基于支持向量机的Ⅱ型糖尿病判别与特征筛选

Identification of TypeⅡ Diabetes and Feature Filtration Based on Support Vector Machine

基于支持向量机理论的分类算法,由于其完善的理论基础和良好的试验结果,目前已逐渐引起国内外研究者的关注。文中采用支持向量机技术,对436个病例的14个特征建立了Ⅱ型糖尿病的“预测性”分类模型,进行全面的数据挖掘和分析,探寻与Ⅱ型糖尿病判别相关联的重要病例特征。同时,还采用决策树、多层感知器方法进行了试验,结果表明支持向量机的效果最好。当输入向量为腰围、腰围/臀围、舒张血压、年龄时,敏感度、特异性、准确率最高,分别为0.8666、0.6420、0.7014。结论表明,支持向量机对Ⅱ型糖尿病特征筛选、分类识别是一种有效的方法,为Ⅱ型糖尿病强相关病例特征鉴别探索了一条有效途径。

Support Vector Machine （ SVM）, a kind of machine learning method, can efficiently solve the classification problem. It is based on structure risk minimum principal, attracting more and more people. A classification,prediction model by using SVM for data mining, is developed ed It could identify type Ⅱ diabetes and select the best subset from the 14 features for classification in a dataset of 436 cases/controls. It turns out that the best sensitivity, specificity and accuracy are 0. 866 6, 0. 6420 and 0. 701 4 respectively, with the subset consisting of waistline, waistline/hip-girth, diastolic blood pressure and age. In addition, the performance of SVM was superior to two other modem techniques, Decision Tree and Multilayer Perceptron is found. It suggests that SVM should be an efficient method to identify type Ⅱ diabetes and select the best subset from some relevant features for the identification.