CN114722904A - Sparse crowd sensing-oriented participant optimization selection method - Google Patents

Sparse crowd sensing-oriented participant optimization selection method Download PDF

Info

Publication number
CN114722904A
CN114722904A CN202210227653.8A CN202210227653A CN114722904A CN 114722904 A CN114722904 A CN 114722904A CN 202210227653 A CN202210227653 A CN 202210227653A CN 114722904 A CN114722904 A CN 114722904A
Authority
CN
China
Prior art keywords
participant
participants
sensing
data
perception
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210227653.8A
Other languages
Chinese (zh)
Inventor
葛惠杰
王健
赵国生
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin University of Science and Technology
Original Assignee
Harbin University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Harbin University of Science and Technology filed Critical Harbin University of Science and Technology
Priority to CN202210227653.8A priority Critical patent/CN114722904A/en
Publication of CN114722904A publication Critical patent/CN114722904A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
    • G16Y20/00—Information sensed or collected by the things
    • G16Y20/40—Information sensed or collected by the things relating to personal data, e.g. biometric data, records or preferences
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computing Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The perception cost and the perception quality are important points of sparse crowd sensing, and the existing sparse crowd sensing participant selection method selects a small number of participants with the largest coverage area according to the mobility of the participants, but ignores the willingness and the perception capability of the participants. In this respect, the invention firstly classifies the sub-regions based on the k-means algorithm by utilizing the space-time correlation among the sub-region perception data. Secondly, analyzing historical behavior information of the user to obtain multiple characteristics such as willingness degree, credit degree and task completion ability, and establishing a three-decision participant selection model. And finally, selecting the participants according to the model and the result of the sub-region classification. The method provided by the invention can deeply analyze the multiple characteristics of the participants, excavate the relationship between the participants and the target task and select the high-quality participants meeting the target task. Different data set experiment results show that the quality of the perception data is obviously improved by selecting an optimization scheme by the participator in the scheme, and the information quality is higher than that of other comparison methods.

Description

Sparse crowd sensing-oriented participant optimization selection method
Technical Field
The invention relates to the field of sparse crowd sensing, in particular to a participant optimization selection method facing sparse crowd sensing.
Background
With the rapid development of embedded devices, wireless sensor networks, internet of things, intelligent mobile terminals and the like, ubiquitous intelligent systems integrating sensing, computing and communication capabilities are being widely deployed and gradually integrated into the daily living environment of human beings, and the capability of ubiquitous computing for acquiring data is greatly enhanced. In this context, mobile crowd sensing is productive. The crowd sensing is a technology which utilizes a large number of mobile participants carrying intelligent equipment in a network space as basic units and cooperates through the Internet of things to realize the distribution of sensing tasks and the collection and utilization of sensing data, and finally completes large-scale and complex urban and social sensing tasks.
Perceptual cost and data quality are two key points of attention for crowd-sourcing awareness. In order to obtain high-quality perception results, crowd sensing generally requires the recruitment of a large number of participants equipped with rich sensing devices to participate in the perception task, aiming at improving the coverage of the participants in the perception area. However, in real-world situations the budget and number of participants are limited. In order to solve the problems, the royal music industry provides a concept of sparse crowd-sourcing perception, and the concept of sparse crowd-sourcing perception is used for remarkably reducing the number of required perception tasks while ensuring the data quality by utilizing the space-time correlation among data perceived by different subregions, so that the overall perception cost (such as energy consumption and excitation of a smart phone) is reduced. In short, sparse crowd sensing is to intelligently select a small part of a target region for sensing, and simultaneously perform high-precision inference on data of the remaining non-sensed region.
Under the budget limit, improving the perception data quality is the final goal of sparse crowd sensing. Participant selection is an important aspect that affects the quality of sparse crowd sensing data. However, most of the existing sparse crowd-sourcing perceptions focus on data reasoning, and have less concern for participant selection problems. Rana et al neglected participant recruitment and mainly used incomplete and randomly collected data to restore the complete sensorgram. HeandShin et al devised an incentive mechanism to guide participants in perceiving partitioned data with more value differences between the recent and current perception periods. But these studies directly neglect participant selection and assume that the selected partitions all have participants to collect perception data. In response to the deficiencies of the above documents, WenbinLiu et al proposed a participant selection strategy that takes into account the mobility of participants, in three phases of participant selection, partition selection and participant-partition cross-selection. The platform is directed to recruit the best set of participants (participant-partition cross-selection) by first selecting the set of candidate participants that covers the most sub-regions under budget constraints (participant selection), and then estimating from the selected set of candidate participants which sub-regions are more useful in data speculation (partition selection). This approach improves the perceived data quality under budget constraints, but ignores the quality of the selected participants and the potential relationship of the participants to the task.
Disclosure of Invention
The invention aims to solve the defects of the existing participant selection strategy in sparse crowd sensing, and provides a sparse crowd sensing participant optimization selection strategy based on three decisions. The method is characterized in that the space-time correlation and historical data correlation among the partitioned sensing data are fully utilized, a sub-region classification algorithm is realized, and the accuracy of participant selection is improved. In addition, the multi-features of participants (willingness degree, credibility degree and task completion capability) are fused, a three-decision participant classification model is established, and the misclassification cost is reduced, so that the goal of improving the perception data quality under the budget constraint is realized. The logical framework of the present invention is shown in fig. 1, and comprises the following steps:
and (3) classifying the sub-regions: the sparse crowd sensing only selects a small part of area to sense, and deduces the rest of the non-sensed area. The space-time correlation existing between the sub-region perception data enables data inference, and if only the participant selection problem is considered, the phenomenon that the participants are too concentrated can occur, so that the data inference is more complicated. In order to guide participant selection more accurately, the method is proposed to cluster the perception regions according to the spatiotemporal correlation between the historical perception data of the perception regions based on a k-means algorithm.
And (3) screening participants: the sparse crowd sensing selects the participants with the most coverage areas according to the mobility of the participants, although the sensing data quality can be improved, all the participants participating in the sensing task cannot be ensured to be high-quality participants, the enthusiasm of the participants cannot be ensured, and the problem that the data collection quality is low and even the task cannot be completed is caused. In order to improve the task completion rate and the perception data quality, the invention comprehensively considers the willingness degree, the credit degree, the task completion quality and other multiple characteristics of the participants and combines three decision-making theories to select a high-quality participant candidate set. To avoid perceptual redundancy, the platform chooses at most one participant in each class of partition to perceive.
Compared with the prior art, the invention has the beneficial effects that: based on the space-time correlation among historical perception data of the perception areas, the areas with similar perception values are divided into a class, and the selection of participants can be effectively guided. The method for optimizing and selecting the participants based on the three-branch decision is provided, under the guidance of a sub-region classification result, by analyzing the willingness degree, the credibility degree, the task completion capability and other characteristics of the participants, a participant selection frame is constructed by using a three-branch decision theory, the probability of the participants to complete the task is reasonably evaluated, the appropriate high-quality participants are selected to complete the perception task, the perception cost caused by misclassification is effectively avoided, and the quality of perception data is improved.
Drawings
FIG. 1 is an overall logical framework architecture of the present invention
FIG. 2 is a sub-region partition process diagram of the present invention
FIG. 3 is a three-branch decision model constructed by the present invention
Detailed Description
The present invention will now be described in more detail with reference to the accompanying drawings.
And (3) classifying the sub-regions: due to the fact that historical data correlation and space-time correlation exist between the sub-region sensing data, the fact that only a small part of regions are sensed and data of the non-sensing regions are inferred based on the correlation is made possible. Although the accuracy of participant selection is improved by using an existing method for guiding the participant selection by judging an effective region by using a deep Q network in reinforcement learning in sparse crowd sensing participant selection, the scheme requires that an intelligent agent continuously interacts with the environment to continuously improve the learned data, so that the time complexity is increased. According to the method, the characteristics of the historical data correlation and the space-time correlation of the perception data are combined, the areas with high similarity are classified into one class, the process of distinguishing the effective areas is effectively simplified, and the process of selecting participants is guided more accurately.
In sparse crowd sensing, classifying sensing areas to guide participant selection is a very important process. Extracting data correlations and spatio-temporal correlations between perceptual regions from historical perceptual data may optimize the process of participant selection. If participants who select sub-regions with greater similarity collect perception data, the difficulty of data inference is increased, and therefore the accuracy of the perception data is reduced. In order to improve the selection efficiency of participants and improve the accuracy of data reasoning, the invention provides a subregion classification algorithm.
Fig. 2 shows an example of sub-region classification, before each sensing task starts, a platform performs fine-grained division on a sensing region, and then calculates a similarity matrix by using data correlation and spatio-temporal correlation existing between data of a historical sensing period, where the similarity matrix is calculated as follows:
R(i,j)＝ρRt(i,j)+(1-ρ)Rhis(i,j) (1)
wherein R ist(i, j) represents the spatio-temporal correlation between region i and region j; rhis(i, j) represents a data correlation between the region i and the region j; ρ is a parameter that balances spatio-temporal correlation and data correlation. Time correlation between region i and region j:
Figure BDA0003536721930000021
where dis [ i, j ] represents the Euclidean distance between region i and region j. Data correlation between region i and region j:
Figure BDA0003536721930000022
wherein
Figure BDA0003536721930000023
Representing the perceptual data of sub-region i at period t;
Figure BDA0003536721930000024
representing the perceptual data of sub-region j at period t; num represents the last period of the current sensing period; since the influence of the sensing data of different sensing periods on the current sensing data is different, the longer the sensing data time is,the smaller the reference value. The correlation matrix is calculated herein taking only the perceptual data of the first four periods of the current perceptual period. And finally, classifying the sub-regions based on a k-means algorithm by combining a correlation matrix, and classifying the regions with higher similarity into a class in the process.
And (3) selecting the participants: sparse crowd-sourcing perception only needs to recruit a small amount of appropriate high-quality participants to perceive partial sub-regions to collect perception data, then a fusion data reasoning algorithm carries out high-precision inference on the data of the rest non-perceived regions by utilizing the space-time correlation among the sub-region data, and the perception cost is reduced while the data quality is improved. Improving the quality of perceptual data is the ultimate goal of sparse crowd sensing. In sparse crowd sensing, there are many factors that affect the quality of the sensed data, such as participant quality, region selection, and task allocation, among which the most direct factor is the quality of the participant who provided the collected data. Therefore, the selection of the participants is particularly important in the whole sparse crowd sensing process. Previous participant selection problems only consider the participant's coverage of the perception area and neglect participant quality and participant preferences. Ignoring the quality of the participant may result in poor quality of the collected data, whereas ignoring the participant preferences may result in poor enthusiasm or even incomplete task completion by the participant. To avoid the above, the present document extracts participant multi-features from the historical behavior of the participant.
The ultimate goal of the participant selection problem is to pick the participants for those active sub-regions to collect perception data and use these collected perception data to infer data for the non-perceived regions. Only by selecting proper high-quality participants to participate in the perception task, the perception data meeting the requirements of the perception task can be collected. Therefore, the selection of a proper sensing participant with high quality is a precondition for completing the sensing task, and is a key point for guaranteeing the quality of collected data. The invention comprehensively considers the willingness degree of participants and the optimization method of the participants.
Fig. 3 shows a specific process of constructing the three-branch decision model. Firstly, extracting multi-feature information such as willingness degree and credibility of a participant and capability of the participant to complete a task according to historical behavior information of the participant, wherein the willingness degree of the participant is related to task position, task content and task time. The sparse crowd sensing tasks are all location dependent, and participants prefer tasks within their acceptance range. Second, the task content that the participant is interested in can affect his preferences, and whether the participant is in an idle state can also affect how willing the participant perceives the task. The credibility of the participants represents the rate of the participants completing the perception tasks, and is an important mark for measuring whether the participants can complete the perception tasks or not; in addition, the ability of the participants to complete the sensing task represents whether the participants can complete the sensing task with high quality, and is the most direct factor influencing the sensing quality. And secondly, calculating the probability of the participant being valuable to the task according to the fusion of the multiple characteristics of the participant. Finally, two thresholds with three decisions are compared. According to the probability that the participants are 'valuable' to the perception task and decision region division threshold values alpha and beta and the result of region classification, dividing a candidate participant set into a positive domain, a boundary domain and a negative domain, accepting the users in the positive domain to join a candidate participant queue, and rejecting the users in the negative domain to join the candidate participant queue, and for the users who can not make a decision in the boundary domain, performing secondary classification by adjusting a multi-feature ratio to obtain the result selected by the final user.
The willingness w of the participant is as followsiThe calculation is as follows:
Figure BDA0003536721930000031
wherein, wpRepresenting the relationship between task location and participant willingness:
Figure BDA0003536721930000032
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003536721930000033
represents participant uiSense of participationAwareness task vjThe distance to be moved is expressed by Euclidean distance;
Figure BDA0003536721930000034
covering the radius for the task; w is ai,jRepresents participant uiFor task vjThe probability of interest. Set x (v)j) Set of attributes, x (u), representing perceptual tasksk) Representing a set of participant preference attributes, participant uiFor task vjThe probability of interest is:
Figure BDA0003536721930000035
Figure BDA0003536721930000036
represents participant uiParticipating in a perceptual task vjIn a time ratio of (1), wherein
Figure BDA0003536721930000037
Which indicates the end time of the task,
Figure BDA0003536721930000038
indicating the start time of the task. The larger the time fraction, the more aggressive the participant is in participating in the perception task.
We combine the ability of a participant to complete a task with the reputation of the participant, collectively referred to as trustworthiness. Degree of reliability
Figure BDA00035367219300000312
The calculation formula of (a) is as follows:
Figure BDA0003536721930000039
wherein the content of the first and second substances,
Figure BDA00035367219300000310
is participant uiThe number of the sensing tasks to be completed,
Figure BDA00035367219300000311
is participant uiThe number of all the perceptual tasks that are accepted,
Figure BDA0003536721930000041
is participant uiParticipating in a perceptual task vjThe time-dependent data collected is used as the time-dependent data,
Figure BDA0003536721930000042
is participant uiParticipating in a perceptual task vjReal sensing data of the ground. Probability of participant being valuable for perceiving task
Figure BDA0003536721930000043
Figure BDA0003536721930000044

Claims (4)

1. A sparse crowd sensing-oriented participant optimization selection method is characterized by comprising a sub-region classification module and a participant screening module.
2. The sparse crowd sensing-oriented participant optimal selection method according to claim 1, wherein the whole sensing area is classified through a sub-area classification module, a similarity matrix is calculated according to the time-space correlation and historical data correlation among the partitioned sensing data, and a sub-area classification result is formed based on a k-means algorithm according to the similarity matrix, so that the accuracy of participant selection is improved.
3. The sparse crowd sensing-oriented participant optimization selection method according to claim 1 is characterized in that the multiple characteristics of participants (willingness degree, credibility degree and task completion capability) are fused, a three-decision participant classification model is established, a high-quality participant candidate set suitable for a target sensing task is selected, the misclassification cost is reduced, and the sensing data quality is improved.
4. The sparse crowd sensing oriented participant optimal selection method of claim 1, wherein consideration is given to using spatiotemporal characteristics of sub-regions to guide the process of participant selection.
CN202210227653.8A 2022-03-08 2022-03-08 Sparse crowd sensing-oriented participant optimization selection method Pending CN114722904A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210227653.8A CN114722904A (en) 2022-03-08 2022-03-08 Sparse crowd sensing-oriented participant optimization selection method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210227653.8A CN114722904A (en) 2022-03-08 2022-03-08 Sparse crowd sensing-oriented participant optimization selection method

Publications (1)

Publication Number Publication Date
CN114722904A true CN114722904A (en) 2022-07-08

Family

ID=82237770

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210227653.8A Pending CN114722904A (en) 2022-03-08 2022-03-08 Sparse crowd sensing-oriented participant optimization selection method

Country Status (1)

Country Link
CN (1) CN114722904A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115358781A (en) * 2022-08-22 2022-11-18 陕西师范大学 Crowd sensing noise monitoring task recommendation method based on limited rational decision model

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115358781A (en) * 2022-08-22 2022-11-18 陕西师范大学 Crowd sensing noise monitoring task recommendation method based on limited rational decision model
CN115358781B (en) * 2022-08-22 2023-04-07 陕西师范大学 Crowd sensing noise monitoring task recommendation method based on limited rational decision model

Similar Documents

Publication Publication Date Title
CN104881800B (en) A kind of incentive mechanism implementation method based on mobile intelligent perception
CN110210903B (en) Crowd sensing excitation method based on user position and credit
CN111339818A (en) Face multi-attribute recognition system
CN112801411B (en) Network traffic prediction method based on generation of countermeasure network
Jiau et al. Services-oriented computing using the compact genetic algorithm for solving the carpool services problem
Gong et al. Adaptive client clustering for efficient federated learning over non-iid and imbalanced data
CN112508292A (en) Intelligent charging station site selection optimization method based on fuzzy TOPSIS method
CN114722904A (en) Sparse crowd sensing-oriented participant optimization selection method
Jiang et al. Federated learning based mobile crowd sensing with unreliable user data
CN114520743A (en) Method and system for detecting network abnormal flow and storable medium
Lin et al. Multi-level cluster-based satellite-terrestrial integrated communication in internet of vehicles
CN108960486A (en) Interactive set evolvement method based on grey support vector regression prediction adaptive value
CN109325177B (en) Information pushing method, system, readable storage medium and computer equipment
CN114093419A (en) RBP binding site prediction method based on multitask deep learning
Gong et al. Adaptive interactive genetic algorithms with individual interval fitness
Gao et al. Deep learning based location prediction with multiple features in communication network
CN108776863A (en) One kind being based on the maximized intelligent perception motivational techniques of user base number
CN116939866A (en) Wireless federal learning efficiency improving method based on collaborative computing and resource allocation joint optimization
CN111405605A (en) Wireless network interruption detection method based on self-organizing mapping
CN108631895B (en) Spectrum sensing secondary user excitation method based on Bayesian game
Xu et al. PSARE: A RL-Based Online Participant Selection Scheme Incorporating Area Coverage Ratio and Degree in Mobile Crowdsensing
CN114971047A (en) Mobile crowd sensing-oriented user collaborative optimization method
CN115545960A (en) Electronic information data interaction system and method
CN114268625B (en) Feature selection method, device, equipment and storage medium
CN115002159A (en) Sparse crowd sensing system-oriented community classification and user selection method

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