Research On The Classification Of Stroke TCD Data Based On Extreme Learning Machine

Transcranial Doppler(TCD)is widely used in the diagnosis of clinical stroke.It is of great practical significance to classify TCD data by artificial intelligence technology so as to realize the auxiliary diagnosis of stroke and save valuable medical resources.In this paper,the extreme learning machine model is studied and improved,and it is applied to the classification of stroke TCD data.the main work is as follows:In view of the fact that the extreme learning machine is sensitive to the selection of input weights and threshold parameters of the hidden layer,the bat algorithm was used to optimize the input weights and thresholds of the hidden layer of the extreme learning machine.In this paper,a stroke TCD data classification model based on bat algorithm optimization extreme learning machine was proposed.TCD data of normal people and patients with stenosis collected in Shanxi people's Hospital was classified and identified by the proposed BA-ELM model,and PSO-ELM,DE-ELM and ELM models are compared with BA-ELM.The experimental results show that the BA-ELM algorithm has the highest classification accuracy and thehighest stability,but the number of neurons in the hidden layer is the least,and the training time is lower than that of the PSO-ELM,DE-ELM model.It shows the effectiveness of BA-ELM in TCD data classification.The kernel extreme learning machine derived from the extreme learning machine was studied,and the bat algorithm was improved.Time-varying inertia weight was introduced into the speed updating formula of the bat algorithm.Four standard test functions were used to verify the strong global search ability and fast convergence speed of the improved bat algorithm.The regularization coefficient and gauss kernel function parameters of the kernel extreme learning machine were optimized by the improved bat algorithm combined with 5-fold cross validation.IBA-KELM model was used to classify four TCD datasets,and BA-KELM,BA-ELM,ELM models were compared with IBA-KELM.The results show that the performance of IBA-KELM model is better than the other three models.In view of the fact that most of the medical data is imbalanced,the imbalanced learning was studied,and a classification model of cost adjustment extreme learning machine based on maximizing Gmean was proposed,with the optimization goal of maximizing Gmean.The cost parameters of CCR-ELM were optimized by the improved bat algorithm combined with 5-fold cross validation,and imbalanced binary datasets are tested by the proposed MG-CCR-ELM model.Two evaluation indexes,classification accuracy and Gmean value,were used to analyze the classification results,and the superior imbalanced learning ability of the model was verified.Finally,the model was used to classify the TCD datasets,and good classification results are obtained,which proves the effectiveness of the proposed method for the classification of imbalanced TCD data.