Stroke risk prediction model based on demographic data

Nowadays stroke is the third leading cause of mortality of all life periods. The statistics from the Office of the National Economic and Social Development Board (NESDB) between 1994 and 2013 found that the stroke caused 255,307 cases mortality. Period of treatment in stroke patients depends on symptom and damage of organs. It seems to be beneficial if the data analysis method likes data mining can be used to predict stroke disease to reduce amount of risk patients before initial disease. In this study, three classification algorithms: Decision Tree, Naive Bayes and Neural Network are used for predicting stroke which are model-based, superior to general statistics, and got a proper model for identification. The scope of data use is the demographic information of patients. This work was initialized by attributes selection, grouping, and resampling before modeling. This study uses the accuracy and area under ROC curve (AUC) as the indicators for evaluation. Decision tree is the most accurate and Naive Bayes is the best in AUC. The further research should also include patients´ diagnosis.