期刊
INTERDISCIPLINARY SCIENCES-COMPUTATIONAL LIFE SCIENCES
卷 10, 期 1, 页码 126-130出版社
SPRINGER HEIDELBERG
DOI: 10.1007/s12539-017-0271-2
关键词
Support vector machine essential hypertension; Predicting model; Bioinformatics
资金
- The BJAST innovation team project funding [IG201406N] Funding Source: Medline
- The BJAST Backbone Youth Project [BJAST-BYP-201620] Funding Source: Medline
- Beijing Talents Project Funding [2015400685627G238] Funding Source: Medline
Essential hypertension (EH) has become a major chronic disease around the world. To build a risk-predicting model for EH can help to interpose people's lifestyle and dietary habit to decrease the risk of getting EH. In this study, we constructed a EH risk-predicting model considering both environmental and genetic factors with support vector machine (SVM). The data were collected through Epidemiological investigation questionnaire from Beijing Chinese Han population. After data cleaning, we finally selected 9 environmental factors and 12 genetic factors to construct the predicting model based on 1200 samples, including 559 essential hypertension patients and 641 controls. Using radial basis kernel function, predictive accuracy via SVM with function with only environmental factor and only genetic factor were 72.8 and 54.4%, respectively; after considering both environmental and genetic factor the accuracy improved to 76.3%. Using the model via SVM with Laplacian function, the accuracy with only environmental factor and only genetic factor were 76.9 and 57.7%, respectively; after combining environmental and genetic factor, the accuracy improved to 80.1%. The predictive accuracy of SVM model constructed based on Laplacian function was higher than radial basis kernel function, as well as sensitivity and specificity, which were 63.3 and 86.7%, respectively. In conclusion, the model based on SVM with Laplacian kernel function had better performance in predicting risk of hypertension. And SVM model considering both environmental and genetic factors had better performance than the model with environmental or genetic factors only.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据