CN107609836A - A kind of extensive mass-rent task method of diffusion based on linear probability - Google Patents

Abstract

The invention discloses a large-scale crowdsourcing task diffusion method based on linear probability. In this method, a reverse auction process is embodied between the crowdsourcing platform and registered users. Registered users submit a bidding document containing quotations to the crowdsourcing platform, and the crowdsourcing platform predicts its influence for each registered user, selects a batch of candidates from the registered users according to the prediction results, and calculates the remuneration of each registered user. The large-scale crowdsourcing task diffusion method proposed in the present invention has reward incentives for selected candidates, and satisfies computational effectiveness, personal rationality and anti-deception. This method can be widely used in large-scale crowdsourcing systems.

Description

A Diffusion Method for Large-Scale Crowdsourcing Tasks Based on Linear Probability

technical field

The invention belongs to the intersection field of the Internet and algorithmic game theory, and in particular relates to a large-scale crowdsourcing task diffusion method based on linear probability.

Background technique

Crowdsourcing refers to the practice that a company or institution outsources work tasks performed by employees in the past to a non-specific public network in a free and voluntary manner. The advantage of crowdsourcing is to use the wisdom of individuals, teams, and communities to complete tasks. In recent years, crowdsourcing has been widely used in many fields, such as video analysis, smart city, machine learning, information retrieval, social network and other fields. With the popularity of smartphones, mobile crowdsourcing has gradually become an effective method to meet the needs of a wide range of sensing tasks.

However, these applications currently assume that there are enough participants to perform crowdsourcing tasks, which is often unrealistic, especially for some crowdsourcing platforms that have large-scale crowdsourcing task completion requirements. Therefore, the most effective solution is to select participants with greater influence in the social network to spread the crowdsourcing task, so that more people can participate in it. In order to select influential participants, it is very important to evaluate the influence of participants in the social network.

Because the selected diffusers need to consume equipment energy, computing power, storage space, data traffic, etc. to diffuse crowdsourcing tasks, participants need to get a certain amount of incentives to offset these losses. The successful implementation of the diffusion of crowdsourcing tasks depends on the number of diffusers and the size of their influence, and neither of the above two points can be guaranteed without incentives. Therefore, the design of the incentive mechanism is very important in the diffusion of crowdsourcing tasks.

The invention discloses a large-scale crowdsourcing task diffusion method based on linear probability. In this method, a reverse auction process is embodied between the crowdsourcing platform and registered users. Registered users submit a bidding document containing quotations to the crowdsourcing platform, and the crowdsourcing platform predicts its influence for each registered user, selects a batch of candidates from the registered users according to the prediction results, and calculates the remuneration of each registered user. The large-scale crowdsourcing task diffusion method proposed in the present invention has reward incentives for selected candidates, and satisfies computational effectiveness, personal rationality and anti-deception.

Contents of the invention

The technical problem to be solved by the present invention is to provide a large-scale crowdsourcing task diffusion method based on linear probability for the defects involved in the background technology.

Technical solution of the present invention is:

Consider a crowdsourcing platform that is owned by a social networking site and has a set of registered users who are a subset of the entire social network's users. When the crowdsourcing platform releases a batch of large-scale crowdsourcing tasks, the registered users themselves are not enough to complete the batch of tasks. At this time, the crowdsourcing platform will select a group of users from registered users, and use the influence of these users in the social network to spread large-scale crowdsourcing tasks in the social network.

A large-scale crowdsourcing task diffusion method based on linear probability described in the present invention is embodied as a reverse auction process between the crowdsourcing platform and registered users, and the steps are as follows:

Step 201: The social network application platform publishes a task set T={t ₁ ,...,t _m }, and for each task t _j ∈ T requires at least r _j people other than registered users to complete, where r _j is called the spread factor of task t _j ,;

Step 202: Set the set of registered users of the crowdsourcing platform as U _R ={1,2,...,n}, each registered user _i∈UR submits a bid Bi =(T _i , _{bi )} , where is the set of tasks that registered user i is willing to diffuse. b _i is the minimum reward that the registered user i wants to get for performing tasks in the task set T _i .

Step 203: The crowdsourcing platform predicts the influence of each registered user i on its social neighbors _v∈US Where U _S is the set of social neighbors of all registered users minus the registered user itself, and the size of U _S is set to q. The social network neighbors of the registered user i refer to the friends of the registered user i in the social network.

Step 204: The crowdsourcing platform calculates the set S of candidates;

Step 205: The crowdsourcing platform calculates the reward p _i for each user _i∈UR ;

Step 206: The crowdsourcing platform notifies the selected candidates, and the selected candidates perform task diffusion in the social network.

Step 207: The crowdsourcing platform pays the selected candidates.

In step 203, the crowdsourcing platform predicts the influence of each registered user i on its social neighbor v The steps are as follows:

Step 301: For any registered user i and its social neighbor v, calculate the Jaccard similarity coefficient Among them, N _i and N _v represent the set of social neighbors of registered user i and social neighbor v respectively.

Step 302: Calculate the influence of any registered user i on its social neighbor _v∈US : End.

In step 204, the crowdsourcing platform calculates the set S of candidates as follows:

Step 401: Initialize the candidate set

Step 402: Check whether each task t _j and the corresponding diffusion factor r _j are all 0; if yes, return to set S and end;

Step 403: Find from the set U _R \S The user i with the smallest value, where k∈U _R \S,

Step 404: Let S=S∪{i};

Step 405: For all tasks in T _i , update the diffusion factor r _j =r _j -min{r _j ,v _i }, and execute step 402.

In step 205, the crowdsourcing platform calculates the reward p _i for each user _i∈UR as follows:

Step 501: For any user _i∈UR , set its reward p _i =0;

Step 502: Execute Step 503 to Step 507 for all candidates i∈S;

Step 503: Let U _R '←U _R \{i},

Step 504: Check whether each task t _j and the corresponding diffusion factor r _j are all 0; if yes, execute step 502;

Step 505: Find U _R '\S' from the set The user i _k with the smallest value, where k∈U _R '\S',

Step 506: let S'=S'∪{i _k }; let

Step 507: For all The task in , update the diffusion factor Execute step 504;

Step 508: Return reward vector p=(p ₁ ,p ₂ ,...,p _n ) of all users, end.

Beneficial effect

Compared with the prior art, the present invention adopts the above technical scheme and has the following technical effects:

1. For the first time in a social network environment, a crowdsourcing task diffusion method with incentives is proposed. Even if there are only a few registered users, participants can be found through diffusion, which improves the task completion rate;

2. Provides a method to predict the influence of registered users, and selects registered users with high influence as candidates, which improves the efficiency of task diffusion;

3. The computational time complexity is low, and the time complexity of the reverse auction in this task diffusion method is O(n ² ·max{nq,m}). is a fully polynomial-time method that is computationally efficient;

4. The incentive for task diffusion is personal and rational, that is, the amount of remuneration paid by the platform to each selected candidate must be greater than or equal to the real cost of the registered user, so it has a positive effect on attracting a large number of registered users and improving the diffusion effect;

5. The incentive method in the crowdsourcing task diffusion method is anti-fraud. When other registered users submit their own real quotations, even if a user adopts a certain strategy to falsely report the quotation, it will not make the utility of the registered user becomes high, so registered users tend to submit their own real offers. Anti-deception plays an important role in preventing market monopoly or collusion.

Description of drawings

Fig. 1 is a reverse auction execution process embodied between the crowdsourcing platform and registered users in the present invention;

Figure 2 is the crowdsourcing platform in the present invention predicts the influence of each registered user i on its social neighbor v execution process;

Fig. 3 is the execution process of the crowdsourcing platform calculating the set S of candidates in the present invention;

Fig. 4 is the execution flow of calculating the reward p _i of each user _i∈UR by the crowdsourcing platform in the present invention.

detailed description

Below in conjunction with accompanying drawing, technical scheme of the present invention is described in further detail:

In the present invention, the social network application platform divides the users into multiple compatible user groups according to the compatible user sets submitted by the users. Winners are then selected and the compensation for each winner is calculated.

Noun description:

Crowdsourcing platform: A system that publishes tasks on the Internet and selects participants from the Internet to complete the tasks. In the present invention, the crowdsourcing platform belongs to a certain social network website, and the registered users of the crowdsourcing platform are a subset of the registered users of the social network. Crowdsourcing platforms can obtain certain social network information, such as the topology of the social network.

Candidates: The registered users selected by the task diffusion method proposed in the present invention are the final diffusers of crowdsourcing tasks.

User's utility: the difference between the user's reward and the cost paid by the user. In the anti-spoofing incentive method, the user's cost is equal to the user's offer.

A large-scale crowdsourcing task diffusion method based on linear probability described in the present invention is embodied as a reverse auction process between the crowdsourcing platform and registered users. The process is shown in Figure 1, and the steps are as follows:

Step 201: The social network application platform publishes a task set T={t ₁ ,...,t _m }, and for each task t _j ∈ T requires at least r _j people other than registered users to complete, where r _j is called the spread factor of task t _j ,;

Step 202: Set the set of registered users of the crowdsourcing platform as U _R ={1,2,...,n}, each registered user _i∈UR submits a bid Bi =(T _i , _{bi )} , where is the set of tasks that registered user i is willing to diffuse. b _i is the minimum reward that the registered user i wants to get for performing tasks in the task set T _i .

Step 203: The crowdsourcing platform predicts the influence of each registered user on its social neighbors _v∈US Where U _S is the set of social neighbors of all registered users minus the registered user itself, and the size of U _S is set to q. The social network neighbors of the registered user i refer to the friends of the registered user i in the social network.

Step 204: The crowdsourcing platform calculates the set S of candidates;

Step 205: The crowdsourcing platform calculates the reward p _i for each user _i∈UR ;

Step 206: The crowdsourcing platform notifies the selected candidates, and the selected candidates perform task diffusion in the social network.

Step 207: The crowdsourcing platform pays the selected candidates.

In step 203, the crowdsourcing platform predicts the influence of each registered user i on its social neighbor v The process flow is shown in Figure 2, and the steps are as follows:

Step 301: For any registered user i and its social neighbor v, calculate the Jaccard similarity coefficient Among them, N _i and N _v represent the set of social neighbors of registered user i and social neighbor v respectively.

Step 302: Calculate the influence of any registered user i on its social neighbor _v∈US : End.

In step 204, the flow of the crowdsourcing platform to calculate the candidate set S is shown in Figure 3, and the steps are as follows:

Step 401: Initialize the candidate set

Step 402: Check whether each task t _j and the corresponding diffusion factor r _j are all 0; if yes, return to set S and end;

Step 403: Find from the set U _R \S The user i with the smallest value, where k∈U _R \S,

Step 404: Let S=S∪{i};

Step 405: For all tasks in T _i , update the diffusion factor r _j =r _j -min{r _j ,v _i }, and execute step 402.

In step 205, the crowdsourcing platform calculates the reward p _i for each user _i∈UR as shown in Figure 4, and the steps are as follows:

Step 501: For any user _i∈UR , set its reward p _i =0;

Step 502: Execute Step 503 to Step 507 for all candidates i∈S;

Step 503: Let U _R '←U _R \{i},

Step 504: Check whether each task t _j and the corresponding diffusion factor r _j are all 0; if yes, execute step 502;

Step 505: Find U _R '\S' from the set The user i _k with the smallest value, where k∈U _R '\S',

Step 506: let S'=S'∪{i _k }; let

Step 507: For all The task in , update the diffusion factor Execute step 504;

Step 508: Return reward vector p=(p ₁ ,p ₂ ,...,p _n ) of all users, end.

Those skilled in the art can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It should also be understood that terms such as those defined in commonly used dictionaries should be understood to have a meaning consistent with the meaning in the context of the prior art, and will not be interpreted in an idealized or overly formal sense unless defined as herein Explanation.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (4)

1. a kind of extensive mass-rent task method of diffusion based on linear probability, it is characterised in that mass-rent platform and registered user Between be presented as a reverse auction process, step is as follows：

Step 201：Social networking application platform issues a set of tasks T={ t₁,...,t_m, for each task t_j∈ T are needed At least r that will be in addition to registered user_jIndividual goes to complete, wherein r_jIt is referred to as task t_jInvasin；

Step 202：If mass-rent platform registered user collection is combined into U_R={ 1,2 ..., n }, each registered user i ∈ U_RSubmit one Bidding documents B_i=(T_i,b_i), whereinThe set of tasks of diffusion is ready for registered user i；b_iTask-set is performed for registered user i T_iThe minimum remuneration that middle task is gone for；

Step 203：Mass-rent platform predicts each registered user i to its social neighbours v ∈ U_SInfluence powerWherein U_SIt is all The social neighbours of registered user subtract registered user in itself after set, if U_SSize be q；Registered user i social network Network neighbours refer to good friends of the registered user i in social networks；

Step 204：Mass-rent platform calculates those selected set S；

Step 205：Mass-rent platform calculates each user i ∈ U_RRemuneration p_i；

Step 206：Mass-rent platform notifies those selected, and those selected carries out task diffusion in social networks；

Step 207：Mass-rent platform is to those selected payt.

A kind of 2. extensive mass-rent task method of diffusion based on linear probability as claimed in claim 1, it is characterised in that In step 203, mass-rent platform predicts influence powers of each registered user i to its social neighbours vThe step of it is as follows：

Step 301：For any registered user i and its social neighbours v, Jie Kade similarity factors are calculatedWherein N_iAnd N_vRegistered user i and social neighbours v social neighbours are represented respectively Set；

Step 302：Any registered user i is calculated to its social neighbours v ∈ U_SInfluence power：Terminate.

A kind of 3. extensive mass-rent task method of diffusion based on linear probability as claimed in claim 1, it is characterised in that In step 204, the step of mass-rent platform calculates those selected set S, is as follows：

Step 401：Those selected is initialized to gather

Step 402：Check each task t_jWith corresponding invasin r_jWhether all it is 0；If it is, returning to set S, terminate；

Step 403：The U from set_RS findThe minimum user i, wherein k ∈ U of value_RS,

Step 404：Make S=S ∪ { i }；

Step 405：To all T_iIn task, renewal invasin r_j=r_j-min{r_j,v_i, perform step 402.

A kind of 4. extensive mass-rent task method of diffusion based on linear probability as claimed in claim 1, it is characterised in that In step 205, mass-rent platform calculates each user i ∈ U_RRemuneration p_iThe step of it is as follows：

Step 501：To any user i ∈ U_R, make its remuneration p_i=0；

Step 502：Step 503 is performed to those selected all i ∈ S and arrives step 507；

Step 503：Order

Step 504：Check each task t_jWith corresponding invasin r_jWhether all it is 0；If it is, perform step 502；

Step 505：The U from set_R' S' findThe minimum user i of value_k, wherein k ∈ U_R' S',

Step 506：Make S'=S' ∪ { i_k}；Order

Step 507：To all T_ikIn task, update invasinPerform step 504；

Step 508：Return to the reward vector p=(p of all users₁,p₂,...,p_n), terminate.