Research And Implementation Of Asthma Diagnosis Model Based On Improved Fuzzy Support Vector Machine

Asthma is a chronic airway disease characterized by airflow limitation caused by narrowing of the airway,thickening of the airway wall cavities,and increased mucus.Although there are detection techniques for asthma in the medical field,such as blood gas analysis and lung function tests,most of these tests are time-consuming and laborious and have problems such as missed diagnosis and misdiagnosis.With the rapid development of artificial intelligence,the application of information technology to the detection and treatment of asthma is gradually becoming a trend.This thesis uses blood routines commonly used in asthma detection as input data to try to construct an asthma diagnosis model based on improved fuzzy support vector machine to improve the accuracy of asthma diagnosis.First,a feature selection method based on information theory is proposed to screen out the subset of features that are most relevant to asthma from blood routine indicators;then an asthma diagnosis model based on improved fuzzy support vector machines is proposed to realize the diagnosis of asthma;Finally,design and develop a medical system that assists doctors in making asthma diagnosis decisions to facilitate doctors’ treatment and recording of asthma patients.Main tasks as follows:(1)Aiming at the problem that the existing feature selection algorithm DCSF(Dynamic Change of Selected Feature)cannot effectively eliminate the bias of conditional items to high redundancy features,a dynamic weighted feature selection algorithm based on minimum redundancy,MR-DCSF(Minimal Redundancy-Dynamic Change of Selected Feature)is proposed.the conditional relevance of the selected subset is inversely proportional to the redundant information between the candidate feature and the selected subset.Using this feature,the problem of minimizing redundancy is transformed into a problem of weight distribution of condition-related items.The MRDCSF algorithm proposes two dynamic weights based on information entropy and condition relevance.By increasing the proportion of selected feature condition related items in the evaluation function,the redundant information in the subset is indirectly reduced,and the bias of the original conditional correlation to the highly redundant features is corrected.The experimental results show that the accuracy and the F1 value of the MR-DCSF algorithm have increased compared with the comparison of the four algorithms,which proves that the MR-DCSF algorithm can screen out feature subsets more reasonably.(2)For the fuzzy support vector machine based on the distance measurement of the class center,the distance between the noise point and the class center is the maximum distance,which affects the accuracy of the membership degree,and the fuzzy support vector machine cannot give the support vector higher membership,A fuzzy support vector machine DAD-FSVM(Distance And Density-Fuzzy Support Vector Machine)based on the distance metric of the class center hyperplane and the density distribution of the sample neighborhood is proposed.Firstly,the sample space is divided into three regions by using the hyperplane of two types of centers.For the two regions that do not contain the support vector,all samples in them are given constant membership values.For the samples in the middle area,according to the characteristic that the distance between the noise points in the blood routine and the heteroplane is usually less than the distance from the hyperplane of this type,a membership function based on the difference between the distance between the sample and the two types of hyperplanes is designed to distinguish noise points and normal samples.Then,taking advantage of the small density of similar samples near the support vector,the neighborhood density index is introduced into the above membership function to reduce the membership of non-support vectors and increase the membership of samples near the optimal classification hyperplane in order to distinguish non-support vector and support vector.The experimental results show that the DAD-FSVM can effectively reduce the interference of noise points and have a better classification effect compared to the other four algorithms.(3)On the basis of the above research,an intelligent diagnosis system for asthma is designed and developed.The system uses human blood routine data as input to realize the intelligent diagnosis of asthma.The system can provide doctors with reference opinions on the diagnosis and decision-making of asthma.