Weather Monitoring Data Processing Based On Matrix Completion

How to solve the contradiction between the limitation of energy in wireless sensor network(WSN)and the energy consumption and the network congestion leaded by the large number of data transmission is always the hot spot for WSN.The development of matrix completion theory provides new approaches for data gathering in WSN.The existing matrix completion algorithms for WSN only consider how to reduce the sampling number without considering the real-time performance when recovering the data matrix.In order to guarantee the recovery accuracy and reduce the recovery time consumed simultaneously,we use the newest matrix completion algorithms to recover the weather monitoring data.Only to esure the difference between the recovered data and original data is little as much as possible thus system can get the right conclusion through these data and then guides users to make right dicision.Moreover,long recovery time will delay the data analysis results which will lead to bad user experience.Therefore,it is very necessary to consider the time cost and the accuracy when we use matrix completion to saving the system energy.This paper constructed a energy-saving weather monitoring system based on matrix completion,and researched the Maxide algorithm and the augmented Lagarangian multiplier menthod(ALM)for weather data recoverying.A large amount of experiments have been.carried out to investigate the performance of these matrix completion algorithm for weather data under different sampling ratio,diffrenr sampling models and different parameters setting from the relative residual error,the relative error and the time cost.The main contribution of this paper are as follows:Firstly,aiming at that the existing data acquisition schemes based on matrix completion in WSN systems do not consider the real-time performance of algorithm,this paper introduce the Maxide algorithm and the ALM agorithm into the weather monitoring system saveing energy by close some nodes,and study the data recovery performance as the same time.The experiment results show that the currently used singular value threshold algorithm(SVT)and MC-Weather algorithm are 5 to 16 times as long as using Maxide matrix completion algorithm to complete the weather monitoring data.And using ALM completion algorithm to complete the weather monitoring data is 1/12 to 1/5 times as long as using singular value threshold algorithm and MC-Weather algorithm.Furthemore compared with using SVT algorithm and MC-Weather algorithm,using ALM algotithm to complete weather monitoring data can decrease the relative error by 1%~6%.And in view of the performance of the matrix completion algorithm influced by the sampling rate,the sampling model and parameters settings of the algorithms,we change a single factor in every group experiments and repeate them manny times to test the performance of the Maxide algorithm in weather monitoring data.Secondly,through analysing the relationship between parameters and the recovery relative error and the real-time performance in ALM algrithm,we obtain a rule of parameters and the recovery performance:For the same observation matrix,the convergence relative residual error and the convergence relative error will decrease with the drop of the value of the parameter ?,while the iterations needed for convergence increase.Conversely,the convergence relative residual error and the convergence relative error will increase with the up of the value of the parameter ?,while the iterations needed for convergence decrease.And an adaptive parameter adjusting algorithm is designed based on it.The experimental results show that this algorithm can well balance the recovery accurancy and time cost of augmented Lagrange method.It makes the completion speeds of ALM algorithm up to 1.6 times,while the accuracy only dropped 2 percentage points.