CN115358781B - Crowd sensing noise monitoring task recommendation method based on limited rational decision model - Google Patents
Crowd sensing noise monitoring task recommendation method based on limited rational decision model Download PDFInfo
- Publication number
- CN115358781B CN115358781B CN202211004808.8A CN202211004808A CN115358781B CN 115358781 B CN115358781 B CN 115358781B CN 202211004808 A CN202211004808 A CN 202211004808A CN 115358781 B CN115358781 B CN 115358781B
- Authority
- CN
- China
- Prior art keywords
- task
- participant
- decision model
- value
- participants
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0207—Discounts or incentives, e.g. coupons or rebates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Abstract
A crowd sensing noise monitoring task recommendation method based on a rational decision model is composed of initializing a participant decision model and credibility, issuing a task to read participant information, matching participants and recommending the task, selecting task completion and updating credibility, and updating the participant decision model. The method constructs a decision model of limited authority of participants, regularly updates the decision model of the participants, reduces the cost, ensures the accuracy of task-participant matching, provides high-quality data meeting requirements for the tasks, avoids the participants doing useless work, reduces the resource waste, initializes the decision model and the credibility of the participants by using the registered and submitted task preference information and the social friend task information of the newly added participants, ensures that the new participants can be properly recommended, and promotes the noise monitoring task to be better completed.
Description
Technical Field
The invention relates to the technical field of crowd sensing network application, in particular to a crowd sensing noise monitoring task recommendation method based on a limited rational decision model.
Background
In recent years, with the vigorous development of the industrial process in China, various noises from traffic vehicles, industrial manufacturing, building construction and social life often interfere with the daily life of people, and harm to human health to a certain extent. Accurate noise monitoring plays an important role in controlling the source of the noise. The traditional noise monitoring is mainly based on static perception and is easily limited by insufficient node coverage, high installation and management cost and the like, and the important research value and practical significance are realized on how to reasonably utilize the existing group resources to efficiently monitor the noise information.
Crowd sensing utilizes existing mobile devices of people to complete large-scale tasks with low cost and high efficiency, and is a new emerging mode applied to environmental monitoring (such as noise monitoring, PM2.5 monitoring and the like) at present. A typical crowd sensing system comprises a task requester, a server platform and task participants, wherein the requester issues crowd sensing tasks through the server platform, the platform distributes the tasks to the participants, and the participants utilize own sensing equipment to execute the tasks and obtain benefits. Because the number of tasks and participants in the server platform is huge, how to recommend the tasks to the proper participants is a key link to efficiently complete the tasks.
Most of the existing recommendation methods are respectively used for performing complex matching calculation on each perception task to select participants, and only a certain characteristic of the participants is considered to optimize a matching process, so that the limited rationality of the participants in the selection process is ignored, namely, the criteria which are searched by people in the decision process are not the maximum or optimal criteria but are just the satisfactory criteria, some tasks which should be recommended are not recommended, and the participant income is reduced invisibly. Therefore, a technical problem to be solved at present is to provide a recommendation method for crowd sensing noise monitoring task with high efficiency and considering the limited rationality of participants.
Disclosure of Invention
The technical problem to be solved by the invention is to overcome the defects of the technical problems and provide a recommendation method of a crowd sensing noise monitoring task based on a limited rational decision model, which is efficient and considers the limited rationality of participants.
The technical scheme adopted for solving the technical problems comprises the following steps:
(1) Initializing participant decision models and goodwill
Platform participant p m The information is as follows:
p m ={Dec,Cr,St,Lo} (1)
wherein m is the number of platform participants, dec is a participant decision model, cr represents the credit degree, cr belongs to [0,1], st is the current state, and the platform has two states of idle and busy, and Lo is the current position.
Determining an initial value of the credit Cr according to the formula (2):
wherein, cr i And s is the credibility of the ith friend of the participant, and the number of the friends of the participant.
New joining participant p n With newly joining participant p n Task information submitted at registration and newly joining participant p n Constructing a participant decision model Dec by historical task information of social friends, and adding a participant p newly n Is a newly joined participant p n Task information submitted during registration with newly joining participant p n Weighted average of social friend historical task information, determining newly joining participant p according to equation (3) n Initial task information Infi of jth value of task characteristic k k,j :
Wherein, infra k,j For newly joining a participant p n Information of jth dereferencing of task characteristics k submitted in registration, wherein k belongs to { t, l, pr }, and t, l and pr represent task characteristics which are respectively task time, task place and task profit, and m 1 For newly joining a participant p n Total number of tasks performed, vk j Representing the value result of the task characteristic k, taking alpha and beta as parameters, and belonging to alpha [0.3,1.0 ∈ ]],β∈[0,0.7]And α + β is 1,k j The j-th value of the task characteristic k is taken.
The method for constructing the participant decision model Dec is as follows:
1) Determining task characteristic information entropy Ent according to formula (4) k :
Wherein n is 1 Representing the number of values of the task feature k, pk j Representing the probability of the jth value of the task characteristic k; inf k,j The j-th value of the task characteristic k is task information m 2 The total number of tasks performed for the participants of the participant decision model Dec.
2) Determining a decision result Res of the participant decision model Dec according to equation (5):
Br=Dt+DI+Dr
wherein Br is the sum of the preference degrees of the characteristics of each task, Q is a result threshold, Q belongs to [3,6],1 represents that the task is suitable, and 0 represents that the task is not suitable; dt is the preference degree of the task time, dl is the preference degree of the task place, and Dr is the preference degree of the task profit.
Each task feature represents a non-leaf node of one layer in the participant decision model Dec, and the entropy Ent of the task feature information obtained by equation (4) k Layering from small to large, task characteristic information entropy Ent k In the small upper layer, the preference degree value of the task characteristics is taken as a branch of each non-leaf node, and the preference degree value is the same as the preference degree value of each non-leaf node and is taken as the same branch; 2) The decision result Res in the step is a leaf node of the participant decision model Dec, a branch of each non-leaf node is connected with a non-leaf node or a leaf node of one layer in the participant decision model Dec, and the operation is repeated until a branch of each non-leaf node is connected with a leaf node of one layer in the participant decision model Dec to construct the participant decision model Dec.
(2) Publishing task read participant information
Publishing the crowd sourcing task T according to equation (6):
T={Ti,Loc,Rew,Num,Cre} (6)
wherein Ti is the time of the task, loc is the place of the task, rew is the reward of the task, num is the number of the required participants, and Cre is the credit degree required by the participants of the task.
And reading all participant decision models Dec and the credibility Cr from the platform.
(3) Matching participants and recommending tasks
When a new task is issued, matching with a participant decision model Dec in a participant set P in sequence according to task characteristics:
P={p 1 ,p 2 ,...,p m } (7)
wherein m is a finite positive integer; predictive platform participant p m If it is appropriate for this task, if platform participant p m Suitable for this task, the platform participant p m Joining a participant pre-selection set P according to equation (8) p :
P p ={p p1 ,p p2 ,...,p pn } (8)
Wherein p is pn Representing a preselected participant, n is a finite positive integer, and n ≦ m.
Determining preselected participant p pn If the credit Cre requirement required by the task participant is met, the participant p is preselected pn The task to meet the requirement is added to the preselected participant p according to equation (9) pn Task recommendation list T k :
T k ={t 1 ,t 2 ,...,t a } (9)
Wherein, t a To suit a preselected participant p pn A is a finite positive integer.
Preselection of participants p pn Task recommendation list T k Ranking by profit from more to less, determining preselected participant p according to equation (10) pn Gain Pro pn,a :
c pn,a =λ 1 ct pn,a +λ 2 cl pn,a
Wherein, rew a As task t a A reward of, c pn,a For pre-selected participants p pn Performing task t a Cost consumed, ct pn,a For pre-selected participants p pn Performing task t a Time taken, cl pn,a Is a pre-selected participant p pn Performing task t a Cost of distance used, λ 1 、λ 2 For cost estimation of the parameters, λ 1 ∈[0,1],λ 2 ∈[0,1]And λ of 1 +λ 2 Is 1,lon pn 、lon a Are respectively preselected participants p pn And task t a Position longitude of (1), lat pn 、lat a Respectively representing preselected participants p pn And task t a The location latitude of (a).
(4) Selecting task completion and updating reputation
Platform participant p m Selecting task completion according to self current state St, and updating platform participant p according to formula (11) m Reputation degree Cr' m :
Wherein, cr m To update the reputation before updating, q m,a As a platform participant p m For task t a Quality of submitted data, q avg,a As task t a All participants submitted an average of the data quality.
(5) Updating participant decision models
Platform participant p m When the using time of the participant decision model Dec reaches M months, M belongs to [1,3 ]]M is a positive integer, using platform participant p m Task information update platform participant p m By the participant decision model Dec, platform participant p is determined as in equation (12) m Task information Inf of j-th value of task characteristic k m,k,j :
Wherein m is 3 As a platform participant p m Total number of tasks performed, vk j Representing the result of the task feature k, k j The j-th value of the task characteristic k is taken.
And (3) repeating the construction method of the participant decision model Dec in the step (1) to obtain an updated participant decision model Dec.
In formula (3) in step (1) of the present invention, α and β are parameters, α ∈ [0.3,1.0], β ∈ [0,0.7], and α + β is 1, and the value of β is determined according to formula (13):
where s is the number of friends of the participant.
In the formula (5) of the step (1), the values of the preference degree Dt of the task time, the preference degree Dl of the task location and the preference degree Dr of the task income are as follows:
wherein, pt j Probability of j-th value of task time, mu 1 、θ 1 A probability threshold value divided for degree, and mu 1 >θ 1 ,μ 1 ∈[0.1,0.2],θ 1 ∈[0.005,0.01]。
Wherein, pl j Probability of the j-th value of the task site, mu 2 、θ 2 A probability threshold value divided for degree, and mu 2 >θ 2 ,μ 2 ∈[0.1,0.25],θ 2 ∈[0,0.01]。
And Vr is a value of the task income, avg is an average value of all historical task income values, and min is a minimum value of all historical task income values.
According to the method, the preference degree of the participants to the tasks is extracted through implicit feedback of historical task information, a decision model of the participants with limited rationality is constructed, the appropriate tasks are actively recommended to the participants, the decision model of the participants is updated regularly, resource consumption is reduced, and the accuracy of task-participant matching is ensured; in the task recommendation process, the credibility of the participants is considered, the credibility of the participants is updated according to the data quality of the participants, high-quality data meeting requirements are provided for the tasks, the participants are prevented from doing useless work, and the resource waste is reduced; and initializing a participant decision model and the credibility of the newly added participant by using the registered submitted task preference information and the social friend task information of the newly added participant, ensuring that the newly added participant can be recommended more appropriately, and promoting the noise monitoring task to be completed better.
Drawings
FIG. 1 is a flowchart of example 1 of the present invention.
Fig. 2 is a graph of the gains from participants completing 20 tasks.
FIG. 3 is a graph of task-participant matching time consumption versus participant population.
FIG. 4 is a graph of participant decision model prediction accuracy results.
FIG. 5 is a graph of participant reputation update for three different cases.
Detailed Description
The present invention will be described in further detail below with reference to the drawings and examples, but the present invention is not limited to the embodiments described below.
Example 1
Taking 2045 users whose check-in times exceed 300 in 2009 in a Brightkite real data set and 1806242 total completed tasks as an example, the steps of the crowd sensing noise monitoring task recommendation method based on the rational decision model in this embodiment are as follows (see fig. 1):
(1) Initializing participant decision models and goodwill
Platform participant p m The information is as follows:
p m ={Dec,Cr,St,Lo} (1)
wherein m is the number of platform participants, dec is a participant decision model, cr represents the credit degree, cr belongs to [0,1], st is the current state, and the platform has two states of idle and busy, and Lo is the current position.
Determining an initial value of the credit Cr according to the formula (2):
wherein, cr i And the credibility of the ith friend of the participant is s, and the number of friends of the participant is s.
New joining participant p n With newly joining participant p n Task information submitted at registration and newly joining participant p n Constructing a participant decision model Dec by historical task information of social friends, and adding a participant p newly n Is a newly joined participant p n Task information submitted during registration with newly joining participant p n Weighted average of social friend historical task information, determining newly joining participant p according to equation (3) n Initial task information Infi of jth value of task characteristic k k,j :
Wherein, infra k,j For newly joining a participant p n The j-th value information of the task characteristics k submitted during registration, wherein k belongs to { t, l, pr }, t, l and pr represent the task characteristics which are respectively task time, task place and task income, and m is 1 For newly joining participant p n Total number of tasks performed, vk j Representing the value result of the task characteristic k, taking alpha and beta as parameters, and belonging to alpha [0.3,1.0 ∈ ]],β∈[0,0.7]And α + β is 1,k j The j-th value of the task characteristic k is taken.
In this embodiment, the value of β is determined according to equation (13):
wherein s is the number of friends of the participant, and after the number of friends of the participant is determined, the value of β is determined, so that the value of α can be obtained.
The method for constructing the participant decision model Dec is as follows:
1) Determining task characteristic information entropy Ent according to formula (4) k :
Wherein n is 1 Representing the number of values of the task feature k, pk j Representing the probability of the jth value of the task characteristic k; inf (Inf) k,j The j-th value of the task characteristic k is task information m 2 The total number of tasks performed for the participants of the participant decision model Dec.
2) Determining a decision result Res of the participant decision model Dec according to equation (5):
Br=Dt+Dl+Dr
wherein Br is the sum of preference degrees of characteristics of each task, Q is a result threshold, Q belongs to [3,6], a value of 5,1 of Q in this embodiment indicates that the task is suitable, and 0 indicates that the task is not suitable; dt is the preference degree of the task time, dl is the preference degree of the task location, and Dr is the preference degree of the task income.
The preference degree Dt of the task time, the preference degree Dl of the task place and the preference degree Dr of the task income of the embodiment take the following values:
wherein, pt j Probability of the j-th value of the task time, mu 1 、θ 1 A probability threshold value divided for degree, and mu 1 >θ 1 ,μ 1 ∈[0.1,0.2]μ of the present embodiment 1 Value of 0.15, theta 1 ∈[0.005,0.01]Theta of the present embodiment 1 The value is 0.008.
Wherein, pl j Probability of j-th value of task location, mu 2 、θ 2 A probability threshold value divided for degree, and mu 2 >θ 2 ,μ 2 ∈[0.1,0.25],θ 2 ∈[0,0.01](ii) a μ of the present example 2 Value of 0.18, theta 2 The value is 0.005.
And Vr is a value of the task income, avg is an average value of all historical task income values, and min is a minimum value of all historical task income values.
Each task feature represents a non-leaf node of one layer in the participant decision model Dec, and the entropy Ent of the task feature information obtained by equation (4) k Layering from small to large, task characteristic information entropy Ent k In the upper layer, the preference degree value of the task characteristics is taken as a branch of each non-leaf node, and the preference degree value is taken as the same branch; 2) The decision result Res in the step is a leaf node of the participant decision model Dec, a branch of each non-leaf node is connected with a non-leaf node or a leaf node of one layer in the participant decision model Dec, and the operation is repeated until a branch of each non-leaf node is connected with a leaf node of one layer in the participant decision model Dec to construct the participant decision model Dec.
(2) Publishing task read participant information
And (5) issuing a crowd sourcing task T according to the formula (6):
T={Ti,Loc,Rew,Num,Cre} (6)
wherein Ti is the task time, loc is the task place, rew is the reward of the task, num is the number of required participants, and Cre is the credit degree required by the task participants.
And reading all participant decision models Dec and the credibility Cr from the platform.
(3) Matching participants and recommending tasks
When a new task is issued, sequentially matching with a participant decision model Dec in a participant set P according to a task characteristic in an equation (7):
P={p 1 ,p 2 ,...,p m } (7)
wherein m is a finite positive integer; predictive platform participant p m If it is appropriate for this task, if platform participant p m Suitable for this task, the platform participant p m Joining a Pre-selection set of participants P according to equation (8) p :
P p ={p p1 ,p p2 ,...,p pn } (8)
Wherein p is pn Representing a preselected participant, n is a finite positive integer, and n ≦ m.
Determining preselected participant p pn If the credit Cre requirement required by the task participant is met, the participant p is preselected pn The on-demand task is added to the preselected participant p according to equation (9) pn Task recommendation list T k :
T k ={t 1 ,t 2 ,...,t a } (9)
Wherein, t a To suit a preselected participant p pn A is a finite positive integer.
Preselection of participants p pn Task recommendation list T k Ranking by profit from more to less, determining preselected participant p according to equation (10) pn Profit Pro pn,a :
c pn,a =λ 1 ct pn,a +λ 2 cl pn,a
Wherein, rew a As task t a Reward of (c) pn,a Is a pre-selected participant p pn Performing task t a Cost consumed, ct pn,a Is a pre-selected participant p pn Performing task t a Time taken, cl pn,a Is a pre-selected participant p pn Performing task t a Cost of distance used, λ 1 、λ 2 For cost estimation of the parameters, λ 1 ∈[0,1],λ 2 ∈[0,1]λ of the present embodiment 1 The value of 0.5, λ 2 A value of 0.5, and λ 1 +λ 2 Is 1,lon pn 、lon a Are respectively preselected participants p pn And task t a Position longitude of (1), lat pn 、lat a Respectively representing preselected participants p pn And task t a The location latitude of (c).
(4) Selecting task completion and updating reputation
Platform participant p m Selecting task completion according to self current state St, and updating platform participant p according to formula (11) m Reputation degree Cr' m :
Wherein, cr m To update the degree of reputation before update, q m,a As a platform participant p m For task t a Quality of submitted data, q avg,a As task t a All participants submitted an average of the data quality.
(5) Updating participant decision models
Platform participant p m When the using time of the participant decision model Dec reaches M months, M belongs to [1,3 ]]M is a positive integer, M of the embodiment takes a value of 2, and a platform participant p is used m Task information update platform participant p m By the participant decision model Dec, platform participant p is determined as in equation (12) m Task information Inf of j-th value of task characteristic k m,k,j :
Wherein m is 3 As a platform participant p m Total number of tasks performed, vk j Representing the result of the task feature k, k j Is a task ofThe jth value of the transaction characteristic k.
And (3) repeating the construction method of the participant decision model Dec in the step (1) to obtain an updated participant decision model Dec.
And finishing the recommendation method of the crowd sensing noise monitoring task based on the limited rational decision model.
Example 2
Taking 2045 users whose sign-in times exceed 300 in 2009 in a Brightkite real data set and 1806242 total completed tasks as an example, the crowd sensing noise monitoring task recommendation method based on the rational decision model of the embodiment includes the following steps:
(1) Initializing participant decision models and goodwill
Platform participant pm information is as follows:
p m ={Dec,Cr,St,Lo} (1)
wherein m is the number of platform participants, dec is a participant decision model, cr represents the credit degree, cr belongs to [0,1], st is the current state, and the platform has two states of idle and busy, and Lo is the current position.
Determining an initial value of the credit Cr according to the formula (2):
wherein, cr i And s is the credibility of the ith friend of the participant, and the number of the friends of the participant.
New joining participant p n With newly joining participant p n Task information submitted at registration and newly joining participant p n Constructing a participant decision model Dec by historical task information of social friends, and adding a participant p newly n Is a newly joined participant p n Task information submitted during registration with newly joining participant p n Weighted average of social friend historical task information, determining newly joining participant p according to equation (3) n Initial task information Infi of jth value of task characteristic k k,j :
Wherein, infra k,j For newly joining a participant p n The j-th value information of the task characteristics k submitted during registration, wherein k belongs to { t, l, pr }, t, l and pr represent the task characteristics which are respectively task time, task place and task income, and m is 1 For newly joining participant p n Total number of tasks performed, vk j Representing the value result of the task characteristic k, taking alpha and beta as parameters, and belonging to alpha [0.3,1.0 ∈ ]],β∈[0,0.7]And α + β is 1,k j The j-th value of the task characteristic k is taken.
In this embodiment, the value of β is determined according to equation (13):
wherein s is the number of friends of the participant, and after the number of friends of the participant is determined, the value of β is determined, so that the value of α can be obtained.
The method for constructing the participant decision model Dec is as follows:
1) Determining task characteristic information entropy Ent according to formula (4) k :
Wherein n is 1 Representing the number of values of the task feature k, pk j Representing the probability of the jth value of the task characteristic k; inf k,j The j-th value of the task characteristic k is task information m 2 Decide for the participantTotal number of tasks performed by participants of the policy model Dec.
2) Determining a decision result Res of the participant decision model Dec according to equation (5):
Br=Dt+Dl+Dr
wherein Br is the sum of preference degrees of characteristics of each task, Q is a result threshold, Q belongs to [3,6], a value of 3,1 of Q in this embodiment indicates that the task is suitable, and 0 indicates that the task is not suitable; dt is the preference degree of the task time, dl is the preference degree of the task place, and Dr is the preference degree of the task profit.
The values of the preference degree Dt of the task time, the preference degree Dl of the task place and the preference degree Dr of the task income of the embodiment are as follows:
wherein, pt j Probability of the j-th value of the task time, mu 1 、θ 1 A probability threshold value divided for degree, and mu 1 >θ 1 ,μ 1 ∈[0.1,0.2]μ of the present embodiment 1 The value of 0.1, theta 1 ∈[0.005,0.01]Theta of the present embodiment 1 The value is 0.005.
Wherein, pl j Probability of the j-th value of the task site, mu 2 、θ 2 A probability threshold value divided for degree, and mu 2 >θ 2 ,μ 2 ∈[0.1,0.25],θ 2 ∈[0,0.01](ii) a μ of the present example 2 The value of 0.1, theta 2 The value is 0.
Vr is a value of the task income, avg is an average value of values of all historical task income, and min is a minimum value of values of all historical task income.
The other steps of this step are the same as in example 1, and a participant decision model Dec is constructed.
(2) Publishing task read participant information
This procedure is the same as in example 1.
(3) Matching participants and recommending tasks
This procedure is the same as in example 1.
(4) Selecting task completion and updating credibility
This procedure is the same as in example 1.
(5) Updating participant decision models
Platform participant p m When the using time of the participant decision model Dec reaches M months, M belongs to [1,3 [ ]]M is a positive integer, M of this embodiment takes a value of 1, and a platform participant p is used m Task information update platform participant p m By the participant decision model Dec, platform participant p is determined as in equation (12) m Task information Inf of j-th value of task characteristic k m,k,j :
Wherein m is 3 As a platform participant p m Total number of tasks performed, vk j Representing the result of the task feature k, k j The j-th value of the task characteristic k is taken.
And (3) repeating the construction method of the participant decision model Dec in the step (1) to obtain an updated participant decision model Dec.
And finishing the recommendation method of the crowd sensing noise monitoring task based on the limited rational decision model.
Example 3
Taking 2045 users whose sign-in times exceed 300 in 2009 in a Brightkite real data set and 1806242 total completed tasks as an example, the crowd sensing noise monitoring task recommendation method based on the rational decision model of the embodiment includes the following steps:
(1) Initializing participant decision models and goodwill
Platform participant p m The information is as follows:
p m ={Dec,Cr,St,Lo} (1)
wherein m is the number of platform participants, dec is a participant decision model, cr represents the credit degree, cr belongs to [0,1], st is the current state, and the platform has two states of idle and busy, and Lo is the current position.
Determining an initial value of the credit Cr according to the formula (2):
wherein, cr i And s is the credibility of the ith friend of the participant, and the number of the friends of the participant.
New joining participant p n With newly joining participant p n Task information submitted at registration and newly joining participant p n Constructing a participant decision model Dec by historical task information of social friends, and adding a participant p newly n Is a newly joined participant p n Task information submitted during registration with newly joining participant p n Weighted average of social friend historical task information, determining newly joined participant p according to equation (3) n Initial task information Infi of jth value of task characteristic k k,j :
Wherein, infra k,j For newly joining participant p n The j-th value information of the task characteristics k submitted during registration, wherein k belongs to { t, l, pr }, t, l and pr represent the task characteristics which are respectively task time, task place and task income, and m is 1 For newly joining participant p n Total number of tasks performed, vk j Representing the value result of the task characteristic k, taking alpha and beta as parameters, and belonging to alpha [0.3,1.0 ∈ ]],β∈[0,0.7]And α + β is 1,k j The j-th value of the task characteristic k is taken.
In this embodiment, the value of β is determined according to equation (13):
wherein s is the number of friends of the participant, and after the number of friends of the participant is determined, the value of β is determined, so that the value of α can be obtained.
The method for constructing the participant decision model Dec is as follows:
1) Determining task characteristic information entropy Ent according to formula (4) k :
Wherein n is 1 Represents the number of values, pk, of the task feature k j Representing the probability of the jth value of the task characteristic k; inf k,j The j-th value of the task characteristic k is task information m 2 The total number of tasks performed for the participants of the participant decision model Dec.
2) Determining a decision result Res of the participant decision model Dec according to equation (5):
Br=Dt+Dl+Dr
wherein Br is the sum of preference degrees of characteristics of each task, Q is a result threshold, Q belongs to [3,6], a value of 3,1 of Q in this embodiment indicates that the task is suitable, and 0 indicates that the task is not suitable; dt is the preference degree of the task time, dl is the preference degree of the task place, and Dr is the preference degree of the task profit.
The values of the preference degree Dt of the task time, the preference degree Dl of the task place and the preference degree Dr of the task income of the embodiment are as follows:
wherein, pt j Probability of j-th value of task time, mu 1 、θ 1 A probability threshold value divided for degree, and mu 1 >θ 1 ,μ 1 ∈[0.1,0.2]μ of the present embodiment 1 Value of 0.2, theta 1 ∈[0.005,0.01]θ of the present embodiment 1 The value is 0.01.
Wherein, pl j Probability of j-th value of task location, mu 2 、θ 2 A probability threshold value divided for degree, and mu 2 >θ 2 ,μ 2 ∈[0.1,0.25],θ 2 ∈[0,0.01](ii) a μ of the present example 2 Value of 0.25, theta 2 The value is 0.01.
And Vr is a value of the task income, avg is an average value of all historical task income values, and min is a minimum value of all historical task income values.
The other steps of this step are the same as in example 1, and a participant decision model Dec is constructed.
(2) Publishing task read participant information
This procedure is the same as in example 1.
(3) Matching participants and recommending tasks
This procedure is the same as in example 1.
(4) Selecting task completion and updating credibility
This procedure is the same as in example 1.
(5) Updating participant decision models
Platform participant p m When the using time of the participant decision model Dec reaches M months, M belongs to [1,3 ]]M is a positive integer, M in this embodiment takes the value of 3, and a platform participant p is used m Task information update platform participant p m By the participant decision model Dec, platform participant p is determined as in equation (12) m Task information Inf of j-th value of task characteristic k m,k,j :
Wherein m is 3 Total number of tasks performed for the participant, vk j Representing the result of the task feature k, k j The j-th value of the task characteristic k is taken.
And (3) repeating the construction method of the participant decision model Dec in the step (1) to obtain an updated participant decision model Dec.
And finishing the recommendation method of the crowd sensing noise monitoring task based on the limited rational decision model.
In order to verify the beneficial effects of the present invention, the inventor carried out a simulation experiment by using the recommendation method of the crowd sensing noise monitoring task based on the limited rational decision model in embodiment 1 of the present invention, the situations are as follows:
the gains obtained by the participants completing 20 tasks are shown in fig. 2, the total gains obtained after the 6 different participants participate in 20 tasks are shown in fig. 2, the value range of the single task gain is 1-10 tokens, the round dots represent that the method is used, and the triangles represent that the tasks are randomly selected. As can be seen from FIG. 2, the yield of the participant who completes 20 tasks by using the method of the present invention is significantly higher than that of the randomly selected task, which indicates that the method of the present invention can improve the yield of the participant.
The relationship between the consumption of the task-participant matching time and the number of participants is shown in fig. 3, fig. 3 shows the average time consumed by matching 1000 tasks with the number of different participants for 5 times, and as can be seen from fig. 3, under the condition that the number of the tasks is unchanged, the task-participant matching time is increased along with the increase of the number of the participants, and the time consumed by matching 1000 tasks with 300 participants is 0.631 second, which indicates the high efficiency of the method.
The result of the prediction accuracy of the decision model of the participant is shown in fig. 4, fig. 4 shows the change of the prediction accuracy of the decision model in different time ranges, as can be seen from fig. 4, when the updating interval duration of the prediction accuracy is one month, the prediction accuracy is the highest, and the prediction accuracy decreases with the increase of time from one month to the next, which shows that the method of the present invention has the advantages that the decision model is set to be updated regularly, the method is reasonable and effective, and the prediction accuracy is ensured while the resource consumption is reduced.
The reputation updating of the participant under three different conditions is shown in fig. 5, fig. 5 shows reputation updating curves under three different conditions, where user _1 is a condition that the data quality of the participant is higher than the average data quality, user _2 is a condition that the data quality of the participant fluctuates from the average data quality, and user _3 is a condition that the data quality of the participant is lower than the average data quality; as can be seen from fig. 5, the credibility changes with the change of the data quality, the credibility increases when the quality is high, and the credibility decreases when the quality is low, which indicates the rationality of the credibility update function of the method of the present invention.
Claims (3)
1. A crowd sensing noise monitoring task recommendation method based on a limited rational decision model is characterized by comprising the following steps:
(1) Initializing participant decision models and goodwill
Platform participant p m The information is as follows:
p m ={Dec,Cr,St,Lo} (1)
wherein m is the number of platform participants, dec is a participant decision model, cr represents the credit degree, cr belongs to [0,1], st is the current state, and the current state comprises an idle state and a busy state, and Lo is the current position;
determining an initial value of the credit Cr according to the formula (2):
wherein, cr i The credibility of the ith friend of the participant is shown, and s is the number of friends of the participant;
new joining participant p n With newly joining participant p n Task information submitted at registration and newly joining participant p n Constructing a participant decision model Dec by historical task information of social friends, and adding a participant p newly n Is a newly joined participant p n Task information submitted during registration with newly joining participant p n Weighted average of social friend historical task information, determining newly joining participant p according to equation (3) n Initial task information Infi of jth value of task characteristic k k,j :
Wherein, infra k,j For newly joining a participant p n Task feature k j' th fetch submitted at registrationThe value information, k belongs to { t, l, pr }, and t, l and pr represent task characteristics which are respectively task time, task location and task income, and m is 1 For newly joining participant p n Total number of tasks performed, vk j Representing the value result of the task characteristic k, taking alpha and beta as parameters, and belonging to alpha [0.3,1.0 ∈ ]],β∈[0,0.7]And α + β is 1,k j Taking the j-th value of the task characteristic k;
the method for constructing the participant decision model Dec is as follows:
1) Determining task characteristic information entropy Ent according to formula (4) k :
Wherein n is 1 Representing the number of values of the task feature k, pk j Representing the probability of the jth value of the task characteristic k; inf k,j The j-th value of the task characteristic k is task information m 2 A total number of tasks performed for the participants of the participant decision model Dec;
2) Determining a decision result Res of the participant decision model Dec according to equation (5):
Br=Dt+DI+Dr
wherein Br is the sum of the preference degrees of the characteristics of each task, Q is a result threshold, Q belongs to [3,6],1 represents that the task is suitable, and 0 represents that the task is not suitable; dt is the preference degree of the task time, dl is the preference degree of the task place, and Dr is the preference degree of the task income;
each task feature represents a non-leaf node of one layer in the participant decision model Dec, and the entropy Ent of the task feature information obtained by equation (4) k Layering from small to large, task characteristic information entropy Ent k Is small and smallOn the upper layer, taking the preference degree value of the task characteristics as a branch of each non-leaf node, and taking the same value as the same branch; 2) The decision result Res in the step is a leaf node of the participant decision model Dec, the branch of each non-leaf node is connected with a non-leaf node or a leaf node of one layer in the participant decision model Dec, and the operation is repeated until the branch of each non-leaf node is connected with a leaf node of one layer in the participant decision model Dec to construct the participant decision model Dec;
(2) Publishing task read participant information
And (5) issuing a crowd sourcing task T according to the formula (6):
T={Ti,Loc,Rew,Num,Cre} (6)
wherein Ti is task time, loc is task place, rew is reward of the task, num is number of required participants, and Cre is credit degree required by the task participants;
reading decision models Dec and credit degrees Cr of all participants from a platform;
(3) Matching participants and recommending tasks
When a new task is issued, matching with a participant decision model Dec in a participant set P in sequence according to task characteristics:
P={p 1 ,p 2 ,...,p m } (7)
wherein m is a finite positive integer; predictive platform participant p m If it is appropriate for this task, if platform participant p m Suitable for this task, the platform participant p m Joining a participant pre-selection set P according to equation (8) p :
P p ={p p1 ,p p2 ,...,p pn } (8)
Wherein p is pn Representing a preselected participant, n being a finite positive integer, and n ≦ m;
determining preselected participant p pn If the credit Cre requirement required by the task participant is met, the participant p is preselected pn The task to meet the requirement is added to the preselected participant p according to equation (9) pn Task recommendation list T k :
T k ={t 1 ,t 2 ,...,t a } (9)
Wherein, t a To suit a preselected participant p pn A is a finite positive integer;
preselection of participants p pn Task recommendation list T k Ranking by profit from more to less, determining preselected participant p according to equation (10) pn Profit Pro pn,a :
c pn,a =λ 1 ct pn,a +λ 2 cl pn,a
Wherein, rew a As task t a A reward of, c pn,a Is a pre-selected participant p pn Performing task t a Cost consumed, ct pn,a For pre-selected participants p pn Performing task t a Time taken, cl pn,a For pre-selected participants p pn Performing task t a Cost of distance used, λ 1 、λ 2 For cost estimating the parameters, λ 1 ∈[0,1],λ 2 ∈[0,1]And λ 1 +λ 2 Is 1,lon pn 、lon a Are respectively preselected participants p pn And task t a Position of longitude, lat pn 、lat a Respectively representing preselected participants p pn And task t a The location latitude of (a);
(4) Selecting task completion and updating reputation
Platform participant p m Selecting task completion according to self current state St, and updating platform participant p according to formula (11) m Reputation degree Cr' m :
Wherein, cr m To update the reputation before updating, q m,a As a platform participant p m For task t a Quality of submitted data, q avg,a As task t a All participants submitted an average of the data quality;
(5) Updating participant decision models
Platform participant p m When the using time of the participant decision model Dec reaches M months, M belongs to [1,3 ]]M is a positive integer, with platform participant p m Task information update platform participant p m By the participant decision model Dec, platform participant p is determined as in equation (12) m Task information Inf of j-th value of task characteristic k m,k,j :
Wherein m is 3 As a platform participant p m Total number of tasks performed, vk j Representing the result of the task feature k, k j Taking the j-th value of the task characteristic k;
and (3) repeating the construction method of the participant decision model Dec in the step (1) to obtain an updated participant decision model Dec.
2. The recommendation method of the crowd-sensing noise monitoring task based on the rationality decision model according to claim 1, characterized in that in the formula (3) in the step (1), α and β are parameters, α e [0.3,1.0], β e [0,0.7], and α + β is 1, and the value of β is determined according to the formula (13):
where s is the number of friends of the participant.
3. The method for recommending the crowd-sourcing aware noise monitoring task based on the rational decision model as claimed in claim 1, wherein: in the formula (5) in the step (1), the values of the preference degree Dt of the task time, the preference degree Dl of the task place and the preference degree Dr of the task income are as follows:
wherein, pt j Probability of j-th value of task time, mu 1 、θ 1 A probability threshold value divided for degree, and mu 1 >θ 1 ,μ 1 ∈[0.1,0.2],θ 1 ∈[0.005,0.01];
Wherein, pl j Probability of j-th value of task location, mu 2 、θ 2 A probability threshold value divided for degree, and mu 2 >θ 2 ,μ 2 ∈[0.1,0.25],θ 2 ∈[0,0.01];
Vr is a value of the task income, avg is an average value of values of all historical task income, and min is a minimum value of values of all historical task income.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211004808.8A CN115358781B (en) | 2022-08-22 | 2022-08-22 | Crowd sensing noise monitoring task recommendation method based on limited rational decision model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211004808.8A CN115358781B (en) | 2022-08-22 | 2022-08-22 | Crowd sensing noise monitoring task recommendation method based on limited rational decision model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115358781A CN115358781A (en) | 2022-11-18 |
CN115358781B true CN115358781B (en) | 2023-04-07 |
Family
ID=84002923
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211004808.8A Active CN115358781B (en) | 2022-08-22 | 2022-08-22 | Crowd sensing noise monitoring task recommendation method based on limited rational decision model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115358781B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2924623A1 (en) * | 2014-03-27 | 2015-09-30 | Korea Electronics Technology Institute | Context based service technology |
CN108038622A (en) * | 2017-12-26 | 2018-05-15 | 北京理工大学 | A kind of intelligent perception system recommendation user method |
CN109347924A (en) * | 2018-09-20 | 2019-02-15 | 西北大学 | A kind of recommended method based on intelligent perception |
CN111797331A (en) * | 2020-06-09 | 2020-10-20 | 安徽师范大学 | Multi-target multi-constraint route recommendation method based on crowd sensing |
CN112927037A (en) * | 2021-02-04 | 2021-06-08 | 中南大学 | Supplier recommendation method and system |
CN114722904A (en) * | 2022-03-08 | 2022-07-08 | 哈尔滨理工大学 | Sparse crowd sensing-oriented participant optimization selection method |
-
2022
- 2022-08-22 CN CN202211004808.8A patent/CN115358781B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2924623A1 (en) * | 2014-03-27 | 2015-09-30 | Korea Electronics Technology Institute | Context based service technology |
CN108038622A (en) * | 2017-12-26 | 2018-05-15 | 北京理工大学 | A kind of intelligent perception system recommendation user method |
CN109347924A (en) * | 2018-09-20 | 2019-02-15 | 西北大学 | A kind of recommended method based on intelligent perception |
CN111797331A (en) * | 2020-06-09 | 2020-10-20 | 安徽师范大学 | Multi-target multi-constraint route recommendation method based on crowd sensing |
CN112927037A (en) * | 2021-02-04 | 2021-06-08 | 中南大学 | Supplier recommendation method and system |
CN114722904A (en) * | 2022-03-08 | 2022-07-08 | 哈尔滨理工大学 | Sparse crowd sensing-oriented participant optimization selection method |
Also Published As
Publication number | Publication date |
---|---|
CN115358781A (en) | 2022-11-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108055119B (en) | Safety excitation method and system based on block chain in crowd sensing application | |
Yang et al. | On designing data quality-aware truth estimation and surplus sharing method for mobile crowdsensing | |
Li et al. | A distributed online pricing strategy for demand response programs | |
Swarup et al. | Project delivery metrics for sustainable, high-performance buildings | |
US11489682B2 (en) | Predicting utilization of a shared collaboration resource | |
Wong et al. | An agent-based negotiation approach to integrate process planning and scheduling | |
CN109068288B (en) | Method and system for selecting mobile crowd sensing incentive mechanism based on multi-attribute user | |
US20130317874A1 (en) | Meeting attendance planner | |
CN105096063A (en) | Bid evaluation expert management method and system thereof | |
Jin et al. | Data-driven pricing for sensing effort elicitation in mobile crowd sensing systems | |
KR20180068174A (en) | contractor selection server for carrying out at construction work and method | |
CN110097190B (en) | Crowd sensing task allocation method based on double time limits | |
Gong et al. | Truthful mobile crowdsensing for strategic users with private data quality | |
Khanfor et al. | A trustworthy recruitment process for spatial mobile crowdsourcing in large-scale social IoT | |
CN111565188B (en) | VANET trust model working method based on combination of message type and trust value confidence | |
Ye et al. | Incentivizing semisupervised vehicular federated learning: A multidimensional contract approach with bounded rationality | |
CN112101785A (en) | Method for analyzing comprehensive benefits of power and communication sharing iron tower | |
CN115358781B (en) | Crowd sensing noise monitoring task recommendation method based on limited rational decision model | |
CN115270782A (en) | Event propagation popularity prediction method based on graph neural network | |
WO2022216490A1 (en) | Intelligent scheduling using a prediction model | |
Zheng et al. | A team-based multitask data acquisition scheme under time constraints in mobile crowd sensing | |
Ding et al. | Optimal incentive and load design for distributed coded machine learning | |
Luo et al. | Parametric prediction from parametric agents? | |
CN113298668A (en) | Mobile crowd-sourcing aware user large-scale rapid recruitment method considering social network | |
Yang | Utility-based decision support system for schedule optimization |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |