Research And Implementation Of Improved WSNs Incomplete Data Filling Algorithms And Its Parallelization

Wireless Sensor Networks(WSNs)is a hotspot of information technology research.Due to the lack of WSNs data,it is difficult for further data analysis and processing.The current solution is not to deal with it,or to fill incomplete data in WSNs network with some traditional methods.However,the traditional filling method has the problems of low accuracy and long calculation time.The thesis studies the filling algorithm of incomplete data in WSNs network and its parallelization,trying to improve the filling accuracy and speed at the same time.The first goal of the thesis is to improve the accuracy of incomplete data filling in WSNs.To overcome the shortcomings of existing Mahalanobis distance filling algorithm(MKNN filling algorithm)which does not fully consider the strong correlation between distance and weighting coefficient and WSNs data,the thesis improves the calculation method of variation degree and weighting coefficient between neighbor variables,and proposes a new Mahalanobis distance filling algorithm,namely NMKNN filling algorithm.Through theoretical analysis and simulation experiments on WSNs data,the predicted value of this algorithm is closer to the original value than that of Mahalanobis distance filling algorithm,which shows that the accuracy of incomplete data filling is improved.The second goal of the thesis is to improve the speed of incomplete data filling in WSNs.This goal is achieved by parallelizing the filling algorithm of NMKNN.According to the characteristics of WSNs data set,the data set is divided into columns and rows,and the data set is filled in the cluster by data parallel method,which improves the speed of data processing and reduces the time of incomplete data filling.Through theoretical analysis and simulation experiments on WSNs data sets,compared with the serialization algorithm,the parallel algorithm proposed in the thesis improves the filling accuracy and reduces the computing time,and increases the running speed by 64.96%.It shows that the proposed algorithm is feasible and effective.