Tackling Data Sparsity For Truth Discovery In Location-aware Mobile Crowdsensing

In recent years,due to the widespread popularity of the mobile devices equipped with various sensors,we can collect sensing data of people's surrounding environment,like the moisture,the temperature and the traffic conditions and so on.Hence the concept of mobile crowdsensing was introduced,it refers to a practice of using people's smartphones to collect environmental data,then upload it to a data processing center,thus to execute some sensing tasks collaboratively.Among the data collected from multiple contributors,inconsistency often occurs due to noise,different sensor precision,or contributors' heterogeneous behaviors.To tackle the inconsistency,truth discovery has been widely studied to jointly infer the underlying ground truths and the contributors' data qualities.Existing truth discovery algorithms are generally based on the aggregation of large amounts of data so as to achieve accurate estimations.That is,there are large enough data contribution at any point of interest(PoI).In practice,the collected data are usually sparsely distributed among a large sensing area,where each PoI may receive only a few reports.In this case,traditional algorithms may not provide an accurate truth estimation for each PoI without the support of mass data.To tackle this challenge,we propose an effective truth discovery method,namely Holmes,which takes advantage of the spatial correlations of the monitored phenomena by reusing each contributor's data for multiple nearby PoIs,so it solves the problem caused by the data sparsity.We also take the issue of long-tail data phenomenon,which means most participants usually provide few sensing data.However,the more data a participant has,the more accurate we will be to estimate his reliability.So we propose Holmes-LT to improve the accuracy of the estimation of contributors' data quality levels.In the online streaming data scenarios,it is inefficient to infer the truths by continuous iteration.We further propose Holmes-OL to realize accurate truth estimation with lower time complexity.We evaluate the performance of our algorithms on both real and synthetic datasets.The results demonstrate that our algorithms achieve significant performance improvements in terms of estimation accuracy over the existing algorithms.Besides,it has strong scalability and a great improvement in efficiency.