Assessment of the risk factors in the daily life of stroke patients based on an optimized decision tree.

BACKGROUND: Stroke is a leading cause of mortality and disability, which can be affected by people's daily living habits. OBJECTIVE: To investigate the effects of main daily living habits (smoking, drinking, diet, vegetable and fruits consumption, and exercise) on stroke risk in patients and provide the scientific basis for the assessment of the risk factors, a novel risk analysis model of the stroke is proposed. METHODS: A data mining method using decision trees which adopted the optimized C4.5 algorithm is presented. It is able to deal with the unbalanced data problem of the classification. Meanwhile, the proposed method has been verified on a clinical dataset of 23,682 patients with 21 risk factors. RESULTS: The overall accuracy and kappa coefficient for stroke risk classification has reached 84.88% and 0.7763, respectively. Through the generated knowledge rules, it demonstrates that the behavioral habits in daily life have an indirect effect on the risk of stroke. While, it has an obvious effect on stroke when hypertension, diabetes mellitus, hypercholesterolemia, and BMI risk factors exist. In addition, it was observed that the aforementioned five daily living habits have a decreased impact on the stroke. CONCLUSIONS: It is anticipated that the proposed system could help in reducing the risk, mortality, and disability of stroke, and provide clinical decision support for the treatment of stroke.