Research On TCD Data Classification Of Stroke And Its Implementation In Hadoop Distributed System

Stroke,as a common nervous system disease,seriously affects people's normal life because of its high incidence,high mortality,high disability rate and high recurrence rate.Transcranial Doppler(TCD)is often used in the early auxiliary diagnosis of stroke because of its noninvasive,convenient and accurate characteristics and advantages.However,at present,the analysis of TCD data by medical staff is often carried out by means of manual interpretation combined with clinical experience diagnosis,which is easy to be affected by subjective judgment and whether they have rich experience or not.How to realize the accurate analysis of stroke TCD data in order to quickly and accurately assist the diagnosis of stroke is one of the important research directions of experts at home and abroad.With the development of artificial intelligence technology,its application in medical auxiliary diagnosis has gradually become a new development trend to improve the efficiency of medical diagnosis.Therefore,how to use artificial intelligence technology to analyze the TCD data of stroke patients in order to improve the accuracy of stroke diagnosis and speed up the diagnosis efficiency become a research topic.The main purpose of this paper is to study the classification of stroke TCD data.The work mainly includes the construction of stroke TCD database,the construction of unbalanced stroke TCD data set classification model and the construction of distributed classificationmodel in massive big data scene.The specific contents of the study are as follows:A stroke TCD database based on Django architecture is designed and built.The development of the database adopts B/S architecture,including front-end interface design,Django framework development and MySQL database design.According to the follow-up requirements,the core module of the database is developed and tested,and the collection and collation of stroke TCD data is realized.An improved fuzzy support vector machine(FSVM)model for stroke TCD data classification is proposed.By analyzing and studying the traditional method of design fuzzy membership function,the existing fuzzy support vector machine is improved.The information entropy is used as the standard to measure the uncertainty of sample points.Combined with the unbalanced adjustment factor,the classification ability of FSVM for a small number of samples is improved.Through the classification experiments of the common unbalanced data set and the stroke TCD unbalanced data set,it is proved that the proposed method can effectively deal with the stroke TCD unbalanced data set.Compared with the general FSVM,in the selected common data set,the classification performance of the model using this method is improved,up to17.99%,and for the stroke TCD data set,the classification performance can be improved by 3.26%.The Hadoop platform of big data distributed processing architecture is designed and implemented,and the construction of the required distributed programming model is completed.In order to satisfy the classification of stroke TCD data,the paper constructs a Hadoop distributed processing system by using the hardware facilities of the laboratory to ensure the high efficiency and reliability of the experiment.On the one hand,the distributed grid optimizationalgorithm based on SVM is implemented.The experimental results show that the distributed grid optimization algorithm can effectively shorten the training time under the premise of the same accuracy.On the other hand,the distributed programming model of SVM is designed,the relationship between training time and dataset size is studied,and the feasibility of processing large-scale stroke TCD data based on Hadoop distributed processing architecture is verified.