| In recent years,with the widespread popularization and application of big da-ta,the value of data has been gradually gained attention and recognized,and the demand for data trading is also increasing.Data trading platform is an important carrier of data trading behavior,which can promote the integration of data resources,standardize trading behavior and reduce transaction costs.It has become one of the main measures to promote the circulation of data in various regions.However,from the overall development level of the data platform,the current big data trading is still face many problems:(1)the asymmetry of data supply and demand makes it difficult for the data trading to meet the effective needs of the society,and the data trading rate and trading amount are not high.(2)the data opening process is slow.The phenomenon of "data island" restricts the overall scale of data trading to some extent and affects data monetizing.(3)data pricing problem cannot be ef-fectively solved in data trading process.In view of the above problems,this thesis proposes to apply mobile sensing technology in data trading market to solve the im-balance between supply and demand of data resources.However,the construction of data trading market based on mobilesensing still faces challenges that traditional mobile sensing and data trading platform have never encountered.In the mobile sensing,population is required to participate in the sensing task to collect and share data,which brings more opportunities and challenges to data collection.Low qual-ity of data,difficulty to measure the value of data and the lack of effective data pricing rules have become unfavorable factors restricting data trading and circula-tion.Because of this,how to choose the right participants to provide high-quality sensing data and how to set a reasonable price for the collected data are important challenges which the mobile sensing data market should face and solve as soon as possible.This paper conducts in-depth research on the selection of participant and data pricing in the mobile sensing data market,and makes the following four major contributions:In order to effectively select the participants to achieve high-quality data col-lection,this thesis proposes two effective strategies to solve the participant selection problem in constrained and unconstrained environments respectively:(1)We proposed a participant selection model based on spatial coverage.In order to maximize the benefits of data platform in the context of limited budget,starting from the predictable mobility of sensing participants,spatial coverage is proposed as an important measurement index for the quality of data contributed by participants,so as to build a participant selection model based on spatial cover-age.Therefore,a novel greedy genetic algorithm is designed to get the best solution.Experiments on real and simulated data sets show that the proposed algorithm can obtain more high-quality data with limited budget.(2)We proposed the participant selection model based on marginalism princi-ple.Firstly,the combination of reputation and willingness to participate is taken as a measure to evaluate the quality of sensing data,and it is modeled as the quality of service(QoS)of participants to build multiple benefit-cost models.Secondly,to maximize the overall benefits of the data platform,this paper draws lessons from the principle of marginalism in economics and hopes to select a new participan-t only when the marginal benefit of a newly added participant is higher than the cost of payment.In this thesis,the scheme is modeled as the marginal problem of sensing market participant selection.In order to solve the problem effectively,a greedy random adaptive search algorithm(GRASP-AR)with annealing mechanis-m is proposed to maximize the benefits of sensing data.Finally,the effectiveness of the proposed algorithm in the participant selection can be clearly verified by a large number of experimental evaluations on real and sensing data sets.In view of how to make a reasonable price for data,this thesis also proposes two effective data pricing models:(3)We proposed data pricing model based on quality level.In this thesis,a weighted linear data quality score model is firstly proposed,which includes 14 da-ta quality standards,most of which can be evaluated by calculation.Based on the score for each quality criterion,the overall quality score for the dataset can be calcu-lated.Secondly,in order to realize automatic grade differentiation of data products,this thesis discretizes quality scores into quality grades and proposes a utility mod-el based on quality grades.From the perspective of data science,the rationality of candidate utility function is verified.In order to take advantage of consumer-s' willingness to pay,allow data platforms to sell data products to data consumers by adjusting data quality.Finally,this thesis also constructs a profit maximization model of data platform and deduces the optimal data pricing according to KKT conditions.(4)We proposed an anonymous personal data pricing model based on sensi-tivity level.Personal data has a strong privacy attribute.Therefore,this thesis first proposes a fair compensation scheme for information loss.From the perspective of data providers,they can get corresponding compensation according to their sensi-tive attitudes.From the perspective of data market,their operating costs have been effectively controlled.Secondly,a nonlinear mathematical model is proposed to describe the relationship between consumer self-selection behavior and the utility of personal data.Finally,we developed a bi-level optimization model with a data platform as the leader and consumers as the followers to maximize the profit of the data market.The data platform collects personal data from the private-conscious population and provides subscription services to the data consumers.Profit maxi-mization is achieved by optimizing the optimal purchase cost and data sensitivity level. |
PDF Full Text Request