Stroke Prediction Based On Machine Learning

Stroke,also known as cerebral apoplexy,has a high incidence,disability and recurrence rate,which seriously endangers people’s health.Therefore,the prediction of stroke disease is particularly important.At present,the prediction of stroke mainly adopts the methods of medical field and traditional machine learning,and the integrated learning is seldom used in the research.In this paper,SMOTE series algorithm,Easy Ensemble,Self-Paced Ensemble is used for imbalanced data processing,and SVM is used as the classification algorithm of SMOTE series algorithm.Ada Boost,XGBoost,Light GBM and Cat Boost were used as the base learners of the imbalanced ensemble classification framework.Three groups of comparative experiments were carried out to compare the model evaluation indicators such as G-mean and AUC values among each model.The results show that the clustering accuracy of self-Paced Ensemble model based on Cat Boost learner is the best,and the accuracy of the test set is 0.7671,Recall value is 0.8009,Gmean value is 0.7835,AUC value is 0.7837.Therefore,a modeling method combining self-Paced Ensemble framework with Cat Boost algorithm can predict stroke disease well.