Research On Missing Data Reconstruction Algorithm In Wireless Sensor Network

With the development of communication and sensor technology,Wireless Sensor Network(WSN)has become the focus of the industry,and has been widely used in many fields,such as military defense,biomedical,environmental monitoring and so on.Obtaining complete and accurate perceptual data through WSN is the basic premise of data analysis and application decision.However,the sensor data loss is inevitably duing to environmental factors and the limitations of sensor nodes' resources.Besides,most of the existing missing data reconstruction algorithms are based on a single data attribute or only considering static scenes,which are difficult to meet the requirements of reconstruction of real and complete data.Therefore,on the basis of studying the data characteristics of WSN and the related missing data reconstruction algorithm,this thesis discusses how to efficiently reconstruct the loss data of the sensor nodes by combining the spatio-temporal and multi-attribute correlations of the sensor data.In view of the fact that most of the existing missing data reconstruction algorithms are based on single attribute data,we improve a multi-attribute missing data reconstruction algorithm based on spatial correlation by combining the multi-attribute correlation of node data in the same cluster and the data's characteristic of joint low rank.The algorithm firstly applies the K-means clustering algorithm to divide the nodes into different clusters according to the physical location of the sensor nodes,then decomposes the data of the nodes in each cluster.Besides,the weighted nuclear norm minimization theory is used to assigned different weights to each singular value adaptively,and then the global optimal solution is obtained by the Alternating Direction Method of Multipliers algorithm.Finally,the data of the Berkeley laboratory is used to verify that the proposed algorithm reconstruct missing data well both in the mode of continuous and random missing data.In view of the fact that most of the sensor nodes in the current reconstruction algorithms are in static scenes,we improve a missing data reconstruction algorithm in mobile scenes based on Bayesian network model.The algorithm includes pre-data collection,optimal trust node determination and subsequent missing data reconstruction methods based on Bayesian network model.If a node finds its data is missing during the data collection phase,it first requests data from the neighboring nodes,and determines the best candidate node by evaluating spatio-temporal correlations,trajectory behavior,quantity and quality of data,and the number of hops traveled by the receive data from the source.Then we establish a Bayesian network model at the node,and the data of the best candidate node is introduced as an auxiliary variable to determine the range of missing data.After that the conditional probability of each value in this range is calculated and the data corresponding to the maximum probability is selected as the reconstruction data of missing data.Finally,we use the data collected by the University of Melbourne to verify that the proposed algorithm can reconstruct the missing data better when the node is under the moving scene,and the stability of the algorithm is higher than other similar algorithms.