Compressed Sensing Reconstruction And Rough Set Attribute Reduction And Their Application

With the rapid development of information technology,data in various fields are growing explosively.These data contain a large number of redundant data and missing data.These data will have an impact on our correct decision-making.Therefore,it will become more urgent for us to process these data.Compressed sensing can reconstruct the original signal accurately with low sampling rate,and also reconstruct the missing data,but its reconstructing accuracy is affected by the selection of candidate sets.Rough set theory can achieve good application results for redundant attributes,but its performance is affected by partition of knowledge particles.Therefore,in order to improve the accuracy of reconstruction,the S-type function is used to improve the Stagewise Weak Orthogonal Matching Pursuit(SWOMP)reconstruction algorithm.To improve the performance of attribute reduction,the Neighborhood Rough Set(NRS)attribute reduction algorithm based on Spielman correlation coefficient is studied.These algorithms are applied to oil logging.The main work or innovations are as follows:(1)Comparative analysis based on base pursuit and orthogonal matching pursuit algorithms.Since the collected or collected data is affected by the noise such as missing values,the deviation of the decision results often occurs.For the same reconstruction precision,find an algorithm that performs more efficient execution of missing value reconstruction.And the base pursuit and orthogonal matching pursuit are compared.The results show that the missing value reconstruction using the orthogonal matching pursuit algorithm is faster and more stable.(2)Research on the Stagewise Weak Orthogonal Matching Pursuit reconstruction technique improved by S-type function.In view of the fact that SWOMP adopts fixed threshold parameters in iteration,it is easy to underestimate or overestimate the candidate set.It is inspired by the rule of “fast approach in the beginning stage and stepwise approximation in the final stage”.It is proposed to improve the threshold setting by using S-type function in iteration.Through experimental analysis and comparison,the reconstruction accuracy is improved.(3)Research on attribute reduction of Neighborhood Rough Set(SCCNRS)based on Spearman correlation coefficient.In order to construct a distance measure more suitable for similarity between samples,according to the rule of " small spacing within the class and large spacing between classes ",Spielman correlation coefficient is used to improve the euclidean distance measure in neighborhood rough sets and determine the importance of its attributes.Finally,k-Nearest Neighbor algorithm(KNN),Gaussian Naive Bayes and Support Vector Machine(SVM)are used to verify the accuracy of attribute reduction.Through the simulation analysis of UCI data,the effectiveness and superiority of SCCNRS attribute reduction are proved.(4)Practical application research on classification and identification of oil and gas layers.In order to improve the accuracy of oil and gas layer classification and identification,a classification model based on segmented weak orthogonal matching tracking reconstruction improved by S-type function and SCCNRS attribute reduction technology was established.The application of oil and gas layer data in an X well logging shows that by using SWOMP improved by S-type function to reconstruct missing values and using SCCNRS to reduce redundant attributes,the accuracy of classification and recognition is improved,which has a wide application prospect.