The Research Of EM Algorithm In Incomplete Monitoring Data Processing

As we all konw,when we are surveying and mapping it will be affected by terrain conditions,weather,environment,human factors,etc.These factors often result in a lack of observational data or a coarse difference and make the observational data incomplete.Now,most of the data processing methods for deformation monitoring are based on complete data,Without dealing with the absence of data,the accuracy of the results will be often affected.When the data is missing from the data,the deletion method,the normal filling method,the fitting process or predictive method are used to deal with the missing data,then using conventional methods to modeling analysis.But these approaches have certain limitations.Delete method implementation is simple,and fast but it led to the waste of resources,when the missing data have a lot of or in a more important position,the method may lead to the result error.The normal filling method,fitting process and prediction method can improve the quality of the deformation monitoring data to a certain extent,but the result may not be the optimal result.Because all of these approaches have to fill in the missing data then the modeling analysis is predicted.Because of the limitations of these approaches,there is a certain amount of uncertainty in the data that it fills,using it for modeling analysis can lead to deviations in results.Based on the above situation,this paper was chosed EM algorithm which are classical algorithm in the statistical domain,after analyzing the mechanism of missing data and the processing methods of missing data the method of dealing with deformation monitoring data by using EM algorithm was discussed.Aiming at the research of EM algorithm in incomplete monitoring data processing,the following work is done in this paper:(1)This paper discusses the data processing methods commonly used in deformation monitoring,summarizes and compares these methods,and analyzes the applicable situation,advantages and disadvantages of various methods in mapping data processing.(2)According to the mechanism of missing data,the methods of dealing with incomplete measurement data were introduced,and the applicability of various methods is analyzed by comparing various missing data processing methods.(3)The principles and properties of EM algorithm are introduced,and the steps of EM algorithm and the commonly used prediction model AR(P)model are introduced in detail to deal with incomplete monitoring data.By comparing the prediction effects of the deletion method and the regression filling method in the case of single deletion and multiple deletions,the reliability of the EM algorithm in the incomplete data processing of deformation monitoring is proved.(4)In the case of complete data were used to GM(1,1)grey model and BP neural network model to predict the settlement data,comparison of data in a single deletion and multiple deletion situations using EM algorithm to estimate the AR(p)prediction model.Through comparison and analysis,it is found that the prediction effect of the 4 methods is not same,and the prediction accuracy of the AR(p)model estimated by EM algorithm is more higher.