Functional Nonparametric Regression Model And Its Application In Medical Data

Functional data analysis analyzes ordered data with continuous changes recorded at high frequencies as a whole function,and has wide applications in fields such as biology,medicine,and economics.Among them,the functional data regression model is an important method for predicting and classifying functional data.This article mainly studies functional non-parametric regression models and their applications in medical data.This article firstly converts the original discrete time series data into functional data based on the basis function and rough penalty method;Then,for functional data with scalar response variables,a non-parametric regression model is applied for prediction and effective classification is performed using the differences in prediction results.In the simulation experiment,we generate curve data with mean difference,structure difference and obvious interaction,respectively,to verify the effectiveness of the classification method.The analysis of dialysis pressure data shows that the accuracy of real-time coagulation warning using rough penalty method and first-order derivative combined with sliding window is significantly better than existing general warning methods.For the problem of outlier detection of functional data whose response variable is a function,a non-parametric regression model is used to predict the curve,and a method of anomaly curve detection is established according to the difference of changes before and after the curve.Firstly,the simulation data with shape outlier and position outlier are used to verify the ability of the method to identify abnormal curves;Then,this method was applied to the recognition of abnormal curves in electrocardiogram data,and the average classification accuracy achieved was 98.13%.The results showed that this method can effectively identify patients with cardiac abnormalities and has high application value.In this paper,the original discrete time series data is converted into functional data.For functional data whose response variables are scalars and functions,nonparametric regression models are used to predict,and classification and outlier detection rules are designed.Finally,this rule is applied to medical data for classification and anomaly curve detection,achieving high accuracy.