CN104954986A - Opportunity-type data transmission method based on multiple behavior sites - Google Patents

Abstract

The invention provides an opportunity-type data transmission method based on multiple behavior sites, aiming to solve the problem of incapability of accurately sending data to attribute users of the behavior sites. According to the method, the user performs site behavior registration on a cloud server, the cloud server stores a user/behavior site combined matrix, and a server updates the matrix timely and returns a newest time slice vector to the user. A source user starts a data forwarding process after applying data from the cloud server. In the process of data forwarding, a circuit is initiated and messages are transmitted to the attribute users of h different behavior sites, wherein the h refers to the number of the sites. In the progressive gradient increase stage of forwarding, the messages are forwarded to intention receivers of the behavior sites layer by layer. The data are sent to the intention receivers of all the surrounding behavior sites in a multicast mode. The data are forwarded to the users of the behavior sites by taking user behavior attribute weights as forwarding conditions, so that data transmission rate is improved effectively and energy consumption is lowered.

Description

A kind of is the opportunistic data transmission method in place based on multirow

Technical field

The present invention proposes a kind of is the opportunistic data transmission method in place based on multirow, belongs to the crossing domain of opportunistic network and social networks.

Background technology

At present, traditional sensor network comes with some shortcomings: (1) needs manually to dispose a large amount of awareness apparatus in advance, and not only expensive, along with the expansion of scale, the complexity of system also sharply increases.(2) awareness apparatus static state is disposed, generally can not dense deployment, makes perception visual angle single, cannot comprehensive thorough perceptive object.(3) carry out forwarding data according to node ID, be not suitable for the network scenarios of height dynamic mobile, also cannot be applied to the network that there is special interests preference.

Along with the development of technology, emerge in multitude with the intelligent movable equipment enriching transducer, intelligent perception, as cognition technology of future generation, has become study hotspot in recent years.Smart mobile phone, as the major product of mobile device, can reflect the behavior property of holder truly, for intelligent perception technology brings opportunities and challenges.The high performance of mobile equipments spread all over the world has erected huge community network, mobile subscriber can obtain in community network, sharing data, makes intelligent perception technology create good using value by the collection to data, analysis for user's customized privatization service.

Intelligent perception is by carrying sharing individual by mobile Internet or obtaining the social perception task that data are carried out consciously, unconsciously cooperation has carried out complexity of awareness apparatus.Intelligent perception technology carries out perception activity by a large amount of universal mobile terminal device, overcomes the problem of awareness apparatus costliness, utilizes the mobile wisdom of people to carry out perception data in all directions.By analyzing the social of user, setting up the data-collecting mode being characterized as guide with user behavior, improve user's participation and Data Collection performance.

Data Collection is as a large key technology of intelligent perception, and the opportunistic Data Collection based on social action analysis becomes the main way of intelligent perception Data Collection.But opportunistic data collection techniques still faces some technical barriers at present: its calculating of the embedded device that (1) mobile subscriber adopts, storage capacity and battery power exist limitation.(2) how the social primary feature being intelligent perception and being different from other data perception patterns, extract the social property of people and to be quantified as computable unit be problem demanding prompt solution.(3) under general network condition, sensing node is the interaction data information when meeting only, and exchanges data chance is very valuable, and reduction time delay, the efficiency of transmission that improves become another challenge in opportunistic Data Collection.

Summary of the invention

The object of the invention is to solve how data to be precisely sent to by the mode that opportunistic forwards in intelligent perception application and there is the problem that multirow is properties user group.The present invention is by portraying BM25 model according to user property, calculate the similarity weights of user about specific behavior place, and send data to according to these weights that to have multirow be properties user, optimize the quantity of message copy in data transmission procedure, the data collection algorithm similar relative to other, the method has low-cost, high transmission rates characteristic.

This invention only considers the problem of single behavior property for the Data Transport Protocol in existing intelligent perception application, utilize the stability of user behavior, adopt server/user model, by server extraction user behavior attribute, and portray user behavior attribute by BM25 model, user/behavior place confederate matrix is stored in the middle of server formation, and the behavioural information immediate updating confederate matrix that server is uploaded according to user, guarantees the real-time of data, accuracy.When source user needs information to be sent to targeted customer, source user need ask for global information to server, and sends data by source user opportunistic, and user data privacy is guaranteed.Adopt ladder to increase progressively the mode combined with multicast in data transmission phase, effectively reduce message copy quantity, reduce and send energy consumption.According to relatively carrying out data retransmission about the correlation weights in behavior place between user, improve data transmission success.

Technical solution of the present invention is:

The invention provides a kind of is the opportunistic data transmission method in place based on multirow.Build Cloud Server, portray user property formation user/behavior place confederate matrix by BM25 and be stored in server.Source user to server request for data, and carries out opportunistic data retransmission according to user about the similarity weights of appointed place.

Explanation of nouns:

Timeslice ts _ij: the time that user i consumes at behavior place j;

User's mean activity timeslice number uts _i: user's sheet quantity of average activity time every day;

All user's mean activity timeslice number avts: all users sheet quantity of average activity time every day;

Duration regulates attribute E: for balancing Active duration between all users and unique user;

Behavior property BeA _ij: user i is about the correlation of behavior place j;

Place balanced nature BA _j: for balancing more with the less activity venue factor of flow of the people, the distribution situation of reflection people's behavior place in all users;

Behavior place weight w eight _ij: user i is to behavior place j correlation weights;

Behavior place threshold values th _j: j border, behavior place weights, user i behavior place weight w eight _ij>=th _jtime, think that user i is the Interests User of behavior place j.

Concrete steps:

Step 1: set number of users as n, place quantity is m, user i ∈ 1 ..., n} behavior place j ∈ 1 ..., the time that m} consumes is with timeslice quantity (time slice) ts _ijtolerance.The length (length of time slice) of timeslice is fixed as tl, if user i is T in the total time that behavior place j consumes _ij, then ts _ij=T _ij/ tl.Ts _ijreflection behavior place j is to the correlation of user i, and timeslice quantity is larger, and user is more relevant to this place.Each user i regularly submits the time T consumed about place j to server _ij, be called that behavior place is registered;

Step 2: the mobile trajectory data that server is submitted to according to user i, counting user i is to the timeslice vector T S in certain behavior place _i=(ts _i1, ts _i2... ts _im), and result of calculation is stored in server end with user/behavior place confederate matrix form.Server upgrades the information that in user/behavior place confederate matrix, this user is corresponding at every turn immediately when receiving the track data of user i, and will add up gained timeslice vector T S _isend to user.Wherein user/behavior place confederate matrix form is as shown in table 1:

Table 1: user/behavior place confederate matrix

	Behavior place 1	Behavior place 2	Behavior place 3	Behavior place 4
					User 1	Timeslice ts 11	Timeslice ts 12	Timeslice ts 13	Timeslice ts 14
User 2	Timeslice ts 21	Timeslice ts 22	Timeslice ts 23	Timeslice ts 24
					User 3	Timeslice ts 31	Timeslice ts 32	Timeslice ts 33	Timeslice ts 34
User 4	Timeslice ts 41	Timeslice ts 42	Timeslice ts 43	Timeslice ts 44

Step 3: source user s specifies TBP=((p ₁, th ₁), (p ₂, th ₂) ..., (p _h, th _h)), h≤m, wherein p _jand th _j, j ∈ 1 ..., h} is respectively behavior place and corresponding behavior place threshold values; S sends behavior place vector P to server;

Step 4: server calculates all user's mean activity timeslice number avts, and sends to source user s.Wherein the computational methods of user's mean activity timeslice number avts are:

$\mathrm{avtts}=\left(\underset{i=1}{\overset{i=n}{\mathrm{\Σ}}}\underset{j=1}{\overset{j=m}{\mathrm{\Σ}}}{\mathrm{ts}}_{\mathrm{ij}}\right)/n---\left(1\right)$

Step 5: server calculates place balanced nature vector BA (the place balanced nature BA that wherein BA vector all behavior places of specifying for source user are corresponding according to user/behavior place confederate matrix _jset), and be sent to source user s.Wherein, place balanced nature BA _jcomputational methods be:

${\mathrm{BA}}_j=\log \frac{n-{\mathrm{pv}}_j+0.5}{{\mathrm{pv}}_j+0.5}---\left(2\right)$

Pv _jfor the total number of persons (flow of the people) of average access every day of behavior place j, namely in confederate matrix, jth row are not the number of 0.BA _jas global factor, it is considered that the relative importance of behavior place j.Known according to formula (2), given area total number of persons, more people arrived this place, and so the value of balance factor will be less.What balance factor reflected is the distribution situation of interested site in all personnel, balances the final behavior weights in the less but important behavior place of flow of the people with this.

Step 6: source user s, by user's mean activity timeslice number avts and place balanced nature vector BA, sends to each user r that meets, calculate separately about place j ∈ 1 ..., the weight w eight of h} _sjand weight _rj; Behavior place weight calculation method is as follows:

Calculate Active duration and regulate attribute E _i:

$E_i=k_1((1-b)+b\·\frac{{\mathrm{uts}}_i}{\mathrm{avts}})---\left(3\right)$

Wherein uts _irepresent that user i consumes activity time sheet sum every day, duration regulates attribute representative user i to the consideration of timeslice sum.K ₁, b is empirical parameter.B is for regulating user i timeslice ratio, and span is [0,1].When b value is 0 in representation formula not service time sheet ratio, will complete timeslice ratio be used when b value is 1

Calculating behavior factor B A:

${\mathrm{BeA}}_{\mathrm{ij}}=\frac{(k_1+1)\·{\mathrm{ts}}_{\mathrm{ij}}}{E_i+{\mathrm{ts}}_{\mathrm{ij}}}---\left(4\right)$

BeA _ijrepresent that user i is to the degree of correlation of behavior place j, its value is higher, illustrates that user i is more relevant to place j.

User i for the computational methods of the weights of behavior place j is:

weight _ij＝BeA _ij·BA _j(5)

Step: 7: enter the initiating line stage, if weight _rj>=weight _sjif this user is r _j, then source user by data retransmission to r _j, otherwise do not forward;

Step 8: repeat step: 6-step 7, until all h behavior place all searches out all corresponding all r _j;

Step 9: source user s deletes the data that self preserves, now source user is initialized h bar and increases progressively route;

Step: 10: the user r on every bar circuit _jinquire the weights Weight of user q about behavior place j that meet _qj;

Step 11: enter gradient incremental stages, if weight _qj>=weight _rjj, then r _jforward the data to q, r _jdelete data; Otherwise do not forward;

Step: 12: repeat step 10-step 11, until weight _qj>=th _j;

Step: 13: enter the multicast stage, in data lifetime, meets weight for any user q _qj>=th _jif meet any user t and meet then user q copies data to user t, terminates.

Beneficial effect

The invention solves in intelligent perception application is place opportunistic data transmission problems about multirow.Through knowing at ONE platform emulation experiment show, be that place opportunistic transmission plan has following significant advantage relative to traditional single file:

1. by portraying many behavior properties of user, using the data retransmission foundation that weights are place as multirow, can effectively transfer data in the customer group with many behavior properties, improve the accuracy rate of transfer rate and data transmission.

2. have employed the data transfer mode that gradient increases progressively and multicast combines in data-gathering process, effectively reduce data trnascription quantity in repeating process, reduce transmission energy consumption.

Accompanying drawing explanation

Fig. 1 is the opportunistic data transmission method flow chart in place based on multirow.

Fig. 2 user/cloud platform interaction models.

Fig. 3 user/behavior place confederate matrix.

Fig. 4 multirow is locality data transmission method schematic diagram.

Embodiment

Embodiments of the present invention are introduced below in conjunction with accompanying drawing.

The invention provides a kind of is the opportunistic data transmission method in place based on multirow, the many behavior properties of the method to mobile subscriber are portrayed, use the data retransmission foundation that weights are place as multirow, can effectively transfer data in the customer group with many behavior properties.The concrete implementing procedure of the method as shown in Figure 1.The method needs Cloud Server support, calculates user data, and carry out real time data interworking in order to store.User/cloud platform interaction models as shown in Figure 2.

Concrete implementation step is:

Step 1: set number of users as n, place quantity is m, user i ∈ 1 ..., n} behavior place j ∈ 1 ..., the time that m} consumes is with timeslice quantity (time slice) ts _ijtolerance.The length (length of time slice) of timeslice is fixed as tl, if user i is T in the total time that behavior place j consumes _ij, then ts _ij=T _ij/ tl.Ts _ijreflection behavior place j is to the correlation of user i, and timeslice quantity is larger, and user is more relevant to this place.Each user i regularly submits the time T consumed about place j to server _ij, be called that behavior place is registered;

Step 2: the mobile trajectory data that server is submitted to according to user i, counting user i is to the timeslice vector T S in certain behavior place _i=(ts _i1, ts _i2... ts _im), and result of calculation is stored in server end with user/behavior place confederate matrix form.User/behavior place confederate matrix as shown in Figure 3.Server upgrades the information that in user/behavior place confederate matrix, this user is corresponding at every turn immediately when receiving the track data of user i, and will add up gained timeslice vector T S _isend to user;

Step 3: source user s specifies TBP=((p ₁, th ₁), (p ₂, th ₂) ..., (p _h, th _h)), h≤m, wherein p _jand th _j, j ∈ 1 ..., h} is respectively behavior place and corresponding behavior place threshold values; S sends behavior place vector P to server;

Step 4: server calculates all user's mean activity timeslice number avts, and sends to source user s.Wherein the computational methods of user's mean activity timeslice number avts are:

$\mathrm{avtts}=\left(\underset{i=1}{\overset{i=n}{\mathrm{\Σ}}}\underset{j=1}{\overset{j=m}{\mathrm{\Σ}}}{\mathrm{ts}}_{\mathrm{ij}}\right)/n---\left(1\right)$

Step 5: server calculates place balanced nature vector BA (the place balanced nature BA that wherein BA vector all behavior places of specifying for source user are corresponding according to user/behavior place confederate matrix _jset), and be sent to source user s.Wherein, place balanced nature BA _jcomputational methods be:

${\mathrm{BA}}_j=\log \frac{n-{\mathrm{pv}}_j+0.5}{{\mathrm{pv}}_j+0.5}---\left(2\right)$

Pv _jfor the total number of persons (flow of the people) of average access every day of behavior place j, namely in confederate matrix, jth row are not the number of 0.BA _jas global factor, it is considered that the relative importance of behavior place j.Known according to formula (2), given area total number of persons, more people arrived this place, and so the value of balance factor will be less.What balance factor reflected is the distribution situation of interested site in all personnel, balances the final behavior weights in the less but important behavior place of flow of the people with this.

Step 6 below-step 12 is data transfer phase, and process as shown in Figure 4.

Step 6: source user s, by user's mean activity timeslice number avts and place balanced nature vector BA, sends to each user r that meets, calculate separately about place j ∈ 1 ..., the weight w eight of h} _sjand weight _rj; Behavior place weight calculation method is as follows:

Calculate Active duration and regulate attribute E _i:

$E_i=k_1((1-b)+b\·\frac{{\mathrm{uts}}_i}{\mathrm{avts}})---\left(3\right)$

Wherein uts _irepresent that user i consumes activity time sheet sum every day, duration regulates attribute representative user i to the consideration of timeslice sum.K ₁, b is empirical parameter.B is for regulating user i timeslice ratio, and span is [0,1].When b value is 0 in representation formula not service time sheet ratio, will complete timeslice ratio be used when b value is 1

Calculating behavior factor B A:

${\mathrm{BeA}}_{\mathrm{ij}}=\frac{(k_1+1)\·{\mathrm{ts}}_{\mathrm{ij}}}{E_i+{\mathrm{ts}}_{\mathrm{ij}}}---\left(4\right)$

BeA _ijrepresent that user i is to the degree of correlation of behavior place j, its value is higher, illustrates that user i is more relevant to place j.

User i for the computational methods of the weights of behavior place j is:

weight _ij＝BeA _ij·BA _j(5)

Step: 7: enter the initiating line stage, if weight _rj>=weight _sjif this user is r _j, then source user by data retransmission to r _j, otherwise do not forward;

Step 8: repeat step: 6-step 7, until all h behavior place all searches out all corresponding all r _j;

Step 9: source user s deletes the data that self preserves, now source user is initialized h bar and increases progressively route;

Step: 10: the user r on every bar circuit _jinquire the weights Weight of user q about behavior place j that meet _qj;

Step 11: enter gradient incremental stages, if weight _qj>=weight _rjj, then r _jforward the data to q, r _jdelete data; Otherwise do not forward;

Step: 12: repeat step 10-step 11, until weight _qj>=th _j;

Step: 13: enter the multicast stage, in data lifetime, meets weight for any user q _qj>=th _jif meet any user t and meet then user q copies data to user t, terminates.

Claims (1)

1. be the opportunistic data transmission method in place based on multirow, it is characterized in that, concrete steps are as follows:

Step 1: set number of users as n, place quantity is m, user i ∈ 1 ..., n} behavior place j ∈ 1 ..., the time that m} consumes is with timeslice quantity (time slice) ts _ijtolerance, the length of timeslice is fixed as tl, if user i is T in the total time that behavior place j consumes _ij, then ts _ij=T _ij/ tl;

Step 2: the mobile trajectory data that server is submitted to according to user i, counting user i is to the timeslice vector T S in certain behavior place _i=(ts _i1, ts _i2... ts _im), and result of calculation is stored in server with user/behavior place confederate matrix form; Server upgrades the information that in user/behavior place confederate matrix, this user is corresponding at every turn immediately when receiving the track data of user i, and will add up gained timeslice vector T S _isend to user;

Step 3: source user s specifies TBP=((p ₁, th ₁), (p ₂, th ₂) ..., (p _h, th _h)), h≤m, wherein p _jand th _j, j ∈ 1 ..., h} is respectively behavior place and corresponding behavior place threshold values; S sends behavior place vector P to server;

Step 4: server calculates all user's mean activity timeslice number avts, and sends to source user s, and wherein the computational methods of user's mean activity timeslice number avts are:

$\mathrm{avtts}=\left(\underset{i=1}{\overset{i=n}{\mathrm{\Σ}}}\underset{j=1}{\overset{j=m}{\mathrm{\Σ}}}{\mathrm{ts}}_{\mathrm{ij}}\right)/n---\left(1\right)$

Step 5: server calculates place balanced nature vector BA according to user/behavior place confederate matrix, and is sent to source user s, wherein, place balanced nature BA _jcomputational methods be:

${\mathrm{BA}}_j=\log \frac{n-{\mathrm{pv}}_j+0.5}{{\mathrm{pv}}_j+0.5}---\left(2\right)$

Pv _jfor the total number of persons of average access every day of behavior place j, namely in confederate matrix, jth row are not the number of 0; BA _jas global factor, it is considered that the relative importance of behavior place j;

Step 6: source user s, by user's mean activity timeslice number avts and place balanced nature vector BA, sends to each user r that meets, calculate separately about place j ∈ 1 ..., the weight w eight of h} _sjand weight _rj; Behavior place weight calculation method is as follows:

Calculate Active duration and regulate attribute E _i:

$E_i=k_1((1-b)+b\·\frac{{\mathrm{uts}}_i}{\mathrm{avts}})---\left(3\right)$

Wherein uts _irepresent that user i consumes activity time sheet sum every day, duration regulates attribute representative user i to the consideration of timeslice sum.K ₁, b is empirical parameter; B is for regulating user i timeslice ratio, and span is [0,1]; When b value is 0 in representation formula not service time sheet ratio, will complete timeslice ratio be used when b value is 1

Calculating behavior factor B A:

${\mathrm{BeA}}_{\mathrm{ij}}=\frac{(k_1+1)\·{\mathrm{ts}}_{\mathrm{ij}}}{E_i+{\mathrm{ts}}_{\mathrm{ij}}}---\left(4\right)$

BeA _ijrepresent that user i is to the degree of correlation of behavior place j, its value is higher, illustrates that user i is more relevant to place j;

User i for the computational methods of the weights of behavior place j is:

weight _ij＝BeA _ij·BA _j(5)

Step: 7: enter the initiating line stage, if weight _rj>=weight _sjif this user is r _j, then source user by data retransmission to r _j, otherwise do not forward;

Step 8: repeat step: 6 to step 7, until all h behavior place all searches out all corresponding all r _j;

Step 9: source user s deletes the data that self preserves, now source user is initialized h bar and increases progressively route;

Step: 10: the user r on every bar circuit _jinquire the weights Weight of user q about behavior place j that meet _qj;

Step 11: enter gradient incremental stages, if weight _qj>=weight _rjj, then r _jforward the data to q, r _jdelete data; Otherwise do not forward;

Step: 12: repeat step 10-step 11, until weight _qj>=th _j;

Step: 13: enter the multicast stage, in data lifetime, meets weight for any user q _qj>=th _jif meet any user t and meet then user q copies data to user t, terminates.