pay close attention to the average and standard deviation of the log series model of person's number respectively；hdRepresent sponsor h powder
Silk number；With log (hd+ 1) come to avoid logarithm be zero；
The scoring for all events held according to event sponsor and scoring number, calculate event sponsor's historical events and comment
Molecular group HF (h) is as follows：
Wherein rijRepresent scoring of j-th of user to i-th of event, miRepresent the scoring sum of i-th of event, nhRepresent
The total number of events that sponsor h is held；
All event contents that event sponsor is held introduce word bag model (bags of words) expression, and will
It utilizes the theme feature of LDA models generation sponsor as sponsor's text set；Similarly, the text of current event is situated between
Continue as text set, calculate the theme feature of current event, and the similarity for calculating the two obtains historical events content factor HC
It is as follows：
HC=sim (Du, Dc)
Wherein DuRepresent the theme distribution that the LDA of user version collection is obtained, DcThe theme obtained for event text by LDA
Distribution, and sim () is then similarity function, is calculated herein using cosine similarity function；
After obtaining the social influence power factor, the historical events scoring factor and historical events content factor, linear combination is proposed
Adaptive enhancing personalized ordering algorithm come integrate this three aspect recommend effectiveness, calculate each user for a certain event
Recommender score in terms of event sponsor；Final sponsor's recommender score is represented by：
Sh(u, e)=λSASA(h)+λHFHF(h)+λHCHC(h)
Wherein λSA、λHF、λHCThree respective weights of the factor are represented respectively；The target of algorithm is exactly to solve λSA、λHF、λHC
So that obtained recommender score meets：The fraction of event of the event score participated in than not participating in is high；In order to solve weight
Parameter, propose adaptive enhancing personalized ordering algorithm；U and H is made to represent user's set and sponsor's set, (u, h) table respectively
Show a user-sponsor couple；Hu +Represent and user u had interactive sponsor's set, Hu -It is then its supplementary set；It is given one
Ranking functions π (), πuhRepresent sorting positions of the sponsor h under user u ranking functions；Utilize TG-AUC (AUC)
To represent the effectiveness of ranking functions, formula is as follows：
Wherein I () represents indicator function, i.e., is 1 when the conditions are met, otherwise be zero；The algorithm passes through with random suitable
Sequence, optimize each ranking functions successively, and weight coefficient is determined according to respective AUC；During study, for not
Bigger weight will be endowed to strengthen the study of next ranking functions by meeting the ranking results of condition；
For each user, ω is useduhRepresent the weight of each user-sponsor couple and be initialized as 1/ | Hu +|；Successively
All event orderings of selected and sorted function pair, and sequence accuracy rate is calculated to adjust weight；The sequence of social influence power is selected first
Function SA () is ranked up to all events, calculates its accuracy rate AUC (SA ()) that sorts；Adjusted according to sequence accuracy rate
The weight of section ranking functions and the weight of user-sponsor couple are as follows：
Wherein D represents user-sponsor to set；Similarly, weight is then counted according to the user-sponsor newly updated
Calculate the ranking results of historical feedback ranking functions and update weight；
Then the weight for updating historical content ranking functions is as follows：
Finally obtain the weight of each ranking functions, it is possible to calculate the recommender score S of final event sponsorh(u,
E) it is as follows：
Sh(u, e)=λSASA(h)+λHFHF(h)+λHCHC(h)。
In the event social networks proposed algorithm that a kind of above-mentioned social information strengthens, combined using coordinate ascent algorithm
The recommender score of each side；Target is to learn the weight of each different recommender scores, the final event that user was participated in
Recommender score is higher than the event that user does not participate in；
S (u, e)=wTS
Wherein S=[St(u, e), Sl(u, e), Sc(u, e), Sh(u, e), Sm(u, e)] represent calculate five recommender scores
The characteristic vector of composition, w=[wt, wl, wc, wh, wm] represent the weight vectors of corresponding recommender score；Utilize coordinate ascent algorithm meter
Calculate weight vectors w；Finally all events are ranked up using overall recommender score, and pushed away what event in the top was used as
Recommend list.
The invention has the advantages that：1st, different from existing assessment models, invention introduces event sponsor
Factor, define the dependent quantization the factor such as scoring of social influence power, history, historical events content etc..2nd, the present invention incorporates
The strong recommendation information of event social networks and propose a kind of model based on event sponsor and user's assessment models come for
The recommendation of event is modeled as social influence power.3rd, the present invention is proposed based on the location of incident assessment models of place popularity with whole
The distributed intelligence of user locations is closed, recommends place that rational event is set for user.4th, the present invention proposes a kind of based on social activity
The social networks proposed algorithm of information enhancement, more accurately calculate recommendation and assess fraction.5th, relative to existing proposed algorithm,
The present invention improves the degree of accuracy of event recommendation so that the information recommended for user more meets the preference of user.
Brief description of the drawings
Fig. 1 is the hierarchical structure of the social networks proposed algorithm based on social information enhancing.
The social influence power that Fig. 2 is event sponsor is analyzed.
Fig. 3 is the social networks based on social information enhancing in the case of different event number when mark ratio is 30%
Proposed algorithm is with the accuracy rate of other algorithms compared with recall rate.
Fig. 4 is that the social networks when marking ratio difference in the case that event number is equal based on social information enhancing pushes away
Algorithm is recommended with the accuracy rate of other algorithms compared with recall rate.
Fig. 5 is the accuracy rate and recall rate ratio of the social networks proposed algorithm based on social information enhancing and its subalgorithm
Compared with.
Embodiment
Social networks proposed algorithm based on social information enhancing comprises the following steps：
1) event recommendation is carried out using the influence of event sponsor and event participant.Due to event, sponsor is normal
Its event that will be held often is promoted using its social account, therefore its social influence degree can have influence on whether other users join
Add the decision-making of the event.
2) information of traditional social networks and event social networks is combined, quantifies influence power and the proposition one of event sponsor
Individual sponsor's assessment models.
3) recommended models of new perception of content are proposed to excavate the preference of user.User is calculated using topic model algorithm
The correlation of the event word content and the word content of event to be recommended of participation, so as to realize according to event content to be pushed away
Recommend.
4) the location of incident assessment models based on place popularity are proposed.In view of user when location of incident is selected
It can take into account factors, such as traffic conditions, meeting room facility etc..When designing proposed algorithm, place is selected to user
Preference is modeled, and recommends place to set rational event for user.
5) according to the assessment models result of above step, the recommender score of event is calculated using searching order algorithm, is obtained
Recommendation list.
A kind of event social networks proposed algorithm of described social information enhancing, it is characterised in that：Utilize group member
Similitude recommended.Because each event belongs to a group, and the user with group had bigger possibility once
Identical event has been participated in, and then has had bigger possibility again while participates in a certain event.Therefore group where event is utilized
The event similarity that member participated in current user to be recommended, to assess influence power of the group member to user's decision-making.Therefore
Group's recommender score is as follows：
Wherein GeThe group belonging to event e to be recommended is represented, u ' represents the other users in the group,Represent current to use
The event vector for the group that family is participated in, sim () is Jaccard coefficient of similarity.
A kind of event social networks proposed algorithm of described social information enhancing, it is characterised in that：Using user when
Between periodically recommended.Event holds whether the time can participate in the event to user and play very big influence.And generally use
The time that event is participated at family all has periodically.For example, free time at weekend is more, therefore the possibility for participating in event is bigger.
All event times that user was participated in form the time arrow of user.The vector is the vector of 24*7 dimensions, represents one respectively
24 hours in 7 days and one day in week.
Time recommender score is calculated by calculating the time arrow of user and the cosine similarity of event time vector.Meter
Calculation method is as follows：
A kind of event social networks proposed algorithm of described social information enhancing, it is characterised in that：Utilize location of incident
Attribute is recommended.Because event social networks is not only related to interaction on the line of user, can also be related under the line of user
Actual exchange.Therefore user can also take the place of event and distance into account while event interested is selected.Example
Such as, many corners in city where the footprint of some users spreads all over, and some users are then only in the place near from company or family
Haunt.Participate in the range distribution of event come analog subscriber using the method for Density Estimator herein, and finally asked according to the distribution
Go out in terms of place, user selects to participate in the probability of the event.Then, L is useduTo represent the place of event that user u participated in
Collection.Point set carries out Density Estimator by the historical events to user, obtains LuCorresponding distribution function fGApproximationSuch as
Under：
Wherein l be a certain place l coordinate, KH() is gaussian kernel function, shaped like：
Wherein H=diag (h1, h2) it is 2 × 2 symmetrical matrixes tieed up, represent the wave peak width of kernel function.Then u pairs of user
Current event e place lePreference probability can useShow.Place recommender score is as follows：
A kind of event social networks proposed algorithm of described social information enhancing, it is characterised in that：Using in event
Hold and introduce to be recommended.Content is the factor that can most show user interest that user selects event.Therefore first to textual data
According to basic natural language processing is carried out, such as go to stop word, stem reduction etc..Then, classics are established to treated data
Word bag model (bags of words), and TFIDF weightings are done to the content introduction vector of each event.Then due to user
Interest often change over time and change, such as interest hair is once elapsed over time to the user A interested of painting
Changing, music is liked now.Therefore, usage time attenuation function is weighted to all historical events vectors of user, shape
Content matrix into user is as follows：
WhereinExpression event e ' TFIDF vectors, δ are time attenuation parameters, and τ (e ') represents that user preengages event e ' and arrived
The now time interval.Matrix ultimately forms the content matrix of performance user interest hobby.Pass through calculatingMatrix is with currently treating
The cosine similarity of recommendation event e TFIDF vectors assesses effectiveness of the event content to user's decision-making.Commending contents fraction is such as
Under：
A kind of event social networks proposed algorithm of described social information enhancing, it is characterised in that：It is proposed a kind of utilize
The algorithm of the social information of event sponsor is simultaneously recommended.It was found that in the event social network environment of reality, event is held
Occasionally there are social networks between person and user, and sponsor generally carries out the popularization of event using its social account.Cause
This analysis attempts to assess event sponsor's in terms of these three from social influence power, historical events scoring, historical events content
Recommend effectiveness.Sponsor h social influence power SA is assessed using the concern number in the Twitter accounts of event sponsor
(h) it is, as follows：
Wherein μ and σ does not pay close attention to the average and standard deviation of the log series model of person's number respectively.hdRepresent sponsor h powder
Silk number.With log (hd+ 1) come to avoid logarithm be zero.
The scoring for all events held according to event sponsor and scoring number, calculate event sponsor's historical events and comment
Molecular group HF (h) is as follows：
Wherein rijRepresent scoring of j-th of user to i-th of event, miRepresent the scoring sum of i-th of event, nhRepresent
The total number of events that sponsor h is held.
All event contents that event sponsor is held introduce word bag model (bags of words) expression, and will
It utilizes the theme feature of LDA models generation sponsor as sponsor's text set.Similarly, the text of current event is situated between
Continue as text set, calculate the theme feature of current event, and the similarity for calculating the two obtains historical events content factor HC
It is as follows：
HC=sim (Du, Dc)
Wherein DuRepresent the theme distribution that the LDA of user version collection is obtained, DcThe theme obtained for event text by LDA
Distribution, and sim () is then similarity function, is calculated herein using cosine similarity function.
After obtaining the social influence power factor, the historical events scoring factor and historical events content factor, linear combination is proposed
Adaptive enhancing personalized ordering algorithm come integrate this three aspect recommend effectiveness, calculate each user for a certain event
Recommender score in terms of event sponsor.Final sponsor's recommender score is represented by：
Sh(u, e)=λSASA(h)+λHFHF(h)+λHCHC(h)
Wherein λSA、λHF、λHCThree respective weights of the factor are represented respectively.The target of algorithm is exactly to solve λSA、λHF、λHC
So that obtained recommender score meets：The fraction of event of the event score participated in than not participating in is high.In order to solve weight
Parameter, propose adaptive enhancing personalized ordering algorithm.U and H is made to represent user's set and sponsor's set, (u, h) table respectively
Show a user-sponsor couple.Hu +Represent and user u had interactive sponsor's set, Hu -It is then its supplementary set.It is given one
Ranking functions π (), πuhRepresent sorting positions of the sponsor h under user u ranking functions.Utilize TG-AUC (AUC)
To represent the effectiveness of ranking functions, formula is as follows：
Wherein I () represents indicator function, i.e., is 1 when the conditions are met, otherwise be zero.The algorithm passes through with random suitable
Sequence, optimize each ranking functions successively, and weight coefficient is determined according to respective AUC.During study, for not
Bigger weight will be endowed to strengthen the study of next ranking functions by meeting the ranking results of condition.
For each user, ω is useduhRepresent the weight of each user-sponsor couple and be initialized as 1/ | Hu +|.Successively
All event orderings of selected and sorted function pair, and sequence accuracy rate is calculated to adjust weight.The sequence of social influence power is selected first
Function SA () is ranked up to all events, calculates its accuracy rate AUC (SA ()) that sorts.Adjusted according to sequence accuracy rate
The weight of section ranking functions and the weight of user-sponsor couple are as follows：
Wherein D represents user-sponsor to set.Similarly, weight is then counted according to the user-sponsor newly updated
Calculate the ranking results of historical feedback ranking functions and update weight.
Then the weight for updating historical content ranking functions is as follows：
Finally obtain the weight of each ranking functions, it is possible to calculate the recommender score S of final event sponsorh(u,
E) it is as follows：
Sh(u, e)=λSASA(h)+λHFHF(h)+λHCHC(h)
A kind of event social networks proposed algorithm of described social information enhancing, it is characterised in that：Risen using coordinate
Algorithm combines the recommender score of each side.Target is to learn the weight of each different recommender scores, finally user is participated in
The recommender score for the event crossed is higher than the event that user does not participate in.
S (u, e)=wTS
Wherein S=[St(u, e), Sl(u, e), Sc(u, e), Sh(u, e), Sm(u, e)] represent calculate five recommender scores
The characteristic vector of composition, w=[wt, wl, wc, wh, wm] represent the weight vectors of corresponding recommender score.Utilize TG-AUC function
(AUC) object function is used as, weight vectors w is calculated using coordinate ascent algorithm.Finally utilize overall recommender score busy to institute
Part is ranked up, and using event in the top as recommendation list.
Embodiment 1：
Now by taking Meetup event social networks as an example, illustrate the specific embodiment of the present invention：
1st, data prediction
The data of Meetup websites are crawled using web crawlers, and are obtained by the Meetup user social contact account accounts provided
The Twitter concern numbers of user.In order to prevent influence of the noise to experimental result, eliminate and participate in the use that event number is less than 5
Family, the number of participant less than 5 event and hold event times be less than 20 group, finally give experimental data set.In order to comment
The recommendation effect of estimation algorithm, the event flag at random participating in certain customers is does not participate in, by itself and other cold-start events
Together as event to be recommended, finally recommended using proposed algorithm.In obtained recommendation list shared by flag event
Ratio is referred to as accuracy rate.
2nd, it is each user, calculates commending contents fraction respectively according to the step described in embodiment, sponsor recommends
Fraction, time recommender score, place recommender score and group's recommender score, finally obtain consequently recommended list.Calculating is based on society
Hand over the AUC of the social networks proposed algorithm of information enhancement.
3rd, in order to prove the validity of algorithm, realize traditional collaborative filtering, the personalized regression algorithm of enhancing and
The event ordering algorithm of multi-context study, contrasted with social information enhancing algorithm.Experimental result is as shown below.
Specific embodiment described herein is only to spirit explanation for example of the invention.Technology belonging to the present invention is led
The technical staff in domain can be made various modifications or supplement to described specific embodiment or be replaced using similar mode
Generation, but without departing from the spiritual of the present invention or surmount scope defined in appended claims.
Claims (5)
1. a kind of event social networks proposed algorithm of social information enhancing, it is characterised in that comprise the following steps：
Step 1, using the influence of event sponsor and event participant carry out event recommendation；Due to event, sponsor is normal
Its event that will be held often is promoted using its social account, therefore its social influence degree can have influence on whether other users join
Add the decision-making of the event；
Step 2, the information with reference to traditional social networks and event social networks, quantify influence power and the proposition one of event sponsor
Individual sponsor's assessment models, are specifically recommended using the similitude of group member；Because each event belongs to a group
Group, and the user with group has bigger possibility once to participate in identical event, and then have bigger possibility again simultaneously
Participate in a certain event；Therefore the event similarity participated in using the member of group where event with current user to be recommended, is come
Assess influence power of the group member to user's decision-making；Therefore group's recommender score is as follows：
<mrow>
<msub>
<mi>S</mi>
<mi>S</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>e</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msub>
<mi>G</mi>
<mi>e</mi>
</msub>
<mo>|</mo>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mrow>
<msup>
<mi>u</mi>
<mo>&prime;</mo>
</msup>
<mo>&Element;</mo>
<msub>
<mi>G</mi>
<mi>e</mi>
</msub>
</mrow>
</munder>
<mi>s</mi>
<mi>i</mi>
<mi>m</mi>
<mo>(</mo>
<mrow>
<mover>
<msub>
<mi>e</mi>
<mi>u</mi>
</msub>
<mo>&RightArrow;</mo>
</mover>
<mo>,</mo>
<mover>
<msub>
<mi>e</mi>
<msup>
<mi>u</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>&RightArrow;</mo>
</mover>
</mrow>
<mo>)</mo>
</mrow>
Wherein GeThe group belonging to event e to be recommended is represented, u ' represents the other users in the group,Represent active user's ginseng
The event vector of the group added, sim () are Jaccard coefficient of similarity；
Step 3, the recommended models of new perception of content are proposed to excavate the preference of user；User is calculated using topic model algorithm
The correlation of the event word content and the word content of event to be recommended of participation, so as to realize according to event content to be pushed away
Recommend；
Step 4, propose the location of incident assessment models based on place popularity；In view of user when location of incident is selected
It can take into account factors；When designing proposed algorithm, the preference for selecting place to user is modeled, and recommends place for user
Rational event is set, specifically recommended using the time cycle property of user；Event holds the time whether user can
Participate in the event and play very big influence；And the time that generally user participates in event all has periodically；For example, during the free time at weekend
Between it is more, therefore participate in event possibility it is bigger；All event times that user was participated in form the time arrow of user；
The vector is the vector of 24*7 dimensions, represents 24 hours in 7 days and one day in one week respectively；
<mrow>
<mover>
<msub>
<mi>u</mi>
<mi>t</mi>
</msub>
<mo>&RightArrow;</mo>
</mover>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msub>
<mi>E</mi>
<mi>u</mi>
</msub>
<mo>|</mo>
</mrow>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mrow>
<msup>
<mi>e</mi>
<mo>&prime;</mo>
</msup>
<mo>&Element;</mo>
<msub>
<mi>E</mi>
<mi>u</mi>
</msub>
</mrow>
</munder>
<mover>
<msup>
<mi>e</mi>
<mo>&prime;</mo>
</msup>
<mo>&RightArrow;</mo>
</mover>
</mrow>
Time recommender score is calculated by calculating the time arrow of user and the cosine similarity of event time vector；Calculating side
Method is as follows：
<mrow>
<msub>
<mi>S</mi>
<mi>T</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>e</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>c</mi>
<mi>o</mi>
<mi>s</mi>
<mrow>
<mo>(</mo>
<mover>
<mi>u</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>,</mo>
<mover>
<mi>e</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>)</mo>
</mrow>
</mrow>
Step 5, established according to temporal regularity, geographic constraint, event group and the content-preference of user's participation event and assess mould
Type, binding events sponsor's assessment models, the recommender score of event is calculated using searching order algorithm, obtains recommendation list.
A kind of 2. event social networks proposed algorithm of social information enhancing according to claim 1, it is characterised in that：Profit
Recommended with event site attribute；Because event social networks is not only related to interaction on the line of user, can also be related to
Actually exchanged under the line of user；Therefore user, also can be by the place of event and distance while event interested is selected
Take into account；For example, the footprint of some users spread all over where city many corners, and some users then only from company or
Family haunts near place；Participate in the range distribution of event come analog subscriber using the method for Density Estimator herein, and according to this
Distribution is finally obtained in terms of place, and user selects to participate in the probability of the event；Then, L is useduParticipated in represent user u
The ground point set of event；Point set carries out Density Estimator by the historical events to user, obtains LuCorresponding distribution function fG's
It is approximateIt is as follows：
<mrow>
<mover>
<msub>
<mi>f</mi>
<mi>G</mi>
</msub>
<mo>^</mo>
</mover>
<mrow>
<mo>(</mo>
<mi>l</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msub>
<mi>L</mi>
<mi>u</mi>
</msub>
<mo>|</mo>
</mrow>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mrow>
<msup>
<mi>l</mi>
<mo>&prime;</mo>
</msup>
<mo>&Element;</mo>
<msub>
<mi>L</mi>
<mi>u</mi>
</msub>
</mrow>
</munder>
<msub>
<mi>K</mi>
<mi>H</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>l</mi>
<mo>-</mo>
<msup>
<mi>l</mi>
<mo>&prime;</mo>
</msup>
<mo>)</mo>
</mrow>
</mrow>
Wherein l be a certain place l coordinate, KH() is gaussian kernel function, shaped like：
<mrow>
<msub>
<mi>K</mi>
<mi>H</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<msqrt>
<mrow>
<mn>2</mn>
<mi>&pi;</mi>
<mo>|</mo>
<mi>H</mi>
<mo>|</mo>
</mrow>
</msqrt>
</mfrac>
<msup>
<mo>&Element;</mo>
<mrow>
<mo>-</mo>
<mfrac>
<mrow>
<msup>
<mi>xx</mi>
<mi>T</mi>
</msup>
</mrow>
<mrow>
<mn>2</mn>
<msqrt>
<mi>H</mi>
</msqrt>
</mrow>
</mfrac>
</mrow>
</msup>
</mrow>
Wherein H=diag (h1, h2) it is 2 × 2 symmetrical matrixes tieed up, represent the wave peak width of kernel function；Then user u is to current
Event e place lePreference probability can useShow；Place recommender score is as follows：
<mrow>
<msub>
<mi>S</mi>
<mi>G</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>e</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mover>
<msub>
<mi>f</mi>
<mi>G</mi>
</msub>
<mo>^</mo>
</mover>
<mrow>
<mo>(</mo>
<msub>
<mi>l</mi>
<mi>e</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>.</mo>
</mrow>
A kind of 3. event social networks proposed algorithm of social information enhancing according to claim 1, it is characterised in that：Profit
Recommended with the content introduction of event；Content is the factor that can most show user interest that user selects event；Therefore it is first
Basic natural language processing is first carried out to text data, such as goes to stop word, stem reduction etc.；Then, to treated data
Classical word bag model (bags of words) is established, and TFIDF weightings are done to the content introduction vector of each event；So
Afterwards because the interest of user often changes over time and changes, such as user A once interested in drawing, over time
Passage interest changes, and likes music now；Therefore, usage time attenuation function enters to all historical events vectors of user
Row weighting, the content matrix for forming user are as follows：
<mrow>
<mover>
<mi>u</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<msup>
<mi>e</mi>
<mo>&prime;</mo>
</msup>
<mo>&Element;</mo>
<msub>
<mi>E</mi>
<mi>u</mi>
</msub>
</mrow>
</munder>
<mfrac>
<mn>1</mn>
<msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<mi>&delta;</mi>
<mo>)</mo>
</mrow>
<mrow>
<mi>&tau;</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>e</mi>
<mo>&prime;</mo>
</msup>
<mo>)</mo>
</mrow>
</mrow>
</msup>
</mfrac>
<mo>&times;</mo>
<mover>
<msup>
<mi>e</mi>
<mo>&prime;</mo>
</msup>
<mo>&RightArrow;</mo>
</mover>
</mrow>
WhereinExpression event e ' TFIDF vectors, δ are time attenuation parameters, and τ (e ') represents that user preengages event e ' till now
Time interval；Matrix ultimately forms the content matrix of performance user interest hobby；Pass through calculatingMatrix with it is current to be recommended
The cosine similarity of event e TFIDF vectors assesses effectiveness of the event content to user's decision-making；Commending contents fraction is as follows：
<mrow>
<msub>
<mi>S</mi>
<mi>C</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>e</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>c</mi>
<mi>o</mi>
<mi>s</mi>
<mrow>
<mo>(</mo>
<mover>
<mi>u</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>,</mo>
<mover>
<mi>e</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>)</mo>
</mrow>
<mo>.</mo>
</mrow>
A kind of 4. event social networks proposed algorithm of social information enhancing according to claim 1, it is characterised in that：Carry
Go out a kind of algorithm of social information using event sponsor and recommended；It was found that in the event social network environment of reality
In, occasionally there are social networks between event sponsor and user, and sponsor generally enters behaviour using its social account
The popularization of part；Therefore analysis attempts to assess thing in terms of these three from social influence power, historical events scoring, historical events content
The recommendation effectiveness of part sponsor；It was found that assess sponsor h's using the concern number in the Twitter accounts of event sponsor
Social influence power factor S A (h), it is as follows：
<mrow>
<mi>S</mi>
<mi>A</mi>
<mrow>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msubsup>
<mo>&Integral;</mo>
<mrow>
<mo>-</mo>
<mi>&infin;</mi>
</mrow>
<mrow>
<mi>l</mi>
<mi>o</mi>
<mi>g</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>h</mi>
<mi>d</mi>
</msub>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</msubsup>
<mfrac>
<mn>1</mn>
<mrow>
<msqrt>
<mrow>
<mn>2</mn>
<mi>&pi;</mi>
</mrow>
</msqrt>
<mi>&sigma;</mi>
</mrow>
</mfrac>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>-</mo>
<mi>&mu;</mi>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mrow>
<mn>2</mn>
<msup>
<mi>&sigma;</mi>
<mn>2</mn>
</msup>
</mrow>
</mfrac>
</mrow>
</msup>
<mi>d</mi>
<mi>x</mi>
</mrow>
Wherein μ and σ does not pay close attention to the average and standard deviation of the log series model of person's number respectively；hdRepresent sponsor h bean vermicelli people
Number；With log (hd+ 1) come to avoid logarithm be zero；
The scorings of all events held according to event sponsor and scoring number, calculate the scoring of event sponsor historical events because
Sub- HF (h) is as follows：
<mrow>
<mi>H</mi>
<mi>F</mi>
<mrow>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<msub>
<mi>n</mi>
<mi>h</mi>
</msub>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<msub>
<mi>m</mi>
<mi>i</mi>
</msub>
</munderover>
<msub>
<mi>r</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
</mrow>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<msub>
<mi>n</mi>
<mi>h</mi>
</msub>
</munderover>
<msub>
<mi>m</mi>
<mi>i</mi>
</msub>
</mrow>
</mfrac>
</mrow>
Wherein rijRepresent scoring of j-th of user to i-th of event, miRepresent the scoring sum of i-th of event, nhExpression is held
The total number of events that person h is held；
All event contents that event sponsor is held introduce word bag model (bags of words) expression, and are made
For sponsor's text set, the theme feature of LDA models generation sponsor is utilized；Similarly, the text presentation of current event is made
For text set, the theme feature of current event is calculated, and to obtain historical events content factor HC as follows for the similarity for calculating the two：
HC=sim (Du, Dc)
Wherein DuRepresent the theme distribution that the LDA of user version collection is obtained, DcThe theme distribution obtained for event text by LDA,
And sim () is then similarity function, calculated herein using cosine similarity function；
After obtaining the social influence power factor, the historical events scoring factor and historical events content factor, linear combination oneself is proposed
Adapt to enhancing personalized ordering algorithm and recommend effectiveness to integrate this three aspect, calculate event of each user for a certain event
Recommender score in terms of sponsor；Final sponsor's recommender score is represented by：
Sh(u, e)=λSASA(h)+λHFHF(h)+λHCHC(h)
Wherein λSA、λHF、λHCThree respective weights of the factor are represented respectively；The target of algorithm is exactly to solve λSA、λHF、λHCSo that
Obtained recommender score meets：The fraction of event of the event score participated in than not participating in is high；In order to solve weight parameter,
It is proposed adaptive enhancing personalized ordering algorithm；U and H is made to represent user's set and sponsor's set respectively, (u, h) represents one
User-sponsor couple；Hu +Represent and user u had interactive sponsor's set, Hu -It is then its supplementary set；Given sequence letter
Number π (), πuhRepresent sorting positions of the sponsor h under user u ranking functions；Represented using TG-AUC (AUC)
The effectiveness of ranking functions, formula are as follows：
<mrow>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>&pi;</mi>
<mo>(</mo>
<mo>&CenterDot;</mo>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<mi>U</mi>
<mo>|</mo>
</mrow>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>u</mi>
<mo>&Element;</mo>
<mi>U</mi>
</mrow>
</munder>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msup>
<msub>
<mi>H</mi>
<mi>u</mi>
</msub>
<mo>+</mo>
</msup>
<mo>|</mo>
</mrow>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>h</mi>
<mo>&Element;</mo>
<msup>
<msub>
<mi>H</mi>
<mi>u</mi>
</msub>
<mo>+</mo>
</msup>
</mrow>
</munder>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msup>
<msub>
<mi>H</mi>
<mi>u</mi>
</msub>
<mo>-</mo>
</msup>
<mo>|</mo>
</mrow>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>h</mi>
<mo>&Element;</mo>
<msup>
<msub>
<mi>H</mi>
<mi>u</mi>
</msub>
<mo>-</mo>
</msup>
</mrow>
</munder>
<mi>I</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>&pi;</mi>
<mrow>
<mi>u</mi>
<mi>h</mi>
</mrow>
</msub>
<mo><</mo>
<msub>
<mi>&pi;</mi>
<mrow>
<msup>
<mi>uh</mi>
<mo>&prime;</mo>
</msup>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
Wherein I () represents indicator function, i.e., is 1 when the conditions are met, otherwise be zero；The algorithm by a random order, according to
Each ranking functions of suboptimization, and weight coefficient is determined according to respective AUC；During study, for being unsatisfactory for
The ranking results of condition will be endowed bigger weight to strengthen the study of next ranking functions；
For each user, ω is useduhRepresent the weight of each user-sponsor couple and be initialized as 1/ | Hu +|；Select successively
Ranking functions calculate sequence accuracy rate to adjust weight to all event orderings；Social influence power ranking functions are selected first
SA () is ranked up to all events, calculates its accuracy rate AUC (SA ()) that sorts；According to sequence accuracy rate regulation row
The weight of order function and the weight of user-sponsor couple are as follows：
<mrow>
<msub>
<mi>&lambda;</mi>
<mrow>
<mi>S</mi>
<mi>A</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mi>l</mi>
<mi>n</mi>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>h</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</msub>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>u</mi>
<mi>h</mi>
</mrow>
</msub>
<mo>{</mo>
<mn>1</mn>
<mo>+</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>S</mi>
<mi>A</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>h</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</msub>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>u</mi>
<mi>h</mi>
</mrow>
</msub>
<mo>{</mo>
<mn>1</mn>
<mo>-</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>S</mi>
<mi>A</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
</mfrac>
</mrow>
<mrow>
<msup>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>u</mi>
<mi>h</mi>
</mrow>
</msub>
<mrow>
<mi>H</mi>
<mi>F</mi>
</mrow>
</msup>
<mo>=</mo>
<mfrac>
<mrow>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msup>
<msub>
<mi>H</mi>
<mi>u</mi>
</msub>
<mo>+</mo>
</msup>
<mo>|</mo>
</mrow>
</mfrac>
<mi>exp</mi>
<mo>{</mo>
<mo>-</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>S</mi>
<mi>A</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>h</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</msub>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msup>
<msub>
<mi>H</mi>
<mi>u</mi>
</msub>
<mo>+</mo>
</msup>
<mo>|</mo>
</mrow>
</mfrac>
<mi>exp</mi>
<mo>{</mo>
<mo>-</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>S</mi>
<mi>A</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
</mfrac>
</mrow>
Wherein D represents user-sponsor to set；Similarly, weight is then calculated and gone through according to the user-sponsor newly updated
The ranking results of history feedback function simultaneously update weight；
<mrow>
<msub>
<mi>&lambda;</mi>
<mrow>
<mi>H</mi>
<mi>F</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mi>l</mi>
<mi>n</mi>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>h</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</msub>
<msup>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>u</mi>
<mi>h</mi>
</mrow>
</msub>
<mrow>
<mi>H</mi>
<mi>F</mi>
</mrow>
</msup>
<mo>{</mo>
<mn>1</mn>
<mo>+</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>H</mi>
<mi>F</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>h</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</msub>
<msup>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>u</mi>
<mi>h</mi>
</mrow>
</msub>
<mrow>
<mi>H</mi>
<mi>F</mi>
</mrow>
</msup>
<mo>{</mo>
<mn>1</mn>
<mo>-</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>H</mi>
<mi>F</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
</mfrac>
</mrow>
<mrow>
<msup>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>u</mi>
<mi>h</mi>
</mrow>
</msub>
<mrow>
<mi>H</mi>
<mi>C</mi>
</mrow>
</msup>
<mo>=</mo>
<mfrac>
<mrow>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msup>
<msub>
<mi>H</mi>
<mi>u</mi>
</msub>
<mo>+</mo>
</msup>
<mo>|</mo>
</mrow>
</mfrac>
<mi>exp</mi>
<mo>{</mo>
<mo>-</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>H</mi>
<mi>F</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>h</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</msub>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msup>
<msub>
<mi>H</mi>
<mi>u</mi>
</msub>
<mo>+</mo>
</msup>
<mo>|</mo>
</mrow>
</mfrac>
<mi>exp</mi>
<mo>{</mo>
<mo>-</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>H</mi>
<mi>F</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
</mfrac>
</mrow>
Then the weight for updating historical content ranking functions is as follows：
<mrow>
<msub>
<mi>&lambda;</mi>
<mrow>
<mi>H</mi>
<mi>C</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mi>l</mi>
<mi>n</mi>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>h</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</msub>
<msup>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>u</mi>
<mi>h</mi>
</mrow>
</msub>
<mrow>
<mi>H</mi>
<mi>C</mi>
</mrow>
</msup>
<mo>{</mo>
<mn>1</mn>
<mo>+</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>H</mi>
<mi>C</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>h</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>D</mi>
</mrow>
</msub>
<msup>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>u</mi>
<mi>h</mi>
</mrow>
</msub>
<mrow>
<mi>H</mi>
<mi>C</mi>
</mrow>
</msup>
<mo>{</mo>
<mn>1</mn>
<mo>-</mo>
<mi>A</mi>
<mi>U</mi>
<mi>C</mi>
<mrow>
<mo>(</mo>
<mi>H</mi>
<mi>C</mi>
<mo>(</mo>
<mi>h</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>}</mo>
</mrow>
</mfrac>
</mrow>
Finally obtain the weight of each ranking functions, it is possible to calculate the recommender score S of final event sponsorh(u, e) such as
Under：
Sh(u, e)=λSASA(h)+λHFHF(h)+λHCHC(h)。
A kind of 5. event social networks proposed algorithm of social information enhancing according to claim 1, it is characterised in that：Profit
With coordinate ascent algorithm come the recommender score with reference to each side；Target is to learn the weight of each different recommender scores, is finally made
The recommender score for obtaining the event that user participated in is higher than the event that user does not participate in；
S (u, e)=wTS
Wherein S=[St(u, e), Sl(u, e), Sc(u, e), Sh(u, e), Sm(u, e)] represent that five recommender scores of calculating form
Characteristic vector, w=[wt, wl, wc, wh, wm] represent the weight vectors of corresponding recommender score；Calculated and weighed using coordinate ascent algorithm
The vectorial w of weight；Finally all events are ranked up using overall recommender score, and the recommendation that event in the top is used as arranges
Table.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710898707.2A CN107657034A (en) | 2017-09-28 | 2017-09-28 | A kind of event social networks proposed algorithm of social information enhancing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710898707.2A CN107657034A (en) | 2017-09-28 | 2017-09-28 | A kind of event social networks proposed algorithm of social information enhancing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107657034A true CN107657034A (en) | 2018-02-02 |
Family
ID=61116788
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710898707.2A Pending CN107657034A (en) | 2017-09-28 | 2017-09-28 | A kind of event social networks proposed algorithm of social information enhancing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107657034A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109614301A (en) * | 2018-11-19 | 2019-04-12 | 微梦创科网络科技（中国）有限公司 | A kind of appraisal procedure and device of information |
| CN109919793A (en) * | 2019-03-12 | 2019-06-21 | 中国科学技术大学 | Activity participates in analysis and recommended method |
| CN110704612A (en) * | 2019-08-09 | 2020-01-17 | 国家计算机网络与信息安全管理中心 | Social group discovery method and device and storage medium |
| CN111428127A (en) * | 2020-01-21 | 2020-07-17 | 江西财经大学 | Personalized event recommendation method and system integrating topic matching and two-way preference |
| CN111475724A (en) * | 2020-04-01 | 2020-07-31 | 上海硕恩网络科技股份有限公司 | Random walk social network event recommendation method based on user similarity |
| CN111611495A (en) * | 2020-04-01 | 2020-09-01 | 西安电子科技大学 | Network information reliability detection method, system, storage medium and terminal |
| CN112380452A (en) * | 2021-01-14 | 2021-02-19 | 北京崔玉涛儿童健康管理中心有限公司 | User interest collection method and device in infant content recommendation |
| CN113094593A (en) * | 2021-03-11 | 2021-07-09 | 西安交通大学 | Social network event recommendation method, system, device and storage medium |
| CN113704635A (en) * | 2021-07-30 | 2021-11-26 | 西安交通大学 | Social network event recommendation method and system |
| CN113706325A (en) * | 2021-07-30 | 2021-11-26 | 西安交通大学 | Planning method and system for event-oriented social network |
| CN114185258A (en) * | 2020-08-25 | 2022-03-15 | Oppo（重庆）智能科技有限公司 | Dial display method, smart watch, and nonvolatile computer-readable storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678672A (en) * | 2013-12-25 | 2014-03-26 | 北京中兴通软件科技股份有限公司 | Method for recommending information |
| CN106339483A (en) * | 2016-08-30 | 2017-01-18 | 电子科技大学 | Social activity recommendation algorithm in mobile social network |
-
2017
- 2017-09-28 CN CN201710898707.2A patent/CN107657034A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678672A (en) * | 2013-12-25 | 2014-03-26 | 北京中兴通软件科技股份有限公司 | Method for recommending information |
| CN106339483A (en) * | 2016-08-30 | 2017-01-18 | 电子科技大学 | Social activity recommendation algorithm in mobile social network |
Non-Patent Citations (2)
| Title |
|---|
| MACEDO等: ""Context-Aware Event Recommendation in Event-based Social Networks"", 《PROCEEDINGS OF THE 9TH ACM CONFERENCE ON RECOMMENDER SYSTEMS》 * |
| ZHIBO WANG等: ""Exploiting Social Influence for Context-Aware Event Recommendation in Event-based Social Networks"", 《IEEE INFOCOM 2017 - IEEE CONFERENCE ON COMPUTER COMMUNICATIONS》 * |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109614301A (en) * | 2018-11-19 | 2019-04-12 | 微梦创科网络科技（中国）有限公司 | A kind of appraisal procedure and device of information |
| CN109614301B (en) * | 2018-11-19 | 2024-01-26 | 微梦创科网络科技（中国）有限公司 | Information evaluation method and device |
| CN109919793A (en) * | 2019-03-12 | 2019-06-21 | 中国科学技术大学 | Activity participates in analysis and recommended method |
| CN110704612A (en) * | 2019-08-09 | 2020-01-17 | 国家计算机网络与信息安全管理中心 | Social group discovery method and device and storage medium |
| CN110704612B (en) * | 2019-08-09 | 2022-09-16 | 国家计算机网络与信息安全管理中心 | Social group discovery method and device and storage medium |
| CN111428127A (en) * | 2020-01-21 | 2020-07-17 | 江西财经大学 | Personalized event recommendation method and system integrating topic matching and two-way preference |
| CN111428127B (en) * | 2020-01-21 | 2023-08-11 | 江西财经大学 | Personalized event recommendation method and system integrating theme matching and bidirectional preference |
| CN111475724A (en) * | 2020-04-01 | 2020-07-31 | 上海硕恩网络科技股份有限公司 | Random walk social network event recommendation method based on user similarity |
| CN111611495A (en) * | 2020-04-01 | 2020-09-01 | 西安电子科技大学 | Network information reliability detection method, system, storage medium and terminal |
| CN114185258B (en) * | 2020-08-25 | 2023-10-17 | Oppo（重庆）智能科技有限公司 | Display method of dial plate, intelligent watch and nonvolatile computer readable storage medium |
| CN114185258A (en) * | 2020-08-25 | 2022-03-15 | Oppo（重庆）智能科技有限公司 | Dial display method, smart watch, and nonvolatile computer-readable storage medium |
| CN112380452A (en) * | 2021-01-14 | 2021-02-19 | 北京崔玉涛儿童健康管理中心有限公司 | User interest collection method and device in infant content recommendation |
| CN113094593B (en) * | 2021-03-11 | 2023-03-28 | 西安交通大学 | Social network event recommendation method, system, device and storage medium |
| CN113094593A (en) * | 2021-03-11 | 2021-07-09 | 西安交通大学 | Social network event recommendation method, system, device and storage medium |
| CN113706325A (en) * | 2021-07-30 | 2021-11-26 | 西安交通大学 | Planning method and system for event-oriented social network |
| CN113704635A (en) * | 2021-07-30 | 2021-11-26 | 西安交通大学 | Social network event recommendation method and system |
| CN113704635B (en) * | 2021-07-30 | 2024-05-24 | 西安交通大学 | Social network event recommendation method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107657034A (en) | A kind of event social networks proposed algorithm of social information enhancing | |
| Pham et al. | A general graph-based model for recommendation in event-based social networks | |
| CN103647671B (en) | A kind of intelligent perception network management and its system based on Gur Game | |
| CN105260390B (en) | A kind of item recommendation method based on joint probability matrix decomposition towards group | |
| CN103377250B (en) | Top k based on neighborhood recommend method | |
| CN110399565A (en) | Based on when null cycle attention mechanism recurrent neural network point of interest recommended method | |
| Wang et al. | Multi-objective optimization for location-based and preferences-aware recommendation | |
| CN109241440A (en) | It is a kind of based on deep learning towards implicit feedback recommended method | |
| CN107515909A (en) | A kind of video recommendation method and system | |
| CN109189944A (en) | Personalized recommending scenery spot method and system based on user's positive and negative feedback portrait coding | |
| CN103617289B (en) | Micro-blog recommendation method based on user characteristics and cyberrelationship | |
| CN101354714B (en) | Method for recommending problem based on probability latent semantic analysis | |
| CN105654388B (en) | A kind of modeling method of dynamic social network Information Propagation Model | |
| CN105117422A (en) | Intelligent social network recommender system | |
| CN102663047B (en) | Method and device for mining social relationship during mobile reading | |
| CN104794367B (en) | Medical treatment resource scoring based on hidden semantic model is with recommending method | |
| CN102999588A (en) | Method and system for recommending multimedia applications | |
| CN101986298A (en) | Information real-time recommendation method for online forum | |
| CN111428127B (en) | Personalized event recommendation method and system integrating theme matching and bidirectional preference | |
| CN110266745A (en) | Information flow recommended method, device, equipment and storage medium based on depth network | |
| CN107657043B (en) | Content-based mixed graph model image recommendation method | |
| CN103177382A (en) | Detection method for key propagation path and central node on microblog platform | |
| CN108053050A (en) | Clicking rate predictor method, device, computing device and storage medium | |
| CN106056413A (en) | Interest point recommendation method based on space-time preference | |
| CN106997383A (en) | The problem of based on PageRank algorithms, recommends method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180202 |
|
| RJ01 | Rejection of invention patent application after publication |