Local Validity Of Big Data Mining And Medical Application Research

With the rapid development of information technology and the rapid growth of data volume,big data exhibits multi-source,heterogeneous,multi-modal and other characteristics,which seriously affects the time efficiency of big data mining.Big data mining experts believe that sampling big data,big data analysis through local samples will lose the meaning of big data itself.While from a statistical point of view it is accepted that speckled and known whole seals lead to big data mining via local samples,which can improve the time efficiency of big data processing.In this paper,a local sampling experiment is conducted on large data sets to explore the local effectiveness of big data mining.This paper theoretically discusses the local effectiveness of big data by making big data samples evenly distributed through random scrambling algorithm to facilitate sampling of big data first.After evenly scrambling the bank marketing data set at UCI,local samples were drawn to scale.We use the Apriori algorithm for rule mining on multiple sets of local samples,the Support and Confidence of the association rule for longitudinal contrast,and the sampling error between lateral contrast and some algorithms such as Hash and KDS.The results show that local samples under random uniform distribution can effectively reflect the big data overall situation.At the same time,the local effectiveness theory is applied to a Dn Cnn deep learning based case in medical image denoising.The large number of image slices generated after preprocessing of the CT images of the lungs of COVID-19 were experimentally randomly sampled after uniform scrambling,and the model was trained.Contrasting the different training models,the results show that the control model of the local samples has a denoising effect close to that of the full training set model.In addition,it is also effective in protecting the image details and lesion characteristics of the lungs while well removing the COVID-19 CT image noise.The local effectiveness of big data,that is,local can embody global features,which could improve the time efficiency of big data mining.There exist a certain practical value.