Pattern-Mining-Based Redundancy Elimination Mechanism In WSN

With the rise of the Internet of T hings(Io T),Wireless Sensor Network(WSN)has been continuously research ed and development.WSN is a kind of application oriented network.It contains a large number of densely deployed sensor nodes,which makes the neighbor's data similar and cause a large amount of spatial redundancy.Moreover,Sensor nodes in WSN are usually gathering data periodically and frequently,which makes the recent data similar and produce a large amount of temporal redundancy.Temporal and spatial redundancy waste the limited energy,increase the communication interference and transmission delay.Therefore,how to eliminate the temporal and spatial redundancy is very important for WSN.In this article,based on correlation feature of data,we designed a novel temporal and spatial redundancy elimination mechanism to eliminate the redundancy in WSN.Firstly,by taking advantage of both linear and periodic feature of phenomena,we proposed a Pattern-Statistic-Based temporal redundancy elimination mechanism(PSB).T his mechanism statistics the appearance frequency,appearance time and appearance duration of every linear pattern,and calculates the Weighted Frequency of them,which indicate the linear pattern's appearance possibility in the future.Based on the statistical information,and by combining Least Square method,PSB can get a most probable linear pattern as the prediction model,and use it to predict the future data.If the difference between real data and prediction data less than threshold e,sensor node can suppress the sending,and Sink just use the prediction data.By this way,the temporal redundancy can be eliminated,and the data accuracy can be ensured.Compared with previous methods,PSB can utilize not only the linear phenomena,but also the periodic phenomena,which improve the prediction success rate and accuracy,and effectively reduce the sending times and the energy consumption of sensor node.Simulation result s show that: compared with the other two methods,the performance of PSB has obvious improvement.Secondly,based on the phenomenon that the data differences between neighbor nodes are usually constant,we proposed a method to min e the constant correlation model between nodes.And then we propose d a Correlation-Model-Based Data Aggregation Algorithm(CMB)using the above method.CMB dig s the correlation model of every pair of neighbor nodes,and draw s a correlation graph of the network.Then CMB establish a shortest path routing tree for each connected component of the correlation graph.In the establish process,CMB inform each node of the correlation model information.After that,When the node receives the children's data,if the children's data and its own data can fit the correlation model,CMB will suppress the children's data,which can eliminate the spatial redundancy and achieve the goal of data aggregation.When the aggregation data reach the Sink node,CMB can restore the suppressed data within a specified threshold.The simulation based on real temperature data shows that,CMB algorithm has excellent performance on aggregation degree and data accuracy.