Research On The Application Of Machine Learning Methods In Stroke Risk Prediction

Stroke occurs when a cerebral blood vessel suddenly ruptures or when a proportion of blood cannot flow to the brain due to blockage of the blood vessel.Without a blood supply,brain cells die gradually,affecting areas of the brain and causing disability.If the symptoms of stroke can be recognized early and timely intervention and treatment can be carried out,it is of great significance to improve the treatment effect and reduce the disability rate of patients.Therefore,in this paper,machine learning algorithms are used to build a stroke prediction model,which is an excellent framework for predicting the risk of stroke to assist physicians in making treatment decisions and maximizing early prevention and treatment of patients.The main research of this paper is as follows.Balance data set based on resampling method.Firstly,the healthcare-dataset-stroke-data public stroke dataset is preprocessed,including data cleaning,mean filling and data specification,and two resampling methods are used to solve the data imbalance problem of the stroke dataset.Ten feature metrics are determined as input to the machine learning algorithm model,and the stroke feature indicator is the target prediction category of the machine learning algorithm model.The balanced data processed by under-sampling and oversampling methods are input into the stroke prediction model constructed by support vector machine,random forest algorithm and logistic regression,and the prediction results before and after the optimization of the two resampling methods are compared and analyzed,and the experimental results show that the balanced data significantly improves the classification performance of the prediction model,and the accuracy,precision,recall and ROC value of the data processed by the SMOTE algorithm in the classification prediction model are the highest.Stroke risk prediction model based on machine learning algorithm.Partial machine learning algorithm is used to build a Stacking algorithm model,and the support vector machine,logistic regression,random forest and Stacking algorithms are used to build the disease prediction model for the optimized data set.The experimental results show that the accuracy,precision and ROC value of the Stacking-SMOTE algorithm prediction model are better than those of a single classifier algorithm.Moreover,the results of this study are compared with those of the same dataset,and the method of this study significantly outperforms other methods.In addition,the Stacking algorithm utilizes explainable machine learning techniques to understand the applicability of the model to clinical predictions,which can reveal the reasons behind the predicted results.A comparison of the performance of deep learning algorithms and classical machine learning algorithms in predicting stroke risk.Firstly,the published data set of The International Stroke Trial was preprocessed,and then several deep learning methods and machine learning methods were used to construct stroke risk prediction models,and their prediction results were compared.Finally,the analysis concluded that: The performance of deep learning algorithm in stroke risk prediction is not better than that of classical machine learning algorithm.