Research On Classification Of Dialysis Pressure Curve Based On Functional Data

With the rapid development of information technology,data collection and recording have become more intensive and continuous.Traditional discrete data analysis methods have faced many limitations in information mining in the era of big data.For medical data such as hemodialysis,the patient's measurement indicators are not just traditional discrete data.If the observed data is treated as discrete data without considering the potential intrinsic function characteristics of the data,it may cause problems such as loss of important information.Functional data analysis is a continuous function as the object of analysis,which can grasp the distribution characteristics and changing laws of the data as a whole.It not only relaxes the structural constraints and distribution assumptions of the data,but also allows static and dynamic analysis from the perspective of functions.This article analyzes the dialysis data from the perspective of function.First,the data is preprocessed.By comparing the advantages and disadvantages of three common basis function expansion methods in the smooth fitting of discrete data,choose the B-spline basis function expansion method to convert discrete data into functional data with continuous values in the time interval,and realize the functionalization of discrete data;Secondly,in order to remove the interference of abnormal curves on the classification results,this paper proposes the data processing of abnormal curves based on the basic idea of functional data depth.Then combining the functional data non-parametric k-nearest neighbor(knn)classification method to calculate the posterior probability value of a given curve belonging to each category,and classify the unknown curve according to the classification rule with the largest posterior probability.Finally,the traditional functional non-parametric knn classification method and the non-parametric knn classification method based on functional depth proposed in this paper are applied to simulated data and instance data.Among them,the simulation data analysis results show that the average error rate of the 10 groups of simulation functions without abnormal curves is not much different.Among the 20 groups of simulation functions with abnormal curves,the classification error rate based on the depth removal of abnormal curves is significantly lower;The example mainly analyzes the dialysis pressure data of a kidney disease patient in a hospital in Sichuan Province.The results show that the non-parametric classification knn method based on the function data depth removal of abnormal curves proposed in this paper performs relatively well,and the accuracy of coagulation classification is significantly improved.